Gigacycle Fatigue in

Mechanical PracticE
 MECHANICAL ENGINEERING
A Series of Textbooks and Reference Books

Founding Editor
L. L. Faulkner
Columbus Division, Battelle Memorial Institute
and Department of Mechanical Engineering
The Ohio State University
Columbus, Ohio

1. Spring Designer’s Handbook, Harold Carlson

2. Computer-Aided Graphics and Design, Daniel L. Ryan
3. Lubrication Fundamentals, J. George Wills
4. Solar Engineering for Domestic Buildings, William A. Himmelman
5. Applied Engineering Mechanics: Statics and Dynamics, G. Boothroyd
and C. Poli
6. Centrifugal Pump Clinic, Igor J. Karassik
7. Computer-Aided Kinetics for Machine Design, Daniel L. Ryan
8. Plastics Products Design Handbook, Part A: Materials and Components;
Part B: Processes and Design for Processes, edited by Edward Miller
9. Turbomachinery: Basic Theory and Applications, Earl Logan, Jr.
10. Vibrations of Shells and Plates, Werner Soedel
11. Flat and Corrugated Diaphragm Design Handbook, Mario Di Giovanni
12. Practical Stress Analysis in Engineering Design, Alexander Blake
13. An Introduction to the Design and Behavior of Bolted Joints, John H.
Bickford
14. Optimal Engineering Design: Principles and Applications, James N. Siddall
15. Spring Manufacturing Handbook, Harold Carlson
16. Industrial Noise Control: Fundamentals and Applications, edited by
Lewis H. Bell
17. Gears and Their Vibration: A Basic Approach to Understanding Gear Noise,
J. Derek Smith
18. Chains for Power Transmission and Material Handling: Design
and Applications Handbook, American Chain Association
19. Corrosion and Corrosion Protection Handbook, edited by
Philip A. Schweitzer
20. Gear Drive Systems: Design and Application, Peter Lynwander
21. Controlling In-Plant Airborne Contaminants: Systems Design
and Calculations, John D. Constance
22. CAD/CAM Systems Planning and Implementation, Charles S. Knox
23. Probabilistic Engineering Design: Principles and Applications,
James N. Siddall
24. Traction Drives: Selection and Application, Frederick W. Heilich III
and Eugene E. Shube
25. Finite Element Methods: An Introduction, Ronald L. Huston
and Chris E. Passerello
26. Mechanical Fastening of Plastics: An Engineering Handbook,
Brayton Lincoln, Kenneth J. Gomes, and James F. Braden
27. Lubrication in Practice: Second Edition, edited by W. S. Robertson
28. Principles of Automated Drafting, Daniel L. Ryan
29. Practical Seal Design, edited by Leonard J. Martini
30. Engineering Documentation for CAD/CAM Applications, Charles S. Knox
31. Design Dimensioning with Computer Graphics Applications,
Jerome C. Lange
32. Mechanism Analysis: Simplified Graphical and Analytical Techniques,
Lyndon O. Barton
33. CAD/CAM Systems: Justification, Implementation, Productivity Measure-
ment, Edward J. Preston, George W. Crawford, and Mark E. Coticchia
34. Steam Plant Calculations Manual, V. Ganapathy
35. Design Assurance for Engineers and Managers, John A. Burgess
36. Heat Transfer Fluids and Systems for Process and Energy Applications,
Jasbir Singh
37. Potential Flows: Computer Graphic Solutions, Robert H. Kirchhoff
38. Computer-Aided Graphics and Design: Second Edition, Daniel L. Ryan
39. Electronically Controlled Proportional Valves: Selection and Application,
Michael J. Tonyan, edited by Tobi Goldoftas
40. Pressure Gauge Handbook, AMETEK, U.S. Gauge Division,
edited by Philip W. Harland
41. Fabric Filtration for Combustion Sources: Fundamentals and Basic
Technology, R. P. Donovan
42. Design of Mechanical Joints, Alexander Blake
43. CAD/CAM Dictionary, Edward J. Preston, George W. Crawford,
and Mark E. Coticchia
44. Machinery Adhesives for Locking, Retaining, and Sealing, Girard S. Haviland
45. Couplings and Joints: Design, Selection, and Application, Jon R. Mancuso
46. Shaft Alignment Handbook, John Piotrowski
47. BASIC Programs for Steam Plant Engineers: Boilers, Combustion,
Fluid Flow, and Heat Transfer, V. Ganapathy
48. Solving Mechanical Design Problems with Computer Graphics,
Jerome C. Lange
49. Plastics Gearing: Selection and Application, Clifford E. Adams
50. Clutches and Brakes: Design and Selection, William C. Orthwein
51. Transducers in Mechanical and Electronic Design, Harry L. Trietley
52. Metallurgical Applications of Shock-Wave and High-Strain-Rate Phenomena,
edited by Lawrence E. Murr, Karl P. Staudhammer, and Marc A. Meyers
53. Magnesium Products Design, Robert S. Busk
54. How to Integrate CAD/CAM Systems: Management and Technology,
William D. Engelke
55. Cam Design and Manufacture: Second Edition; with cam design software
for the IBM PC and compatibles, disk included, Preben W. Jensen
56. Solid-State AC Motor Controls: Selection and Application,
Sylvester Campbell
57. Fundamentals of Robotics, David D. Ardayfio
58. Belt Selection and Application for Engineers, edited by Wallace D. Erickson
59. Developing Three-Dimensional CAD Software with the IBM PC, C. Stan Wei
60. Organizing Data for CIM Applications, Charles S. Knox, with contributions by
Thomas C. Boos, Ross S. Culverhouse, and Paul F. Muchnicki
61. Computer-Aided Simulation in Railway Dynamics, by Rao V. Dukkipati and
Joseph R. Amyot
62. Fiber-Reinforced Composites: Materials, Manufacturing, and Design, P. K.
Mallick
63. Photoelectric Sensors and Controls: Selection and Application, Scott M. Juds
64. Finite Element Analysis with Personal Computers, Edward R. Champion, Jr.
and J. Michael Ensminger
65. Ultrasonics: Fundamentals, Technology, Applications: Second Edition,
Revised and Expanded, Dale Ensminger
66. Applied Finite Element Modeling: Practical Problem Solving for Engineers,
Jeffrey M. Steele
67. Measurement and Instrumentation in Engineering: Principles and Basic
Laboratory Experiments, Francis S. Tse and Ivan E. Morse
68. Centrifugal Pump Clinic: Second Edition, Revised and Expanded, Igor J.
Karassik
69. Practical Stress Analysis in Engineering Design: Second Edition, Revised
and Expanded, Alexander Blake
70. An Introduction to the Design and Behavior of Bolted Joints: Second Edition,
Revised and Expanded, John H. Bickford
71. High Vacuum Technology: A Practical Guide, Marsbed H. Hablanian
72. Pressure Sensors: Selection and Application, Duane Tandeske
73. Zinc Handbook: Properties, Processing, and Use in Design, Frank Porter
74. Thermal Fatigue of Metals, Andrzej Weronski and Tadeusz Hejwowski
75. Classical and Modern Mechanisms for Engineers and Inventors, Preben W.
Jensen
76. Handbook of Electronic Package Design, edited by Michael Pecht
77. Shock-Wave and High-Strain-Rate Phenomena in Materials, edited by Marc
A. Meyers, Lawrence E. Murr, and Karl P. Staudhammer
78. Industrial Refrigeration: Principles, Design and Applications, P. C. Koelet
79. Applied Combustion, Eugene L. Keating
80. Engine Oils and Automotive Lubrication, edited by Wilfried J. Bartz
81. Mechanism Analysis: Simplified and Graphical Techniques, Second Edition,
Revised and Expanded, Lyndon O. Barton
82. Fundamental Fluid Mechanics for the Practicing Engineer, James W.
Murdock
83. Fiber-Reinforced Composites: Materials, Manufacturing, and Design,
Second Edition, Revised and Expanded, P. K. Mallick
84. Numerical Methods for Engineering Applications, Edward R. Champion, Jr.
85. Turbomachinery: Basic Theory and Applications, Second Edition, Revised
and Expanded, Earl Logan, Jr.
86. Vibrations of Shells and Plates: Second Edition, Revised and Expanded,
Werner Soedel
87. Steam Plant Calculations Manual: Second Edition, Revised and Expanded,
V. Ganapathy
88. Industrial Noise Control: Fundamentals and Applications, Second Edition,
Revised and Expanded, Lewis H. Bell and Douglas H. Bell
89. Finite Elements: Their Design and Performance, Richard H. MacNeal
90. Mechanical Properties of Polymers and Composites: Second Edition,
Revised and Expanded, Lawrence E. Nielsen and Robert F. Landel
91. Mechanical Wear Prediction and Prevention, Raymond G. Bayer
92. Mechanical Power Transmission Components, edited by David W. South
and Jon R. Mancuso
93. Handbook of Turbomachinery, edited by Earl Logan, Jr.
94. Engineering Documentation Control Practices and Procedures,
Ray E. Monahan
95. Refractory Linings Thermomechanical Design and Applications,
Charles A. Schacht
96. Geometric Dimensioning and Tolerancing: Applications and Techniques
for Use in Design, Manufacturing, and Inspection, James D. Meadows
97. An Introduction to the Design and Behavior of Bolted Joints: Third Edition,
Revised and Expanded, John H. Bickford
98. Shaft Alignment Handbook: Second Edition, Revised and Expanded,
John Piotrowski
99. Computer-Aided Design of Polymer-Matrix Composite Structures,
edited by Suong Van Hoa
100. Friction Science and Technology, Peter J. Blau
101. Introduction to Plastics and Composites: Mechanical Properties
and Engineering Applications, Edward Miller
102. Practical Fracture Mechanics in Design, Alexander Blake
103. Pump Characteristics and Applications, Michael W. Volk
104. Optical Principles and Technology for Engineers, James E. Stewart
105. Optimizing the Shape of Mechanical Elements and Structures, A. A. Seireg
and Jorge Rodriguez
106. Kinematics and Dynamics of Machinery, Vladimír Stejskal
and Michael Valásek
107. Shaft Seals for Dynamic Applications, Les Horve
108. Reliability-Based Mechanical Design, edited by Thomas A. Cruse
109. Mechanical Fastening, Joining, and Assembly, James A. Speck
110. Turbomachinery Fluid Dynamics and Heat Transfer, edited by Chunill Hah
111. High-Vacuum Technology: A Practical Guide, Second Edition, Revised
and Expanded, Marsbed H. Hablanian
112. Geometric Dimensioning and Tolerancing: Workbook and Answerbook,
James D. Meadows
113. Handbook of Materials Selection for Engineering Applications, edited by
G. T. Murray
114. Handbook of Thermoplastic Piping System Design, Thomas Sixsmith
and Reinhard Hanselka
115. Practical Guide to Finite Elements: A Solid Mechanics Approach,
Steven M. Lepi
116. Applied Computational Fluid Dynamics, edited by Vijay K. Garg
117. Fluid Sealing Technology, Heinz K. Muller and Bernard S. Nau
118. Friction and Lubrication in Mechanical Design, A. A. Seireg
119. Influence Functions and Matrices, Yuri A. Melnikov
120. Mechanical Analysis of Electronic Packaging Systems, Stephen A. McKeown
121. Couplings and Joints: Design, Selection, and Application, Second Edition,
Revised and Expanded, Jon R. Mancuso
122. Thermodynamics: Processes and Applications, Earl Logan, Jr.
123. Gear Noise and Vibration, J. Derek Smith
124. Practical Fluid Mechanics for Engineering Applications, John J. Bloomer
125. Handbook of Hydraulic Fluid Technology, edited by George E. Totten
126. Heat Exchanger Design Handbook, T. Kuppan
127. Designing for Product Sound Quality, Richard H. Lyon
128. Probability Applications in Mechanical Design, Franklin E. Fisher and
Joy R. Fisher
129. Nickel Alloys, edited by Ulrich Heubner
130. Rotating Machinery Vibration: Problem Analysis and Troubleshooting,
Maurice L. Adams, Jr.
131. Formulas for Dynamic Analysis, Ronald L. Huston and C. Q. Liu
132. Handbook of Machinery Dynamics, Lynn L. Faulkner and Earl Logan, Jr.
133. Rapid Prototyping Technology: Selection and Application,
Kenneth G. Cooper
134. Reciprocating Machinery Dynamics: Design and Analysis,
Abdulla S. Rangwala
135. Maintenance Excellence: Optimizing Equipment Life-Cycle Decisions, edited
by John D. Campbell and Andrew K. S. Jardine
136. Practical Guide to Industrial Boiler Systems, Ralph L. Vandagriff
137. Lubrication Fundamentals: Second Edition, Revised and Expanded,
D. M. Pirro and A. A. Wessol
138. Mechanical Life Cycle Handbook: Good Environmental Design
and Manufacturing, edited by Mahendra S. Hundal
139. Micromachining of Engineering Materials, edited by Joseph McGeough
140. Control Strategies for Dynamic Systems: Design and Implementation,
John H. Lumkes, Jr.
141. Practical Guide to Pressure Vessel Manufacturing, Sunil Pullarcot
142. Nondestructive Evaluation: Theory, Techniques, and Applications,
edited by Peter J. Shull
143. Diesel Engine Engineering: Thermodynamics, Dynamics, Design,
and Control, Andrei Makartchouk
144. Handbook of Machine Tool Analysis, Ioan D. Marinescu, Constantin Ispas,
and Dan Boboc
145. Implementing Concurrent Engineering in Small Companies, Susan Carlson
Skalak
146. Practical Guide to the Packaging of Electronics: Thermal and Mechanical
Design and Analysis, Ali Jamnia
147. Bearing Design in Machinery: Engineering Tribology and Lubrication,
Avraham Harnoy
148. Mechanical Reliability Improvement: Probability and Statistics for
Experimental Testing, R. E. Little
149. Industrial Boilers and Heat Recovery Steam Generators: Design,
Applications, and Calculations, V. Ganapathy
150. The CAD Guidebook: A Basic Manual for Understanding and Improving
Computer-Aided Design, Stephen J. Schoonmaker
151. Industrial Noise Control and Acoustics, Randall F. Barron
152. Mechanical Properties of Engineered Materials, Wolé Soboyejo
153. Reliability Verification, Testing, and Analysis in Engineering Design,
Gary S. Wasserman
154. Fundamental Mechanics of Fluids: Third Edition, I. G. Currie
155. Intermediate Heat Transfer, Kau-Fui Vincent Wong
156. HVAC Water Chillers and Cooling Towers: Fundamentals, Application,
and Operation, Herbert W. Stanford III
157. Gear Noise and Vibration: Second Edition, Revised and Expanded,
J. Derek Smith
158. Handbook of Turbomachinery: Second Edition, Revised and Expanded,
edited by Earl Logan, Jr. and Ramendra Roy
159. Piping and Pipeline Engineering: Design, Construction, Maintenance,
Integrity, and Repair, George A. Antaki
160. Turbomachinery: Design and Theory, Rama S. R. Gorla
and Aijaz Ahmed Khan
161. Target Costing: Market-Driven Product Design, M. Bradford Clifton,
Henry M. B. Bird, Robert E. Albano, and Wesley P. Townsend
162. Fluidized Bed Combustion, Simeon N. Oka
163. Theory of Dimensioning: An Introduction to Parameterizing Geometric
Models, Vijay Srinivasan
164. Handbook of Mechanical Alloy Design, edited by George E. Totten, Lin Xie,
and Kiyoshi Funatani
165. Structural Analysis of Polymeric Composite Materials, Mark E. Tuttle
166. Modeling and Simulation for Material Selection and Mechanical Design,
edited by George E. Totten, Lin Xie, and Kiyoshi Funatani
167. Handbook of Pneumatic Conveying Engineering, David Mills, Mark G. Jones,
and Vijay K. Agarwal
168. Clutches and Brakes: Design and Selection, Second Edition,
William C. Orthwein
169. Fundamentals of Fluid Film Lubrication: Second Edition,
Bernard J. Hamrock, Steven R. Schmid, and Bo O. Jacobson
170. Handbook of Lead-Free Solder Technology for Microelectronic Assemblies,
edited by Karl J. Puttlitz and Kathleen A. Stalter
171. Vehicle Stability, Dean Karnopp
172. Mechanical Wear Fundamentals and Testing: Second Edition, Revised
and Expanded, Raymond G. Bayer
173. Liquid Pipeline Hydraulics, E. Shashi Menon
174. Solid Fuels Combustion and Gasification, Marcio L. de Souza-Santos
175. Mechanical Tolerance Stackup and Analysis, Bryan R. Fischer
176. Engineering Design for Wear, Raymond G. Bayer
177. Vibrations of Shells and Plates: Third Edition, Revised and Expanded,
Werner Soedel
178. Refractories Handbook, edited by Charles A. Schacht
179. Practical Engineering Failure Analysis, Hani M. Tawancy, Anwar Ul-Hamid,
and Nureddin M. Abbas
180. Mechanical Alloying and Milling, C. Suryanarayana
181. Mechanical Vibration: Analysis, Uncertainties, and Control, Second Edition,
Revised and Expanded, Haym Benaroya
182. Design of Automatic Machinery, Stephen J. Derby
183. Practical Fracture Mechanics in Design: Second Edition, Revised
and Expanded, Arun Shukla
184. Practical Guide to Designed Experiments, Paul D. Funkenbusch
 Additional Volumes in Preparation

Mechanical Engineering Software

Spring Design with an IBM PC, Al Dietrich

Mechanical Design Failure Analysis: With Failure Analysis System Software

for the IBM PC, David G. Ullman
 Gigacycle Fatigue in
Mechanical PracticE

Claude bathias paul C. Paris

Professor of Mechanics Senior Professor of Mechanics
Institute for Technology Department of Mechanical
and Advanced Materials (ITMA) and Aeronautical Engineering
Conservatoire National Washington University in St. Louis
des Arts et Métiers St. Louis, Missouri, U.S.A.
Paris, France

MARCEL DEKKER NEW YORK

Cover: Upper photo: Modern TGV high-speed train. Courtesy of Israel
Marines (CNAM/ITMA, Paris, France).
Lower photo: Stephenson locomotive ca. 1833, © Musée des arts
et métiers/S. Pelly, Paris, France.

Although great care has been taken to provide accurate and current information,
neither the author(s) nor the publisher, nor anyone else associated with this
publication, shall be liable for any loss, damage, or liability directly or indirectly
caused or alleged to be caused by this book. The material contained herein is not
intended to provide specific advice or recommendations for any specific situation.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress.

ISBN: 0-8247-2313-9

Marcel Dekker, 270 Madison Avenue, New York, NY 10016, USA

http://www.dekker.com

Distribution center:
Marcel Dekker, Cimarron Road, Monticello, NY 12701 USA

Copyright © 2005 by Marcel Dekker. All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form
or by any means, electronic or mechanical, including photocopying, microfilming,
and recording, or by any information storage and retrieval system, without
permission in writing from the publisher.

Current printing (last digit):

10 9 8 7 6 5 4 3 2 1

PRINTED IN THE UNITED STATES OF AMERICA

 Dedication

We dedicate this book to the patient encouragement of our

wives—Marie-Claude Bathias and Barbara L. Paris. We also
include in our dedication our children Anne Potter, Claire
Besset, Gail Paris, and Dr. Anthony J. Paris, who have also
greatly inspired our effort.

i
 Contents

Preface vii
Acknowledgments ix
Table of Notation xi

1 Introduction 1

2 Ultrasonic Fatigue Concepts 9

2.1 Introduction 9
2.2 Longitudinal elastic waves and resonance frequency 11
2.3 Analytical solution for the variable section specimen 15
2.4 Stress magnification factor 21
2.5 Analytical solution of resonance length 22
2.6 Methods for calculating crack tip stress intensity factor 30

3 Testing Machines and Their Performance 51

3.1 Introduction 51
3.2 Basic structure 52
3.3 Nonsymmetrical and variable amplitude test equipment 55
3.4 Computer control system 56
3.5 High temperature test equipment 67
3.6 Low temperature test equipment 69
iii
iv Contents

3.7 Thin sheet test equipment 71

3.8 High pressure piezo-electric fatigue machine 73
3.9 Non-axial test equipment 76

4 S-N Curve and Fatigue Strength 87

4.1 Introduction 87
4.2 Ferrous materials 90
4.3 Aluminium matrix composite 105
4.4 Non-ferrous alloys 109
4.5 Alloys at cryogenic temperature 118
4.6 N18 alloy at high temperature 123
4.7 Rotating-bending internal crack stress correction 125
4.8 Ti-Al intermetallic alloys 127

5 Crack Growth and Threshold 133

5.1 Titanium alloys 146
5.2 Nickel-based alloys 151
5.3 Aluminium alloys 159
5.4 Materials of b.c.c. and f.c.c crystalline structure 173
5.5 Low carbon steel sheet 179
5.6 Austenitic stainless steel 181
5.7 Spheroidal graphite cast iron (SGI) 186
5.8 Database of threshold SIF DKth 189
5.9 Other applications: Fretting fatigue 194

6 Frequency and Environmental Effects 207

6.1 Frequency effect 207
6.2 Heat effect 216
6.3 Cryogenic temperature 220
6.4 Environmental effects 223
6.5 S-N curve at room temperature and high pressure
hydrogen for Ti-6A4V 227

7 Microstructural Aspects and Damage to Materials

in the Gigacycle Regime 229
7.1 Gigacycle S-N curve shape 229
7.2 Mechanical aspects of initiation between
106 and 109 cycles 231
7.3 Initiation zone for low cycle to gigacycle failures 241
7.4 Initiation mechanisms at 109 cycles 242
7.5 Role of inclusions 243
Contents v

7.6 Gigacycle fatigue of alloys without inclusions 249

7.7 General discussion of the gigacycle
fatigue mechanisms 253

Appendix 1 Stress Calibration 259

A1.1 Amplifying horn 260
A1.2 First calibration 262
A1.3 Second calibration 265
A1.4 Third calibration 269

Appendix 2 Remarks on the Statistical Prediction 273

A2.1 Remarks on the statistical analysis in
the megacycle regime 274

References 279

Listing of Materials Tested in Gigacycle Fatigue 293

Index 295
 Preface

The photos of locomotives on the cover of this book illustrate

the time period of interest in significant studies of metal fati-
gue. The older Stephenson Locomotive 020 of 1833, displayed
at the CNAM museum in Paris, is an example from the era of
first recognition of fatigue. The French TGV high speed trains
of the late 20th century evolved this interest into the ‘‘giga-
cycle regime’’.
Therefore, over 150 years ago, A. Wohler began his stu-
dies of metal fatigue for application to rail car axles. Others,
such as Bauschinger, also examined fatigue phenomena later
in the 19th century but were limited by the test equipment
and instrumentation available. At about the beginning of
the 20th century, it was found that initiation of fatigue from
a smooth surface was preceded by plastic slip and later by
reversals of this slip at the surface to form an intrusion lead-
ing to a crack growing failure. It was thereafter frequently
concluded that below a certain stress level—the so called
‘‘endurance limit’’—this reversing slip and=or crack initiation
would not occur and fatigue failure could be avoided. This
concept assumed that fatigue crack initiation from imperfec-
vii
viii Preface

tions in the material or due to manufacturing could be

avoided and was accepted well beyond the middle of the
20th century.
However, improvements in test equipment and methods,
as well the motivation for improved metal structures such as
aircraft, commercial power generators, high speed trains, etc.,
led to more intensive analyses of fatigue. In the 1960s and
early 1970s the ‘‘damage tolerance’’ approach to fatigue was
developed, which assumed crack-like flaws initially in a struc-
ture and calculated a safe crack growth life. Some structural
situations required showing that present cracks would not
grow at all or that they would be below the ‘‘crack growth
threshold’’ in size and imposed stress. Both of these methods
employed so called ‘‘fracture mechanics’’ methods in their
approach. For components sustaining extremely high num-
bers of cycles of loads, manufacture without significant flaws
and holding the stress levels low enough to avoid initiation
remains the dominant method of approach. This motivates
studies of ‘‘gigacycle fatigue’’. These requirements have also
motivated this book and its presentation of results of fatigue
under conditions up to 1010 cycles of loading.
The development of piezo-electrically loaded fatigue
machines capable of testing at the ultrasonic frequency of
20 kHz or more in the 1980s made it practical to test to such
high number of cycles of load for fatigue initiation as well as
for very slowly growing cracks to establish thresholds effi-
ciently. Consequently, this is a book that is an exposition of
the new concepts and data established by these new testing
techniques of the last 20 years. This book not only presents
results but also discusses in detail the design of these
machines and the methods of using them to explore high cycle
fatigue phenomena. Environmental testing and results
including vacuum and temperature effects are presented
and discussed. Load ratio effects and variable amplitude
loading are also included.

Paul C. Paris and Claude Bathias

August 2004
 Acknowledgments

The authors wish to acknowledge the special efforts of

Dr. Hiroshi Tada in editorial work and checking the mathe-
matical accuracy of this book. We also wish to express special
thanks to Delphine Martin, Fabrice Montembault and Emin
Bayraktar for their patient assistance in preparing several
drafts of the manuscript as it evolved. The research efforts
of graduate students at CNAM who prepared doctoral disser-
tations on gigacycle fatigue providing data and experience
reflected in the book include K. Saanouni from Tunisia
(1981): X. Kong (1986), J. Ni (1992), T. Wu (1994), H. Tao
(1996), Q. Wang (1998), Z. Sun (2000), and H. Xue (2004) from
China: G. Thanigaiyarasu from Pakistan (1987): K. El Alami
from Maroco (1995), G. Jago (1996), and J. Bonis from France
(1997); and I. Marines from Mexico (2004) are due much
thanks. We are also thankful for the suggestions of numerous
colleagues.
We wish to also acknowledge the assistance of our long-
term friend, John Corrigan, for his help in publishing this
book. Our special thanks to Joanne Jay of Dekker for extraor-
dinary editorial effort in expediting publication.
ix
 Table of Notation

a crack size or radius

b Burger’s vector
a and b constants in S-N curve formulae
da Da
dN or DN crack growth rate
b,h rectangular cross section dimensions
c wave velocity
e, e engineering or true strain
ef, ef strain at fracture
e_ ðx; tÞ strain rate at a specific location and time
o
f ¼ 2p frequency (cycles per second)
k ¼ oc wave vector
l wave length
m exponent in Paris crack growth law or meters
S0 ðxÞ
pðxÞ SðxÞ
r parameter in specimen shape or distance from
crack tip
s standard deviation or seconds
t time
v charge or displacement of crack surface
u,v,w rectangular components of displacements
x,y,z Cartesian coordinates
Ai, Bi constants in displacement expressions

xi
xii Notation

A0 displacement at the end of a bar

A=D, D=A analog to digital or vice versa
Bi thickness of specimen
C1 ; C2 ; C3 ; C4 Elastic constants
Cþþ computer software
0
C; K temperature Celsius or Kelvin
E modulus of elasticity
Ed dynamic modulus of elasticity
F force
Famp ¼ VV12 amplification factor of a horn
G elastic energy release rate at a crack tip
Hv Vickers hardness
J2 a twelve prong connector plug
Hz Hertz frequency
K1 crack tip stress intensity factor
DK, DKth, DKeff stress intensity range (threshold or effective)
L1 ; L2 ; L length (resonance, exponential, or specimen)
N; Nf ; Ni ; Np number of cycles (to failure, to initiation,
or in crack propagation)
Pa Pascal
R cyclic load ratio
RA reduction in area
R1, R2 radius of the specimen at the center and end
S(x) cross sectional area at location x
S-N fatigue stress vs. number of cycles curves
U(x) displacement at location x
UTS ultimate tensile strength
V voltage input
½K
2 ; ½M
c ; ½K
g Matrix (elementary rigidity, elementary mass,
or geometrical)
a; b microstructure in titanium alloys or parameters
in vibration equation solutions
a; b; C constants in Murakami’s equation
l; Dli eigenvalues
r mass density
o frequency
x; Z non-dimensional coordinates
s applied normal stress
sx ; sy ; txy rectangular components of plane stress
sa alternating stress
syp yield point stress
sw ; sd fatigue failure stress
t shear stress
EðxÞ; SðxÞ strain or stress on the reduced section of a specimen
1
Introduction

Initially, it is of interest to note that many structural compo-

nents sustain far beyond 107 cycles of loading, but materials
characterization and fatigue predictions are normally based
upon data limited to between 106 and 107 cycles. This is
because standard fatigue testing equipment prior to the past
decades was limited in speed to less than 200 cycles per
second. Therefore, testing beyond 107 cycles was very time
consuming. However, the fatigue life of current automobile
engines ranges around 108 cycles; big diesel engines for
ships or high speed trains have ranges to 109 cycles. It is
further noted that at this time interest in fatigue life extends
to about 1010 cycles, for example, in turbine engine compo-
nents (Figure 1.1).
From a historical perspective, it was established for the
first time in the mid-1980s by several Japanese researchers
(Ebara, 1987; Kikukawa, 1965; Murakami, 1994) that struc-
tural metal alloys can fail after 107 cycles. More recently,
the phenomena of gigacycle fatigue failures in many
alloys up to 1010 cycles has been extensively established by

1
2 Chapter 1

Figure 1.1 Fatigue life of machines and components.

C. Bathias and co-workers (Kong, 1987; Ni, 1991; Thani-

gaiyarasu, 1988; Wu, 1991).
The S-N (stress=cycles) curve is often still assumed to be
a rectangular hyperbolic relationship, but in reality there is
not a horizontal asymptote. This means that fatigue initiation
mechanisms from 106 to beyond 109 cycles are a topic of great
interest for advanced structural technologies. Consequently
the S-N curve, since it is not asymptotic, must be determined
in order to guarantee the real fatigue strength in the very
high cycle regime.
The preceding view was based on assuming that fatigue
initiation mechanisms leading to growing cracks must be
avoided. However, if an initiated crack or pre-existing crack-like
flaw grows at a very small rate, which will allow a sufficient life,
then failure may also be avoided. For this reason the very slow
growth of cracks in the threshold regime is also of interest
herein. In the 1960s Paris and co-workers (Lindner, 1965)
observed threshold region crack growth rates as low as
0.6  10 11 meters per cycle. At such rates it would usually take
well over 108 cycles to grow to failure. Consequently, the subject
of threshold level crack growth rates is discussed and data
developed by high speed equipment are presented in Chapter 5.
Both initiation and growth of fatigue cracks are important
to develop a full understanding of very high cycle fatigue. If
stresses are low enough to prevent initiation from usual mate-
rial initiation mechanisms, then assuring below-threshold
Introduction 3

Figure 1.2 Typical S-N curve as defined by international standar-

dization.

conditions for any possible defects will guarantee

sufficient life. Design and production conditions will dictate
that considering one or the other alone will be sufficient to
avoid failure in practice. Emphasis will first be placed on
S-N testing curve techniques and results. (Thresholds and
crack growth behavior will be mainly deferred to Chapter 5.)
When the fatigue curve or S-N curve is defined, it is usually
done in reference to carbon steels. The S-N curve data are
generally limited to 107 cycles and it is presumed, according to
the standard, that a horizontal asymptote allows determination
of a fatigue limit value for an alternating stress between 106
and 107 cycles. Beyond 107 cycles (Figure 1.2), it is normally con-
sidered that the fatigue life is infinite. However for other metal
alloys, it is assumed that the asymptote of the S-N curve is not
horizontal.
For fatigue limits to 109 cycles a few results can be
observed in the references (Bathias, 1993, 1998, 2004; Stanzl,
1996). Until recently, the shape of the S-N curve beyond 107
cycles was predicted by using probabilistic methods, which
4 Chapter 1

Figure 1.3 Isoprobability of failure.

is also true for the fatigue limit. In principle, the fatigue limit
is given for a specific number of cycles to failure. Using, for
example, the staircase method, the fatigue limit is given by
the average alternating stress sD and the probability of frac-
ture is given by the standard deviation (s) of the scatter.
A classical way to determine the infinite fatigue life is to
use a Gaussian function. Roughly speaking, it is said that
the mean endurance limit stress sD, minus 3s gives a prob-
ability of fracture close to zero (Figure 1.3). Assuming s is
equal to 10 MPa, the true infinite fatigue limit should be sD
30 MPa. However, experiments data herein will show that
for many alloys between sD for 106 and sD for 109 the
difference is greater than 30 MPa.
This so-called SD approach to the average fatigue limit is
certainly not the best way to reduce the risk of rupture in
fatigue (Figure 1.4) and meant as a last resort. Only direct
experience can remove this ambiguity by providing some
accelerated tests of fatigue.
From a basic point of view, it seems that it is better to
determine the real fatigue strength and not an estimated
fatigue limit for a given number of cycles, especially in the
gigacycle regime (Figure 1.5).
Introduction 5

Figure 1.4 Safe fatigue curve.

Today, this is possible since piezoelectric fatigue

machines are very reliable and capable of producing 1010
cycles in less than 1 week (at 20 kHz), whereas the
conventional systems require more than 3 years of testing
for only one sample (at <200 Hz).

Figure 1.5 The concept of gigacycle S-N curve.

6 Chapter 1

To reemphasize the present situation, it is evident that

the historical concept of a fatigue limit is bound to the
hypothesis of the existence of a horizontal asymptote on the
S-N curve between 106 to 107 cycles (for example, see Figure
1.1). Therefore, if a sample reaches 107 cycles and is not
broken, it is considered to have an infinite life. That is a
convenient and economical assumption, but it is not a rigor-
ous approach. It is important to understand that the staircase
method is popular today to determine an assumed fatigue
limit only because of the convenience of this approximation.
A fatigue limit determined by this method to 107 cycles
requires 30 hours of test with a machine working at 100 Hz
for only one sample. To reach 109 cycles, 3000 hours of testing
would be necessary, which is very time consuming and
expensive.
It is of great importance to understand and predict a
fatigue life in terms of crack initiation and small crack propa-
gation. It has been generally accepted that at high stress
levels, fatigue life is determined primarily by crack growth,
while at low stress levels, the life span is mainly consumed
by the process of crack initiation. Several authors have
demonstrated that the portion of life attributed to crack
nucleation is above 90% in the high cycle regime (106 to 107
cycles) for steel, aluminium, titanium, and nickel alloys.
In cases where the crack nucleates from a defect, such as
an inclusion or pore, it is said that a relation should exist
between the fatigue limit and the crack growth threshold.
However, the relation between crack growth and initiation
is not obvious for many reasons. First, it is not certain that
a sharp defect implies immediate fatigue crack growth from
the very first cycle. Second, when a defect is small, a short
crack does not grow at the same rate as a long crack. In par-
ticular the effects of load ratio, (R), and closure depend on the
crack length. Thus, the relationship between fatigue crack
growth threshold (DKth) and sD remains to be developed.
The relationship between sD and DKth must be established
in the gigacycle regime if a relationship does in fact exist.
Indeed, the experiments show that there are several crack
initiation mechanisms dependent upon the alloys and particu-
Introduction 7

lar types of defects. Consequently, it seems that there is no

general relation between DKth and sD even at 109 cycles.
But in the case when initiation depends on inclusions, a
Murakami type model (see Section 4.2 and Section 7.52)
appears to be sufficient.
For all the reasons stated above, it is necessary to apply
an accelerated fatigue testing method by ultrasonic fatigue
test techniques to investigate the behavior of the S-N curve
into the gigacycle regime. In the earlier chapters of this book,
ultrasonic testing methods will be presented.
2
Ultrasonic Fatigue Concepts

2.1. INTRODUCTION

The ultrasonic fatigue test method differs from the conven-

tional fatigue test method that has frequency limited to
100 Hz of cyclic stressing of material. The frequency of ultra-
sonic fatigue testing ranges from 15 kHz to 30 kHz, with a
typical frequency being 20 kHz. With this high frequency,
the time and cost to obtain a fatigue limit (if one does exist)
or crack growth rate threshold data can be dramatically
reduced. For instance, the test time for 107 cycles is within
9 minutes by ultrasonic method, while conventional fatigue
testing at 100 Hz will take about 12 days. For even higher
cycles—for instance, 109—the ultrasonic method requires
only 14 hours, whereas it would take more than 3 years at
100 Hz for a single specimen. The ultrasonic method also
provides a reliable way of testing at the extremely small
rates of crack growth in the threshold regime, for example
in the range of 10
9 mm=s to 10
11 mm=s.
The application of ultrasonic fatigue testing started near
the beginning of the 20th century (Hopkinson, 1911). Up until
9
10 Chapter 2

then, the highest frequency of fatigue testing with a mechani-

cally driven system did not exceed 33 Hz. That year Hopkin-
son developed a first electromagnetic resonance system of
116 Hz. (Jenkin, 1925) used similar techniques to test
copper, iron, and steel wires at the frequency of 2.5 kHz. In
1929, the test machine of Jenkin and Lehmann (Jenkin,
1929) reached the frequency of 10 kHz with a pulsating air
resonance system. In (Mason, 1950) the test machine marked
an important point in the development of ultrasonic fatigue
testing techniques. He introduced the piezo-electric and mag-
netostrictive types of transducers capable of translating
20 kHz electrical voltage signals into displacement controlled
20 kHz mechanical vibration, and used high power 20 kHz
ultrasonic waves to induce fracture of materials in fatigue.
Shortly afterward, even higher frequencies of fatigue testing
were reached, for example 92 kHz (Girard, 1959) and 199 kHz
(Kikukawa, 1965). However, the design of Mason’s 20 kHz
machine has been used as the basis for most modern ultra-
sonic fatigue testing machines.
In (Neppiras, 1959) a proposition to apply ultrasound to
the determination of S-N curves initiated a series of research
work, most of which aimed at developing methods of measur-
ing fatigue life and fatigue limits under constant amplitude
loading conditions at R ¼
1.
In (Mitsche, 1973) was the first to use ultrasound for the
purpose of fatigue crack propagation testing. They gave the
first ultrasonic data on the curves of Da=DN versus DK, using
standard procedures for low frequency loading (conventional)
as a first approximation to calculate DK. More accurate meth-
ods of calculating DK in high frequency appeared rather
slowly. Important work, such as that from (Saanouni, 1982),
(Shoeck, 1982), (Kong, 1991), (Mayer, 1993), (Wu, 1994),
and (Ni, 1996), was published between 1982 and 1996.
Fatigue properties of materials under variable amplitude
loading are also of great interest in practice. This is another
domain to which the ultrasonic fatigue concept can contri-
bute. Conventional tests often involve the modification of
the load time sequence; however, ultrasonic tests can follow
exactly the load sequences desired. Bathias with (Wu, 1994)
Ultrasonic Fatigue Concepts 11

and (Ni, 1994) developed computer control systems for ultra-

sonic fatigue testing. For this purpose PC 486 (or higher)
computers were used, since the speed of these computers
is adequate to drive a piezo-electrical fatigue machine.
Industrial applications of the ultrasonic fatigue data
(constant amplitude and random amplitude) now have been
extended to aircraft, automobile, railway, offshore, and other
structures. Among them, some in-service loading conditions of
aircraft do fall into the ultrasonic frequency region. In these
domains, the importance of ultrasonic fatigue technologies
is, of course, more direct.
There are several articles reviewing the development of
ultrasonic fatigue techniques; we refer the reader to (Bathias,
2002); (Mayer, 1999); and (Stanzl, 1996).

2.2. LONGITUDINAL ELASTIC WAVES AND

RESONANCE FREQUENCY

In conventional fatigue tests, the frequency is that of the

external load system of the test machine, which is different
from the natural frequencies of the specimen. In other words,
the specimen is in forced vibration. An ultrasonic fatigue test
differs from this in that the external frequency supplied by
the test machine must be one of the natural frequencies of
the specimen. This is the definition of free vibration. To better
understand this phenomenon, it is worth briefly recalling
elastic wave theory.
The differential equations for a general three-dimensional
isotropic elastic body in a Cartesian co-ordinate system are


@2u E 1 @e 2
r 2 ¼ þr u ð2:1aÞ
@t ð1 þ nÞ 1
2n @x


@2v E 1 @e 2
r 2 ¼ þr v ð2:1bÞ
@t ð1 þ nÞ 1
2n @y


@2w E 1 @e 2
r ¼ þ r w ð2:1cÞ
@t2 ð1 þ nÞ 1
2n @z
12 Chapter 2

where u, v, and w are displacements along x, y, and z respec-

tively, E and n the Young’s modulus and Poisson’s ratio, r the
mass density, H2 the Laplacian, and

@u @v @w
e¼ þ þ ð2:2Þ
@x @y @z

is the volume dilatation.

Elastic wave theory indicates that the following two
types of wave may exist in an infinite isotropic elastic body.

2.2.1. Longitudinal Wave

For a longitudinal wave, the curl of the displacement field is
zero. The velocity of wave propagation is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Eð1
nÞ
c¼ ð2:3Þ
ð1 þ nÞð1
2nÞr

2.2.2. Transverse Wave

For a transverse wave, the volume dilatation e vanishes. The
velocity of wave propagation is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
E
c¼ ð2:4Þ
2ð1 þ nÞr

If there is a boundary, a surface wave may also be produced.

This wave is similar to the gravity surface wave in a fluid. The
amplitude of vibration decreases rapidly with the distance
from the surface, and the wave velocity is less than the
velocity inside the body.
To simplify the discussion, let us start with a one-
dimensional specimen of straight cylinder. As shown in
Figure 2.1, a longitudinal elastic wave comes from one end
of the bar and travels through the length l, then it is reflected
from the other end and returns to the initial place of entrance.
Ultrasonic Fatigue Concepts 13

Figure 2.1 Displacement and strain variation along an elastic bar.

The wave velocity c will be determined directly from Eq. 2.3

with n ¼ 0 for consideration of this one-dimensional example.
sffiffiffiffi
E
c¼ ð2:5Þ
r

Also the differential Eqs. 2.1a to 2.1c reduce to a single

equation

@2u E @2u
¼ ð2:6Þ
@t2 r @x2

The solution of Eq. 2.6 is given by

X
1
u¼ un ðx; tÞ ð2:7Þ
n¼1

where


npct npct npx
un ðx; tÞ ¼ An
1 cos þ Bn
1 sin cos ð2:8Þ
l l l
The boundary conditions of ultrasonic fatigue testing
require the displacement to be maximum at both ends
whereas the strain vanishes at the same places. That is


@u
¼0 ð2:9Þ
@x x¼0;l
14 Chapter 2

Thus Eq. 2.8 for the first mode of vibration becomes

uðx; tÞ ¼ A0 cosðkxÞ sinðotÞ ð2:10Þ

where
p pc
k¼ ; o¼ ð2:11Þ
l l
The amplitude of vibration at each point along the bar is
UðxÞ ¼ A0 cosðkxÞ ð2:12Þ

where A0 is the displacement amplitude at the end of the bar.

The strain e of each point is given by
eðx; tÞ ¼
kA0 sinðkxÞ sinðotÞ ð2:13Þ

with its maximum

eðxÞ ¼
kA0 sinðkxÞ ð2:14Þ
The strain rate is
e_ ðx; tÞ ¼
koA0 sinðkxÞ cosðotÞ ð2:15Þ
with its maximum

e_ ðxÞ ¼
koA0 sinðkxÞ ð2:16Þ

From Eqs. 2.5 and 2.11

sffiffiffiffiffiffi
1 Ed
l¼ ð2:17Þ
2f r

for the first mode of vibration, where f ¼ o=2p is the

frequency and Ed ¼ the dynamic elastic modulus for consid-
eration of dynamic effects.
In conclusion, the length of resonance of the one-
dimensional specimen is given by Eq. 2.17 for the first vibra-
tion mode. The displacement node (where displacement
vanishes) at the center of the specimen corresponds to the
maxima of the strain, stress and strain rate. At both ends,
we have the maximum of displacement and the nodes of
Ultrasonic Fatigue Concepts 15

strain, stress and strain rate. That is

u ¼ 0; e ¼
kA0 ; s ¼
Ed kA0 ; e_ ¼
koA0
l ð2:18aÞ
for x ¼
2

U ¼ A0 ; e ¼ 0; s ¼ 0; e_ ¼ 0; for x ¼ 0; l ð2:18bÞ

Eq. 2.17 indicates an important fact: The resonance length is

inversely proportional to the frequency. This explains why
some very high frequencies—for example, 92 kHz and
199 kHz mentioned in the previous section—are not practic-
able. For example, for a typical steel specimen of uniform
section with E ¼ 200 000 MPa, r ¼ 7800 kg=m3, we have (from
Eq. 2.17)
2:5  106
l ðmmÞ
f
and
f (kHz) l (mm)

20 127
92 27.5
199 13.2

The last two values of l obviously lead to difficulties in

machining, displacement, or strain measurements, as well
as energy dissipation.

2.3. ANALYTICAL SOLUTION FOR THE

VARIABLE SECTION SPECIMEN

For a fatigue specimen of variable section, amplitudes of

strain and stress vary at each section. Normally, the reso-
nance amplitude that is a function of the specimen geometry
is determined numerically.
In order to obtain stress concentration in the middle of
the specimen to expedite a fatigue test, the specimen for
ultrasonic fatigue initiation or fatigue crack growth is, in
16 Chapter 2

most cases, designed with a reduced section in the center as

shown in Figure 2.2. We call the length L1 the resonance
length, the determination of which involves a numerical
approach, such as the finite element method (FEM). But if
the center part is in an exponential form, an analytical solu-
tion of the resonance length can be obtained (Kong, 1987).
The longitudinal wave equation for a specimen with a
varying cross section can be written as
@ 2 u @f
rSðxÞ ¼ ð2:19Þ
@t2 @x
where S(x) is the area of cross section at position x, and
@u
f ¼ Ed SðxÞ ð2:20Þ
@x
is the force acting on the section. Thus we have
 
@ 2 uðx; tÞ 2 @uðx; tÞ @ 2 uðx; tÞ
¼ c pðxÞ þ ¼0 ð2:21Þ
@t2 rffiffiffiffiffiffi @x @x2
S0 ðxÞ
where c ¼ Erd , as before and pðxÞ ¼
SðxÞ
Under boundary conditions that ultrasonic fatigue speci-
mens must satisfy, the solution to Eq. 2.21 takes the form
u(x,t) ¼ U(x) sin(ot), and the equation for the amplitude of
vibration U(x) (see Eq. 2.12) at each point along the specimen
can be easily obtained
U 00 ðxÞ þ pðxÞU 0 ðxÞ þ k2 UðxÞ ¼ 0 ð2:22Þ

Figure 2.2 Ultrasonic specimen geometry: A. Endurance speci-

men. B. Crack growth specimen.
Ultrasonic Fatigue Concepts 17

where
o
k¼ ð2:23Þ
c
To obtain the solution of Eq. 2.22, we must define the
curve in the central part of the specimens given in Figure
2.2. For the axisymmetric specimen in Figure 2.2A, if the
curve is of a profile of hyperbolic cosine, i.e.,

yðxÞ ¼ R2 ; L2 < jxj  L ð2:24aÞ

yðxÞ ¼ R1 coshðaxÞ; jxj  L2 ð2:24bÞ

where


1 R2
L ¼ L1 þ L2 ; a¼ arccosh ð2:25Þ
L2 R1

Then from the boundary conditions of Eq. 2.22, we can

find the specimen’s resonance length.
 
1 1
L1 ¼ arctan ½b cothðbL2 Þ
a tanhðaL2 Þ ð2:26Þ
k k

where
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b¼ a2
k2 ð2:27Þ

The solution of Eq. 2.22 is

sinhðbxÞ
UðxÞ ¼ A0 jðL1 ; L2 Þ ; jxj  L2 ð2:28aÞ
coshðaxÞ

UðxÞ ¼ A0 cosðkðL
xÞÞ; L2 < jxj  L ð2:28bÞ

where

cosðkL1 Þ coshðaL2 Þ
jðL1 ; L2 Þ ¼ ð2:29Þ
sinhðbL2 Þ
18 Chapter 2

With this solution, it is easy to obtain the strain and

stress for the reduced section part and for the cylindrical part.
The results are as follows.

For the reduced section part (jxj  L2):

eðxÞ ¼
½b coshðbxÞ coshðaxÞ
a sinhðbxÞ sinhðaxÞ
A0 jðL1 ; L2 Þ
cosh2 ðaxÞ
ð2:30aÞ
sðxÞ¼
½bcoshðbxÞcoshðaxÞ
asinhðbxÞsinhðaxÞ
Ed A0 jðL1 ;L2 Þ
cosh2 ðaxÞ
ð2:30bÞ

For the cylindrical part (L2 < jxj  L):

eðxÞ ¼ kA0 sinðkðL
xÞÞ ð2:31aÞ

sðxÞ ¼ Ed kA0 sinðkðL
xÞÞ ð2:31bÞ

The difference between the revolution surfaces with the

hyperbolic cosine profile and a circular profile is very small
for the axisymmetric ultrasonic fatigue specimen. Therefore,
we can use the analytical solution given above to avoid the
numerical calculation of the resonance length and the stress
field of the specimen.
For the plane stress specimen of Figure 2.2B, if the
central part has an exponential profile, i.e.,
yðxÞ ¼ R2 ; L2 < jxj  L ð2:32aÞ

yðxÞ ¼ R1 expð2a1 xÞ; jxj  L2 ð2:32bÞ

with


1 R2
a1 ¼ ln ð2:33Þ
2L2 R1
Ultrasonic Fatigue Concepts 19

then we can find the specimen’s resonance length as follows.

 
1 1
L1 ¼ arctan ½b1 cothðb1 L2 Þ
a1  ð2:34Þ
k k
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b1 ¼ a21
k2 ð2:35Þ

The solution of Eq. 2.22 is

UðxÞ ¼ A0 j1 ðL1 ; L2 Þ sinhðb1 xÞ expð
a1 xÞ; jxj  L2

ð2:36aÞ

UðxÞ ¼ A0 cosðkðL
xÞÞ; L2 < jxj  L ð2:36bÞ

where

cosðkL1 Þ expða1 L2 Þ
j1 ðL1 ; L2 Þ ¼ ð2:37Þ
sinhðb1 L2 Þ

The strain and stress are as follows.

For the reduced section part (jxj  L2):

eðxÞ ¼ A0 j1 ðL1 ; L2 Þ½b1 coshðb1 xÞ
a1 sinhðb1 xÞ expð
a1 xÞ

ð2:38aÞ

sðxÞ¼Ed A0 j1 ðL1 ;L2 Þ½b1 coshðb1 xÞ
a1 sinhðb1 xÞexpð
a1 xÞ

ð2:38bÞ

The strain and stress in the rectangular part (L2 <jxj  L) are
again given by Eq. 2.31, and the detailed solution procedure is
given in Section 2.5.
Because of the surface finishing difficulties of specimens
with a hyperbolic cosine (catenary) profile in the center, we
examine the difference between this geometry and a circular
profile specimen that is commonly and economically used in
practice.
20 Chapter 2

In Figure 2.3, we draw two curves, a circle

x2 þ ðb
yÞ2 ¼ r2 ð2:39Þ
and a catenary
y ¼ ðb
rÞ coshðaxÞ ð2:40Þ
with b ¼ 11.5, r ¼ 10 and a ¼ 0.249843 (in mm).
The values of these two equations and the difference in y
are summed up in Table 2.1 (Wu, 1992), which shows that the
largest difference between the two curves is approximately
1.8% and situated in 4 mm from the specimen center. We con-
sider therefore that, for a length of 2 mm from the center (a
total of 4 mm), the catenoidal profile agrees well with that
of the circle. In other words, the analytical solution of the
specimen geometry described in this section can be used as
a good approximation for the actual specimen without
performing costly numerical simulations.
Besides catenoidal approximation, other forms of reduced
section profile may be used; for example, the forms of simple
dumbbell, hollow dumbbell, exponential dumbbell, half
dumbbell, and half straight as summarized in (Wu, 1994).
Analytical solutions of resonance length exist for some of
them. For example, (Koslov, 1988) gives an analytical solution
for exponential profile.

Figure 2.3 Catenoidal approximation of a circle.

Ultrasonic Fatigue Concepts 21

Table 2.1 Difference Between a Circle and Its Catenoidal

Approximation
x y circle y catenary Dy Dy (%)

0 1.49999 1.5
1.43051e-5
9.53674e-4

0.5 1.51277 1.51172 1.04845e-3 0.0693545
1 1.55122 1.54706 4.15492e-3 0.268569
1.5 1.61563 1.60658 9.05788e-3 0.563801
2 1.70655 1.69119 0.0153511 0.907705
2.5 1.82472 1.80224 0.0224781 1.24723
3 1.97118 1.94144 0.0297375 1.53172
3.5 2.14731 2.11099 0.0363224 1.72063
4 2.35484 2.31351 0.0413277 1.78636
4.5 2.59604 2.55219 0.0438468 1.71801
5 2.87379 2.83075 0.0430427 1.52054
5.5 3.19186 3.15354 0.0383225 1.21522
6 3.55523 3.52561 0.0296195 0.840123
6.5 3.97063 3.95277 0.0178688 0.452057
7 4.44756 4.44169 5.87225e-3 0.132208
7.5 4.99999 5.00001
2.3365e-5
4.67299e-4

2.4. STRESS MAGNIFICATION FACTOR

The stress magnification factor is defined as the ratio of the

maximum stress in the reduced section specimen to that in
the constant section specimen having the same length, bound-
ary, and excitation conditions.
For the specimen with surface of revolution of a hyper-
bolic cosine profile in the center, the magnification factor of
stress, from Eqs. 2.30b and 2.14, is

b cosðkL1 Þ coshðaL2 Þ
F¼ ð2:41Þ
k sinhðbL2 Þ

For a plane stress specimen, this factor is

b1 cosðkL1 Þ expða1 L2 Þ
F¼ ð2:42Þ
k sinhðb1 L2 Þ
22 Chapter 2

2.5. ANALYTICAL SOLUTION OF

RESONANCE LENGTH

In the following section, we give the complete procedures for

obtaining the solution of a longitudinal wave Eq. 2.22 and
the resonance length L1 for both endurance and crack growth
specimens.

2.5.1. Endurance Specimen

The endurance specimen is axi-symmetric (Figure 2.2A).
From Eq. 2.24 that defines the profile of a longitudinal
section, we have

SðxÞ ¼ pR22 ; L2 < jxj  L ð2:43Þ

SðxÞ ¼ pR21 cosh2 ðaxÞ; jxj  L2 ð2:44Þ

For the cylindrical part, the section area S(x) keeps

constant, so that p(x) in Eq. 2.22 vanishes
S0 ðxÞ
pðxÞ ¼ ¼0 ð2:45Þ
SðxÞ
and Eq. 2.22 becomes

U 00 ðxÞ þ k2 UðxÞ ¼ 0 ð2:46Þ

This ordinary differential equation of second order has

the solution in a general form
UðxÞ ¼ C1 cosðkxÞ þ C2 sinðkxÞ ð2:47Þ

For the reduced section part that has a hyperbolic cosine

profile, we have from Eq. 2.44
S0 ðxÞ
pðxÞ ¼ ¼ 2a tanhðaxÞ ð2:48Þ
SðxÞ
Combining Eq. 2.48 with Eq. 2.22 results in

U 00 ðxÞ þ 2a tanhðaxÞU 0 ðxÞ þ k2 UðxÞ ¼ 0 ð2:49Þ

Ultrasonic Fatigue Concepts 23

Introducing a function
wðxÞ ¼ coshðaxÞUðxÞ ð2:50Þ

leads to

w0 ðxÞ ¼ U 0 ðxÞ coshðaxÞ þ a sinhðaxÞUðxÞ ð2:51Þ

w00 ðxÞ ¼ ½U 00 ðxÞ þ 2a tanhðaxÞU 0 ðxÞ þ a2 UðxÞ coshðaxÞ

ð2:52Þ
Comparing Eq. 2.49 to Eq. 2.52, we get

w00 ðxÞ ¼ ða2
k2 Þ coshðaxÞUðxÞ

or
w00 ðxÞ
ða2
k2 ÞwðxÞ ¼ 0 ð2:53Þ

This equation has the general solution

wðxÞ ¼ C3 expðbxÞ þ C4 expð
bxÞ ð2:54Þ

where b is defined in Eq. 2.27. The definition (Eq. 2.50) gives

the general solution of Eq. 2.49

C3 expðbxÞ þ C4 expð
bxÞ
UðxÞ ¼ ð2:55Þ
coshðaxÞ
This general solution was given in Wu (1992).
Now we have the solutions for the cylindrical part
(Eq. 2.47) and for the reduced section part (Eq. 2.55) with four
constants C1, C2, C3 and C4. These constants can be deter-
mined by appropriate boundary and continuity conditions.
The boundary conditions (Eqs. 2.9 and 2.18b) in the end
of specimen (note that the difference between the co-ordinate
systems of Figures 2.1 and 2.2) require, from the solution to
Eq. 2.47, that
C1 cosðkLÞ þ C2 sinðkLÞ ¼ A0 ð2:56Þ

C2 ¼ C1 tanðkLÞ ð2:57Þ
24 Chapter 2

From these two relations, we have

C1 ½cosðkLÞ þ sinðkLÞ tanðkLÞ ¼ A0 ð2:58Þ

so that
C1 ¼ A0 cosðkLÞ; C2 ¼ A0 sinðkLÞ ð2:59Þ
Consequently, the solution of displacement amplitude for
the cylindrical part is

UðxÞ ¼ A0 cos½kðL
xÞ ð2:60Þ

This is Eq. 2.28b.

The center of the specimen is a node of displacement
[U(0) ¼ 0]; therefore, from the general solution for this part
(Eq. 2.55), we obtain
C3 þ C4 ¼ 0 ð2:61Þ

and Eq. 2.55 is now rewritten as

2C3 sinhðbxÞ
UðxÞ ¼ ð2:62Þ
coshðaxÞ
Now the continuity conditions at x ¼ L2 require that the
displacement amplitude U(L2) and strain amplitude U0 (L2)
calculated from solutions (Eqs. 2.60 and 2.62) must be equal.
This gives
2C3 sinhðbL2 Þ
A0 cos½kðL
L2 Þ ¼ ð2:63Þ
coshðaL2 Þ
kA0 sin½kðL
L2 Þ
2C3 ½b coshðbL2 Þ coshðaL2 Þ
a sinhðbL2 Þ sinhðaL2 Þ
¼
cosh2 ðaL2 Þ
ð2:64Þ

Remember that L ¼ L1 þ L2 and, if we compare the two

expressions of Eqs. 2.63 and 2.64, we have

k tanðkL1 Þ ¼ b cothðbL2 Þ
a tanhðaL2 Þ ð2:65Þ

Ultrasonic Fatigue Concepts 25

or, again, Eq. 2.26 for the resonance length

 
1 1
L1 ¼ arctan ½b cothðbL2 Þ
a tanhðaL2 Þ
k k
Finally, Eqs. 2.61 and 2.63 give constants C3 and C4
A0 cosðkL1 Þ coshðaL2 Þ
C3 ¼
C4 ¼ ð2:66Þ
2 sinhðbL2 Þ
The solution (Eq. 2.62) now reads

cosðkL1 Þ coshðaL2 Þ sinhðbxÞ

UðxÞ ¼ A0 ð2:67Þ
sinhðbL2 Þ coshðaxÞ

which is Eq. 2.28a.

2.5.2. Crack Growth Specimen

The solution steps for the crack growth specimen of Figure
2.2B are similar to those for the endurance specimen
discussed above.
The areas of cross sections are given by
SðxÞ ¼ 2R2 L3 ; L2 < j xj  L ð2:68Þ

SðxÞ ¼ 2R1 L3 expð2a1 xÞ; j xj  L2 ð2:69Þ

and correspondingly
pðxÞ ¼ 0; L2 < j xj  L ð2:70Þ

pðxÞ ¼ 2a1 ; j xj  L2 ð2:71Þ

For the constant section part (L2 < jxj  L), the differen-
tial equation is again Eq. 2.46. Moreover, because the bound-
ary conditions at the end of the specimen are the same as
those for the endurance specimen, so the solution of displace-
ment amplitude is also expressed by Eq. 2.60.
For the central part with exponential profile, the differ-
ential equation is
U 00 ðxÞ þ 2a1 U 0 ðxÞ þ k2 UðxÞ ¼ 0 ð2:72Þ
26 Chapter 2

Now introduce a function

w1 ðxÞ ¼ expða1 xÞUðxÞ ð2:73Þ

Then

w01 ðxÞ ¼ ½U 0 ðxÞ þ a1 UðxÞ expða1 xÞ ð2:74Þ

w001 ðxÞ ¼ ½U 00 ðxÞ þ 2a1 U 0 ðxÞ þ a21 UðxÞ expða1 xÞ ð2:75Þ

Comparing Eq. 2.75 to Eq. 2.72, we have the equation for

w1(x)
 
w001 ð xÞ
a21
k2 w1 ð xÞ ¼ 0 ð2:76Þ

with the general solution

w1 ðxÞ ¼ C5 expðb1 xÞ þ C6 expð
b1 xÞ ð2:77Þ

where b1 is defined by Eq. 2.35.

Substituting Eq. 2.77 into Eq. 2.73 gives the general
solution of Eq. 2.72
U ð xÞ ¼ ½C5 expðb1 xÞ þ C6 expð
b1 xÞ expð
a1 xÞ ð2:78Þ

The condition that U(0) ¼ 0 results in

C5 þ C6 ¼ 0 ð2:79Þ

From Eqs. 2.78 and 2.79, we have

UðxÞ ¼ 2C5 sinhðb1 xÞ expð
a1 xÞ ð2:80Þ

Now our task is to determine the constant C5 from the

continuity conditions that U(x) and U0 (x) must satisfy at
x ¼ L2. This gives, from Eqs. 2.60 and 2.80

A0 cosðkL1 Þ ¼ 2C5 sinhðb1 L2 Þ expð
a1 L2 Þ ð2:81Þ

A0 k sinðkL1 Þ ¼ 2C5 ½b1 coshðb1 L2 Þ

ð2:82Þ

a1 sinhðb1 L2 Þ expð
a1 L2 Þ
Ultrasonic Fatigue Concepts 27

From these two relations, we find the resonance length of the

crack growth specimen and the constant C5
 
1 1
L1 ¼ arctg ½b1 cothðb1 L2 Þ
a1  ð2:83Þ
k k

A0 cosðkL1 Þ
C5 ¼
C6 ¼ expða1 L2 Þ ð2:84Þ
2 sinhðb1 L2 Þ

Substituting Eq. 2.84 into Eq. 2.80 gives the solution of

Eq. 2.72

UðxÞ ¼ A0 j1 ðL1 ; L2 Þ sinhðb1 xÞ expð
a1 xÞ; j xj  L2

ð2:85Þ

This is the solution given in Eq. 2.36a.

2.5.3. Transition Section

In a vibration system, we must reduce the number of dis-
placement nodes because it is easier to excite the system
when the order of vibration mode is low. For a fatigue test,
the specimen center must be a displacement node because of
the mechanical symmetry of the specimen. The vibration at
the two (or one) positions where static load is applied has
to be avoided. Thus, there will be three displacement nodes
in the test domain. Between the two neighboring nodes,
there is a stress node at each end of the specimen (Figure
2.4). To conclude, there are generally four nodes of dis-
placement (DN) and four nodes of stress (SN), as shown in
Figure 2.5.
Thus, the test domain is composed of a specimen in the
middle and two transition sections in two end parts. The
intrinsic frequency for all parts must be the same (20 kHz).
The geometry of the transition section can be determined in
the same way as for the specimen itself.
These discussions are also valid when there is only one
transition section, with the load ratio R ¼
1.
28 Chapter 2

Figure 2.4 Distributions of displacement and stress in test domain.

2.5.4. About the Effect of Transverse Motion

In the one-dimensional formulas of elastic wave discussed so
far, we have neglected the transverse motion of the media.
This implies that we consider a thin bar which is much longer
than it is wide. In other words, physically, we omit the effect
of the Poisson ratio.
Ni (1996) demonstrates that this treatment is reasonable
and the difference in eigenvalues between a one-dimensional
model and a fully three-dimensional model is small.
For a cylinder of length l and radius r, Ni
  (1996) finds
that, if the terms equal to or higher than O2 rl are omitted
in the development of Bessel functions of the solution, the
eigenvalue problems
 for the two cases are identical, and that,
2 r
if the term O l is considered, we have the following relation
(Ni, 1996):

fn 1 r 2
¼ 1
n 2 p2 ð2:86Þ
f 4 l
where fn is the eigen frequency of the first mode of longitudi-
nal vibration taking into account the transverse dimensions
of the specimen, and f is that of a very thin cylinder with a
zero Poisson ratio calculated by the formula in Eq. 2.17.
From the formula in Eq. 2.86, data from Table 2.2, and a
Poisson ratio 0.3, we find that for specimens of T6A4V and
Ultrasonic Fatigue Concepts 29

Figure 2.5 Displacement and stress nodes.

U500, the fn =f values are, respectively, 0.998686 and

0.998547. These results justify the use of one-dimensional
approximation in the determination of resonance length of
an ultrasonic fatigue specimen.
Tables 2.2 and 2.3 list dimensions of some specimens
tested in the CNAM=ITMA laboratory (see Figure 2.2 for
the meaning of symbols). The circular part of the specimen
center is approximated by a catenoidal profile, and the reso-
nance length L1 is determined analytically. The values of
the stress magnification factor range from 5 to 10 for these
30 Chapter 2

Table 2.2 Dimensions of Fatigue Life Specimens (length in mm)

Material L1 L2 B0 B1 B2 r (kg=m3) Ed (GPa)

T6A4V 18.19 14.31 31 1.5 5 4420 110

U500 16.60 14.31 31 1.5 5 8020 214
17-4PH 16.14 14.31 31 1.5 5 7830 203
T6A4V 15.03 14.31 31 1.5 5 4420 108
IN718 18.15 15.00 25 2.5 7.5 8200 215
U500 30.704 7.5 10 1.5 5 8020 214
T6A4V 29.01 7.5 10 1.5 5 4420 108
42CrMo4U-Rep B 16.94 14.31 31 1.5 5 7820 211
42CrMo4U-Rep C 17.14 14.31 31 1.5 5 7870 216

specimens. The dimensions of the ultrasonic fatigue three-

point bending specimen of aluminium alloy-based metal–
matrix composites are given in Chapter 3 and the test results
discussed subsequently.

2.6. METHODS FOR CALCULATING CRACK

TIP STRESS INTENSITY FACTOR

In the field of fracture mechanics, formulas have been

available to calculate stress intensity factor K for various
specimens (Paris, 1965; Tada, 2000). However, it is
generally believed that all formulas were established for
static- or low-frequency cyclic loading rather than vibratory

Table 2.3 Dimensions of Fatigue Crack Growth Specimens

(length in mm)
Material W L1 L2 B0 B1 B2 r (kg=m3) Ed (GPa)

T6A4V 14 41.81 12.20 31 1.5 4 4420 108

U500 14 44.53 12.20 31 1.5 4 8020 214
U500 14 44.53 5 6 1.5 4 8020 214
17-4PH 14 43.75 12.20 31 1.5 4 7830 203
T6A4V 14 43.483 5 6 1.5 4 4420 108
Al2017A 14 64.1 33.9 90 1.5 4
Astroloy 14 44.61 12.20 31 1.5 4 8000 214
Al-Li8090 14 39.71 21.07 90 1.5 4 2350 86
Cr-Si 11 22.31 12.20 31 1.5 4 7850 210
Ultrasonic Fatigue Concepts 31

Table 2.4 Specimen Dimensions (mm)

B B1 B2 L1 L2

U500 31 1.5 4 12.2 45.05

Astroloy (20 C) 31 1.5 4 12.2 42.50
Astroloy (400 C) 31 1.5 4 12.2 42.50
17-4PH 31 1.5 4 12.2 44.40
T6A4V 31 1.5 4 12.2 42.38
Al-Li8090 31 1.5 4 12.2 52.00

wave excitations. In a wave vibration system, the situations

are more complex in the presence of an inertia force.
Since the use of the 20 kHz-resonance method was sug-
gested by the Fatigue Crack Growth (FCG) study in the
1970s, the determination of K1 in a vibration regime has
always been a key problem, that many researchers must solve
in the case of ultrasonic fatigue. It is well known that SIF is a
mechanical parameter characterizing the intensity of a stress
field around a crack tip. In the handbook of Tada, Paris, and
Irwin (Tada, 2000), the geometric shape and the state of
applied force are considered. Formulas are obtained using
numerical methods, such as the boundary collocation, Green’s
function, asymptotic approximation and FEM (finite element
method). However, to our knowledge, literature and hand-
book data do not adequately answer problems related to the
determination of SIF in the regime of ultrasonic fatigue.
We will present methods, developed mainly in the CNMA=
ITMA laboratory by Kong, Wu, and Ni, for calculating SIF in
an ultrasonic frequency regime. These approaches will provide
not only numerical tools such as efficient algorithms and inter-
faces to commercial FEM software but also practical formulas
(Kong, 1987; Ni, 1995, 1996a, 1996b; Bathias, 1997; Wu, 1994).

2.6.1. Two-Dimensional Approach Based on

Fracture Mechanics Concepts
Several researchers (Mayer, 1992; Mayer et al., 1993) have
used an approximation to compute K1:
pffiffiffiffiffiffi
K1 ¼ s paYða=wÞ
32 Chapter 2

where Y(a=w) is established for a static state. However, in a

vibration system, s is not a measurable quantity and is not
the same in two cross-sections along the specimen even if
the area of the section is constant. Because the specimens
used in ultrasonic fatigue testing often have special geome-
tries, certain numerical methods are required to obtain the
stress solution; the FEM is a suitable method (Wu, 1994;
Mayer, 1992).

2.6.1.1. Difficulties of Calculation

Figure 2.6 gives a special type of specimen for FCG test at
20 kHz, designed with the mechanical symmetry about the
crack line and with a variable thickness to reduce the reso-
nance length of the specimen.

Figure 2.6 Specimen geometry.

Ultrasonic Fatigue Concepts 33

To calculate K1 in a two-dimensional vibrating system,

the computation must incorporate the effects of the following
five features:
1. Presence of crack
2. Non-uniform thickness
3. Vibration
4. Higher order vibration mode
5. Influence of temperature on the experiments.

2.6.1.2. Development of Computation

A specimen with a crack is not a linear vibration system when
both opening and closing of the crack occur. In the case of a
small crack, we can consider the vibrating system as two dif-
ferent linear systems corresponding respectively to a system
with an opened and with a closed crack. For the fracture tests,
the opened state is relevant to the SIF.
The computation is based on the FEM using the linear
elasticity theory for the plane strain problem. Eight-node
isoparametric quadratic elements are employed.
For vibration fatigue testing at ultrasonic frequency and
at R ¼
1, K1max is considered to be DK1max and, thus, FEM
can be applied to obtain it. It is well known that there is a sin-
gular stress field around the
pffiffiffi crack tip. The elastic stresses are
inversely proportional to r where r is the distance from the
crack
pffiffiffi tip, whereas the displacements are directly proportional
to r  K1 can be found by stress s or crack-opening displace-
ment v, that is in the polar coordinates (r, y) at the crack tip
(Figure 2.7).

pffiffiffiffiffiffiffiffi
K1 ¼ lim 2prsy ; at y ¼ 0 ð2:87Þ
r!0
rffiffiffiffiffi
E p v
K1 ¼ lim ; at y ¼ p ð2:88Þ
r!0 1
n 2 2r 2

To simulate the singular field of elastic stress, two singu-

lar elements are disposed at the crack tip (Figure 2.7), and the
34 Chapter 2

Figure 2.7 Singular elements.

formula calculating K1 is
rffiffiffiffiffiffi
E p vd
K1 ¼ ð2:89Þ
1
v2 2d 2

where d is the width of the element, and vd is the crack open-

ing displacement at point D, which can be computed by means
of FEM. For a better simulation of the r
1=2 singularity of
elastic stress field around the crack tip, we can also use the
so-called transition elements that are adjacent to the singular
element. This has been done by Wu (1992) in a study of the
stress field for an ultrasonic fatigue crack.
For the specimen in Figure 2.6, the equilibrium
equations become

@sx @txy txy dt

þ þ þ fx ¼ 0 ð2:90Þ
@x @y t dy

@txy @sy sy dt
þ þ þ fy ¼ 0 ð2:91Þ
@x @y t dy
Ultrasonic Fatigue Concepts 35

where fx and fy are the body forces, and t ¼ t(y) is the speci-
men’s thickness. The additional terms

txy dt sy dt
and
t dy t dy

appear for a specimen with non-uniform thickness.

In two-dimensional FEM, the elementary stiffness
matrix and mass matrix are respectively presented by
Z 1 Z 1
e
½K  ¼ ½BT ½ D½BtjJ jdxdZ ð2:92Þ

1
1

Z 1 Z 1
e
½M ¼ r ½ N T ½ D½ N tjJ jdxdZ ð2:93Þ

1
1

with
2x y
x¼
1; Z¼ ; L ¼ L1 þ L2
w L
where r is the mass density; [B], [D], [N] are geometrical,
elasticity, interpolation matrices, and jJj is the Jacobian.
For our specimen, these matrices must be integrated
numerically.
In a specimen with a crack, the vibration problem is com-
plicated. The specimen can be considered to be two linear
mechanical systems—according to whether the crack is
opened or closed—with different intrinsic frequencies and
modes. Because the crack growth takes place when the crack
is opened, we are only interested in the mode for the opened
crack. By means of the FEM, the following numerical equa-
tions of vibration without damping can be used

½ K fug ¼ o2 ½M fug ð2:94Þ

or, with 1=o2 ¼ l,

l½ K fug ¼ ½M fug ð2:95Þ

36 Chapter 2

where [K] and [M] are global stiffness and mass matrices. We
now have the problem of finding the eigenmatrix and eigen-
vector. Solving this consists of finding the values l and fug.
The inverse iteration method can give the largest eigenvalue
l1 and the corresponding eigenvector fu1g (Eq. 2.95); the
largest eigenvalue corresponds to the fundamental frequency
of the mechanical system.
Since the intrinsic frequency of about 20 kHz is not the
fundamental frequency, the inverse iteration method cannot
be directly used and needs a transformation.
Suppose the frequency of about 20 kHz corresponds to
2
oe . We have the following inequalities
0  o21  o22      o2e      o2n ð2:96Þ
If we introduce
1
o2i ¼ o2 þ ð2:97Þ
Dli
then o2 can be chosen so that
jDle j ¼ max jDli j ð2:98Þ
i¼1;...;n

Eq. 2.95 becomes

Dl½K  fug ¼ ½M fug ð2:99Þ

with

½K   ¼ ½ K 
o2 ½M  ð2:100Þ

For Eq. 2.99, the largest eigenvalue Dle corresponds to

the mode with an intrinsic frequency of about 20 kHz. The
convergence of the inverse iteration method is rapid and the
cumulative error is minimized.
In the case of a turbine disk, it is customary to study the
FCG rate at elevated temperatures. The specimen is heated
by an inductor in its center with a temperature gradient
between the middle and the end of the specimen. Since the
end is free, and plasticity and thermal-mechanical coupling
at high strain rate is not taken into account in the present
Ultrasonic Fatigue Concepts 37

elastic analysis, there is no extra thermal stress in the speci-

men. This does not mean that heat effect is totally negligible,
however; for vibration fatigue, the material rigidity, such as
Young’s modulus, is a function of temperature. That is to
say, the vibration mode may change. So we have to consider
the temperature gradient to obtain a correct vibration mode
when the elementary stiffness matrix [K]e is determined.

2.6.1.3. Numerical Results

A typical finite element mesh is presented in Figure 2.8. To
check this mesh, an example is computed in which a constant
thickness specimen is loaded with static tensile stress at the
end. A formula to calculate K for this case is given Wu
(1992). The error of the FEM result is small, less than 5.7%
for a=w ranging from 0.15 to 0.5. The singular element size

Figure 2.8 A finite element mesh.

38 Chapter 2

is small. Figure 2.9 gives vibration modes. Note the boundary

condition that there is no horizontal constraint, and the iner-
tia force balances automatically in the horizontal direction.
For displacement amplitude at the specimen end
U0 ¼ 1 mm, the relations of DK versus a=w are presented in
Figure 2.10. From these curves, the experimental results,
da=dN vs DK, were plotted in Figure 2.11 for Astroloy, at both
room and elevated temperatures. The threshold is usually
lower at 400 C than at 20 C.

2.6.1.4. Formula for K

In order to simplify the calculation of K for ultrasonic fatigue,
we propose a formula. For FCG rate tests at 20 kHz, it is pos-
sible to obtain a standard formula for the specimen shown in
Figure 2.4 at room temperature, where Ed and r are constant

Figure 2.9 Vibration modes.

Ultrasonic Fatigue Concepts 39

Figure 2.10 Curves of DK versus a=w.

and Poisson’s ratio is about 0.3. For this purpose, Eq. 2.94 can
be written as

rffiffiffiffiffiffi2
0 r
½K fug ¼ o ½M 0 fug ð2:101Þ
Ed
Matrices [K0 ] and [M0 ] depend only on geometrical dimen-
sions. For this specimen, the resonance length L1 is uniquely

Figure 2.11 Test data at 20 kHz for Astroloy.

40 Chapter 2

determined so that it p has a 20 kHz intrinsic frequency. It

ffiffiffiffiffiffiffiffiffiffiffi
depends directly upon Ed =r, sound velocity in a material.
Eq. 2.102 gives the relation, calculated by the FEM with the
aid of dichotomizing search.
sffiffiffiffiffiffi
Ed
L1 ¼
20:327 þ 12:765 ð2:102Þ
r

From dimensional analysis considerations, the following

formula is introduced
rffiffiffi
Ed p
DK1 ¼ 2
A0 f ða=wÞ ð2:103Þ
1
n a

where A0 is a measurable quantity and f(a=w) is a dimension-

less correction factor. This formula describes only the positive
value of DK1.
Generally, the function f(a=w) depends on the resonance
length. However, as observed in Figure 2.12, its dependency
pffiffiffiffiffiffiffiffiffiffiffi
on L1 does not seem to be significant for Ed =r ¼
4:8 5:7 km=sec and for a=w ¼ 0 0.5. Therefore, the follow-

Figure 2.12 Correction factors (Kong, 1991).

Ultrasonic Fatigue Concepts 41

ing polynomial expression (Eq. 2.104) is established as correc-

tion factor. When the temperature field is not uniform, it is
difficult to obtain a standard formula. But, if the temperature
field is uniform, the formula (Eq. 2.103) is still suitable even
at a higher temperature.

f ða=wÞ ¼ 0:64ða=wÞ þ 1:73ða=wÞ2

3:98ða=wÞ3 þ 1:96ða=wÞ4 ð2:104Þ

2.6.1.5. Calculation by ANSYS

The previous calculation was for cyclic loading with R ¼
1.
To determine the SIF in the presence of a static load (Sun,
2001), instead of developing our own program, the ANSYS
program is used. In this case, the SIF is not a simple superpo-
sition of static and dynamic components, because a geometri-
cal matrix depending on the static force influences is added to
the vibration equation.
Since the experimental system is mechanically symme-
trical about the crack, only half of the specimen will be taken
as the calculation model, as was done in the above for R ¼
1.
But, if half the specimen alone is taken, a more complex vibra-
tion equation will result with a non-zero matrix on the right
side and the calculation will be more involved. If the specimen
and the cone are calculated together, the equation becomes
simpler

½ K  þ ½ K g fug
o2 ½ M fug ¼ f0g ð2:105Þ

where [K]g is a geometrical matrix. The right side is zero,

because the dynamical displacements are zero at the ends of
this model and the force does no work (Figure 2.13). In this
case, computation is easier.
When using ANSYS, isoparametric elements of eight
nodes (STIF82) simulate the specimen, and beam elements
(STIF3) simulate the cone. An elastic tee is introduced to con-
nect the two types of the elements. To do so, it is necessary to
introduce two degenerated STIF82 elements in the specimen.
When conducting experiments, a screw connects the specimen
42 Chapter 2

Figure 2.13 Calculation model using ANSYS.

and the cone. Since the dynamical stress of the connection

zone is theoretically very small, this tee connection would
not cause significiant errors. The constraint equations repre-
senting the relation of displacement and rotation in the con-
nection are as follows
Ux4 ¼ Ux2 ð2:106Þ

1
Uy4 ¼ ðUy1 þ Uy2 þ Uy3 Þ ð2:107Þ
3

Uy3
Uy1
ROTz4 ¼ ð2:108Þ
x3
x1
where the numerical subscripts 1 through 4 correspond to
those in Figure 2.13. In vibratory fatigue, the stress ratio R
can be defined only by the ratio of stress intensity factors as
Kmin
R¼
Kmax
with Kmin ¼ Ks
Ka and Kmax ¼ Ks þ Ka where Ks and Ka are
computed as follows.
Ultrasonic Fatigue Concepts 43

Figure 2.14 SIF amplitude and static SIF for T6A4V.

First, the static SIF—called mean factor, Ks, which is

proportional to tension force—is determined by Eq. 2.109
½Kfus g ¼ ff g ð2:109Þ
where ffg and fusg are vectors of static force and displace-
ment, respectively. Secondly, the SIF amplitude, Ka, is deter-
mined by Eq. 2.105. Figure 2.14 gives curves determined for
titanium alloy T6A4V. Ka is proportional to the displacement
amplitude at the specimen’s end. An optical sensor measures
this amplitude. We observe that the greater the static force,
the smaller the Ka for a given crack length.

2.6.2. Three-Dimensional Computation of Stress

Intensity Factor
In the previous section, half a specimen is taken into account
and, by a self-compiling two-dimensional finite element code,
K1 values are evaluated using the opening displacement of a
node close to the crack tip.
Although the basic idea of this approach is acceptable
with the newly defined function f(a=w) for use in ultrasonic
fatigue, there is also room for improvement. First, only one
44 Chapter 2

half of the specimen is computed on the assumption that the

displacement and deformation distribution along the speci-
men is symmetric about its center line. However, the lower
side of the specimen is linked elastically with an amplifying
horn while the upper side is free, and so the displacement
and strain will not be symmetric in the upper and the lower
parts of the specimen. Secondly, the analysis is carried out
in two-dimensions, neglecting the Poisson’s contraction effect
in thickness. And thirdly, the finite element mesh close to the
crack tip is not fine enough. In this section, methods for eval-
uating K1 by means of displacement and energy approaches
are described (Ni, 1996; Wu, 1992).

2.6.2.1. Three-Dimensional Finite

Element Models
The ultrasonic FCG specimen, shown in Figure 2.15, is
designed to have exactly two quarter-wavelengths, with sym-
metry about the middle section. A lateral notch is made in the
central section to facilitate and localize the crack initiation.
The specimen is linked elastically with the amplifying
horn(s)—either at one end for load ratio R ¼
1, or at both
ends for R >
1. The dimensions in Figure 2.15 are

Figure 2.15 Ultrasonic FCG specimen coupled with amplifying

horns.
Ultrasonic Fatigue Concepts 45

R0 ¼ 31 mm, B1 ¼ 3.0 mm, B2 ¼ 8.0 mm, and L1 ¼ 12.2 mm, and

W ¼ 14 mm.
And, for the specimen of T6A4V and the amplifying horn
studied here, D1 ¼ 18.0 mm, D2 ¼30.38 mm, and L3 ¼
75.51 mm, with the resonant length L2 determined to be
L2 ¼ 41.88 mm.
Software ALGOR is adopted in computation. The finite
element mesh for the model has the following two features:
1. The whole specimen and the amplifying horn(s) are
considered together. The finite element mesh is
symmetric about the central line of the specimen.
2. A group of finite element meshes with minute
dimensions is installed all around the crack tip,
and it is movable with the crack tip. The smallest
dimension of the element meshes close to the crack
tip is 1=224 of the specimen width, W.

2.6.2.2. SIF Determined by Displacement and

Energy Approaches
Displacement approach involves a singular element modeling
of r
1=2 singularity in the elastic stress field around the crack
tip, the approach used in the previous section. The relations
between the SIF and the crack length obtained by the displa-
cement method are given in Figure 2.16
The energy method is a more global approach than
the displacement method. The basic idea of using strain
energy is particularly well suited for in finite element
simulation.
The elastic energy, G, made available for crack extension
is given by


@V
G¼þ ð2:110Þ
@A

when the load is kept constant, and where V is the elastic

strain energy of the body and A is crack surface area.
In our three-dimensional finite element calculation, the
energy rate, G, is determined as follows. First, the strain
46 Chapter 2

Figure 2.16 K1 versus a by displacement

pffiffiffiffiffi approach: (A) R ¼
1;
(B) R >
1 (U0 ¼ 1 mm, K1 in MPa m).

energy V is computed when the crack area is A by the formula

1
V ¼ fugT ½Kfug ð2:111Þ
2
Note that the stiffness matrix [K] is a function of A.
Then under the same displacement fug, the strain
energy V þ DV corresponding to a crack area A þ DA is calcu-
lated. The elastic energy rate, G, is determined by


DV
G¼
ð2:112Þ
DA
In plane stress conditions, the relation between G and K1 is
given by
pffiffiffiffiffiffiffiffiffiffi
K1 ¼ E d G ð2:113Þ
An interface routine interconnected with ALGOR is compiled
for calculating G as well as K1 values in this ultrasonic FCG
study.
The relationship between K1 and crack length a com-
puted by the energy method is given in Figure 2.17.
The determination of K1 by means of the energy method
may achieve a better precision than the displacement method,
since, in the energy approach, the evaluation of the displace-
ment distribution is taken into consideration over the entire
vibration body (specimen plus horns), while in the displace-
Ultrasonic Fatigue Concepts 47

Figure 2.17 K1 versus a by p energy

ffiffiffiffiffi approach. (A) R ¼
1;
(B) R >
1 (U0 ¼ 1 mm, K1 in MPa m).

ment method, only displacements of nodes near the crack tip

are used.
In the case of R ¼
1, K1 determined by energy approach
increases progressively with the crack length, and then devel-
ops a decreasing tendency for crack lengths greater than
6.5 mm (Figure 2.17). This decreasing characteristic is not
observed in the displacement approach (Figure 2.15). In fact,
as a fatigue crack initiates and then propagates in ultrasonic
resonance vibration, the vibration energy transformed
through the continuous medium is reflected at crack lips. As
a result, the system in resonant vibration becomes more
and more detuned and deviates from its proper frequency,
resulting in a hybrid effect of vibration modes. Figure 2.18 fol-
lows the evolution of the resonance frequency as the crack
length increases. Once the crack length reaches a certain
value, the resonant system in ultrasonic vibration will be
destroyed and the vibration energy will decrease. This effect
has been observed in experiments and is easily simulated
numerically (Figure 2.19).
It is mandatory, therefore, to control carefully the fre-
quency of the machine when the crack is propagating. An
error in the frequency induces an error in the stress intensity
factor. In order to avoid such errors, the piezoelectric machine
must be controlled with a computer and the test has to
be stopped when the crack length reaches one half of the
specimen width.
48 Chapter 2

Figure 2.18 Resonant frequency versus crack length a without

frequency control.

Figure 2.19 Detuning effect of the resonant vibration due to a

crack (a ¼ 5.5 mm).
Ultrasonic Fatigue Concepts 49

In the case of R >
1, K1 increases continuously as the

fatigue crack propagates from 0.5 mm to 7.5 mm. When a sta-
tic tensile load is applied to the ultrasonic vibration system,
the fatigue crack opening displacement at the crack tip may
be greater than that obtained for R ¼
1, even if the vibration
energy sent through the cross-section of the specimen
decreases. Experimental data indicate that the FCG beha-
viors at ultrasonic fatigue frequency are comparable to those
observed in conventional fatigue tests regardless of the R
ratio. In ultrasonic resonance vibration, as a fatigue crack
initiates and then propagates, the vibration energy is trans-
mitted through the continuous medium and is reflected at
crack facets. It is not clear why the FCG at ultrasonic fre-
quency is not affected by the high frequency. The effects of
the following several factors are discussed later.
The effect of the deformation rate de=dt on the plastic
zone size at the crack tip;
The effect of vibration on the residual stresses at the
crack tip;
The effect of the plastic zone size and the residual
stresses on the crack opening and the crack closure;
The effect of the environment at high frequency.
3
Testing Machines and Their
Performance

3.1. INTRODUCTION

Up to now, there have been no standards for testing proce-

dures and testing machines of ultrasonic fatigue, although
efforts are in progress within ASTM to provide a recom-
mended practice and ultimately a testing standard (Bathias,
1998). Because of this, laboratories must develop their own
machines and design practical test procedures. The labora-
tories of Willertz in the United States, Stanzl in Austria,
Bathias in France, Ni in China, Ishii in Japan, and Puskar
in Slovakia are among the leading laboratories in this field.
Although ultrasonic fatigue test machines in these labora-
tories are not the same, some components are common to all
machines. The three most important are: (1) a high
frequency generator that generates 20 kHz sinusoidal electri-
cal signal, (2) a transducer that transforms the electrical
signal into mechanical vibration, and (3) a control unit.
Early ultrasonic fatigue machines performed only uni-axial

51
52 Chapter 3

(one-dimensional) and constant amplitude tests, so the con-

trol unit and other parts were not very complicated. In the
last two decades, progress has been made to extend the ultra-
sonic fatigue technique to variable amplitude loading condi-
tions, low or high temperature environments, torsional or
multi axial tests, and so on. Thus, designing a modern ultra-
sonic fatigue test machine may involve mechanical, electrical,
optical, magnetic, and thermal considerations. In France,
Bathias used a first ultrasonic fatigue test machine in 1967
on the principle used by Mason (Bathias, 1998). As indicated
in an early review paper (Stanzl, 1996), the rather restrictive
uses of the ultrasonic fatigue test method appeared to be
partly due to the lack of commercially available test equip-
ment, forcing the individual investigators to work with impro-
vized facilities not readily amenable to standardized
experimental conditions.

3.2. BASIC STRUCTURE

As stated above, an ultrasonic fatigue test machine must

include the following three common components:
1. A power generator that transforms 50 or 60 Hz
voltage signal into ultrasonic 20 kHz electrical sinu-
soidal signal.
2. A piezoelectric (or magnetostrictive) transducer
excited by the power generator, which transforms
the electrical signal into longitudinal ultrasonic waves
and mechanical vibration of the same frequency.
3. An ultrasonic horn that amplifies the vibration com-
ing from the transducer in order to obtain the
required strain amplitude in the middle section of
the specimen.
These three parts are special devices required for the produc-
tion of ultrasonic fatigue load. Other components of an ultra-
sonic fatigue test machine may include recording systems
(amplitude control unit voltmeter, frequency control unit,
cycle counter and oscilloscope) and measuring systems
(displacement sensor and video camera observation unit).
Testing Machines and Their Performance 53

The function of the system shown in Figure 3.1 is to

make the specimen vibrate in ultrasonic resonance at one of
its longitudinal modes. The displacement amplitude reaches
its maximum U0 at the end of the specimen, which can be
measured by means of a dynamic sensor, while the strain
excitation in push–pull cycles (load ratio R ¼
1) attains the
maximum in the middle section of the specimen that produces
the required high frequency fatigue stress. The video camera
supervision system in Figure 3.2 is used in the fatigue crack
growth test for observing and recording crack initiation and
propagation processes. The information recorded may include

Figure 3.1 Full resonance system in Bathias’s laboratory with

schematic view of apparatus.
54 Chapter 3

Figure 3.2 Diagram of equipment with computer control.

ultrasonic cyclic displacement amplitude U0, the evolution of

the fatigue crack growth, which enables us to determine the
fatigue crack growth rate da=dN, and the stress intensity
factor Kmax, calculated by analytical or numerical methods
(Wu, 1991).
During ultrasonic fatigue tests, the maximum strain
values can be measured directly using miniature strain
gauges, suitably positioned on the sample surface. For exam-
ple, a measuring system consists of a Wheatstone bridge
amplifier, dynamic strain gauges (0.79 mm by 0.81 mm), and
a digital oscilloscope (two channels, 40 K memory for each)
has been built for direct strain measurements in Bathias’s
laboratory.
In the same laboratory, the dynamic displacement ampli-
tude at the specimen extremity, U0, is measured by an optic
fiber sensor, which permits measurements of the displace-
ment from 1 mm to 199.9 mm, with a resolution of 0.1 mm.
The magnification factor of stress can then be calculated
according to these measurements. For a virgin specimen
(i.e., without a crack), the vibratory stress and strain can also
be determined at the midsection. The maximum strain value
thus determined is then confirmed to be accurate by use of the
Testing Machines and Their Performance 55

above-mentioned strain gauge. In addition, a system of

video-camera–television has been used for the control of crack
initiation and propagation. This system refines events to
1=25th of a second and magnifies specimen surface 140 
200 times.

3.3. NONSYMMETRICAL AND VARIABLE

AMPLITUDE TEST EQUIPMENT

Because specimens of ultrasonic fatigue vibrate in resonance,

a free end is sufficient for symmetric loading conditions
(R ¼
1). This avoids the large and cumbersome arrange-
ments for gripping the specimen that is often encountered
in conventional fatigue testing (Figure 3.3). If there is a static
load, the situation is different.

Figure 3.3 Vibratory stress and displacement field and computer

control system.
56 Chapter 3

Figure 3.4 Vibration system for ultrasonic fatigue superposed on

static loading.

Superposing a mean stress or displacement upon the

symmetric tension–compression cycle can create a complex
cyclic loading with R >
1. Therefore, an additional horn will
be added at the other end of the specimen, as shown in
Figure 3.4
Such a test machine is particularly useful to efficiently
study fatigue endurance and fatigue crack growth behavior
of materials subjected to elasto-plastic low cycle fatigue load-
ing superposed on vibration stress cycles with high frequency
(Figure 3.4).

3.4. COMPUTER CONTROL SYSTEM

A computer control system is of great importance in program-

ming and controlling the load as well as in data acquisition.
Before 1981, there was an ultrasonic fatigue machine with a
computer control unit (Kong, 1987). Here, however, we dis-
cuss the machine built in Bathias’ laboratory (Wu, 1992). This
computer control system uses an IBM PC computer and a
data acquisition system composed mainly of a 12 bit A=D con-
verter, a 12 bit digital to analogue (D=A) converter, a strain
gauge board, and a thermocouple board. The A=D board
converts the data from analogue to digital in 20 ms. When
using the program with necessary commands, the acquisition
Testing Machines and Their Performance 57

Figure 3.5 Flow diagram for test control program.

time for TURBOBASIC language is 280 ms. This acquisition

time does not agree with the sampling theory, and with the
additional control operations, the time becomes much longer.
If sampling were directly executed for the 20 kHz sinusoidal
wave, auto-control would be impossible. For fatigue tests,
the most important factor is load; that is, the amplitude of
the sinusoidal wave. Therefore, a rectifier with a filter is
installed between the transducer and A=D converter, from
which the d–c output-voltage proportional to the vibrating
amplitude is obtained. At any time, the A=D converter can
detect the test load, and with this information, the computer
gives a control signal to the power amplifier at the D=A board
to maintain constant amplitude or to change the amplitude.
Figure 3.5 is the flow chart of the control program. When
the operator places the computer on line, by interrupting the
power amplifier potentiometer, the program demands an
input load with the expected amplitude to begin the control
test and the specimen vibrates with the amplitude of the load.
58 Chapter 3

In order to avoid the influence of a parasite signal for the sys-

tem input read by A=D, a hardware filter has been installed
before the A=D. In other words, in this control program there
is another software numerical filter that selects the medium
value among three values of every sampling to obtain a true
load signal. The adjustment of the D=A output voltage is
made to reduce the observed difference between the desired
load and actual feedback reading of the A=D load channel. A
software gain is introduced in the program so that the D=A
command to the amplifier is the product of a software loop
gain and the expected vibration amplitude of the specimen.
The computer is an integration link in this control loop. Cor-
rections to the D=A voltage are made automatically ten times
per second. The testing load can be modified while the test is
in progress. During this mode of operation, the operator has
the option of either changing vibrating amplitude or stopping
the test at the current mean load level.
A control loop is composed of the specimen, transducer,
voltmeter and rectifier, computer, power amplifier, and vibra-
tor. The regulation of some parameters in the program is very
important for this computer control system. The system pre-
sented in Figure 3.5 permits entry at expected vibrating
amplitude by a computer keyboard, so that the input of the
system is a jump function. The output of the system (that
is, the vibrating amplitude measured by a capacitive sensor)
responds differently to the different control parameters in
the program. Figures 3.6 and 3.7, respectively, present the
responses to the first two groups of parameters where the
output voltage (ordinate), which is the input of the A=D con-
verter, is directly proportional to the specimen vibrating
amplitude. Interval 40–60 represents the jump function from
40  60 mV and so on, and the curves represent the auto-
control system response to the jump function. In Figure 3.6,
the amplitude response converges with oscillation and the
convergence is slow (20 s). In Figure 3.7, there is no oscilla-
tion, but the convergence is still slow (13 s). For fatigue tests,
the convergence must be rapid and the curve smooth to avoid
the overload. When the parameters are chosen properly, the
satisfactory system response will be obtained. Figure 3.8 gives
Testing Machines and Their Performance 59

Figure 3.6 Response curve under the first parameters.

the responses under the final parameters chosen; that is, the
signal gain to the D=A converter depends on the difference
between the desired load and actual load. Clearly, the conver-
gence time is significantly shorter—it is 3 seconds, which is
much shorter than the time of manual operation.
Another important aspect is data acquisition. For ultra-
sonic fatigue crack tests, the specimen is relatively small
because it must have a high intrinsic frequency. A typical

Figure 3.7 Response curve under the second parameters.

60 Chapter 3

Figure 3.8 Response curve under the final parameters chosen.

specimen width is approximately 14 mm. The effective crack

propagation range is only about 4 mm. When a crack propa-
gates rapidly under a high load, it may cause a rapid change
in the vibratory amplitude because of the decrease of the spe-
cimen intrinsic frequency. This amplitude change may go
beyond the auto-control range of the computer and an unsa-
tisfactory experiment will result. Therefore, it is important
to distinguish the satisfactory period from the unsatisfactory
period, and to record the results for out-of-line analyses after
the tests. Because the entire process takes only 10 seconds or
so, the data acquisition using this control program is very
helpful. Figure 3.9 shows the testing charge recorded during
60 s and its relation with the crack length recorded by the
video system. These records provide the information needed
in calculation of the stress intensity factor and in determina-
tion of the crack growth rate. The decrease in testing load
after 40 s (when the machine stops automatically in case of
crack) occurs because the intrinsic frequency of the specimen
begins to depart significantly from the designed frequency
(error more that 5%). These results are then unusable.
The input data for the control system are elastic modu-
lus, mass density of the material to be tested, and the desired
Testing Machines and Their Performance 61

Figure 3.9 Charge recorded by computer and corresponding crack

length (without frequency control).

test stress. Several forms of amplifying horn can be presented

over a visual interface on the computer screen. This permits
one to calculate the amplification of displacement and stress
range in comparison with the tension measured by the inter-
face J2 (described later in this chapter).
The computer starts the control process by activating the
weakest vibration with the help of a card piloted by a numer-
ical exit. After 50 ms, the stress of vibration attains the
recorded level picked up from an input curve of stress signal.
Figure 3.10 shows a response signal to the start-up (Wu,
1992). The response signal comes from plug 9 of interface J2
(see Figure 3.13). We observe that the response curve has a
plateau at 50% power before attaining 100% power. The time
for reaching 100% is 85 ms, and there is no overload. Then,
the software maintains constant amplitude of vibration and
numerical filter eliminates false acquisitions. With plugs in
J2, we know the actual frequency and power. As soon as a
macro crack appears, the ultrasonic generator automaticaly
cuts off the current and the computer gives the fatigue
life of the specimen. This function also ensures that the test
conditions are maintained.
62 Chapter 3

Figure 3.10 Response signal to the start-up.

A generator with a converter composed of six piezo-cera-

mics is chosen to provide vibration energy. This ultrasonic
generator 900BA is made by Branson Ultrasonic Corporation.
It has a maximum power of 2 kW and provides a sinusoidal
signal for the converter that is the source of mechanical vibra-
tion. This amplifier automatically maintains the intrinsic
frequency of the mechanical system in the range of
19.5 kHz  20.5 kHz. The converter, horn, and specimen form
a mechanical vibration system with four stress nodes (null
stress) and three displacement nodes (null displacement) for
an intrinsic frequency of 20 kHz. Here, the stress and displa-
cement are considered to be longitudinal. In Figure 3.11,
points B and C (connected points), and point A and converter
top are stress nodes. The specimen center is a displacement
node; there the stress is at the maximum.
The horn must vibrate at a frequency of 20 kHz. Depend-
ing on the specimen loading, the horn is designed so that the
displacement is amplified between B and C, usually 3 to 9
times, meaning that the geometry between B and C must
determined accordingly. The finite element method may be
required when the geometrical shape is complex.
Testing Machines and Their Performance 63

Figure 3.11 Vibratory stress and displacement field.

The mechanical system composed of a converter, a horn,

and a specimen is linear, and all stress and displacement
fields are linear. It is necessary only to measure the ampli-
tude of one of them. To determine an S-N curve, one needs
to know the stress amplitude with good accuracy. However,
at high frequency and low temperature, it is difficult to mea-
sure the stress amplitude. Therefore, the stress in the mid-
section of the specimen is computed from the displacement
of the piezo-ceramics system.
Piezo-ceramics expand or contract when an electrically
induced tension is applied. The tension is proportional to
expansion or contraction; that is, the tension is proportional
to the displacement in the mechanical system. It is strictly
proportional to the expansion or contraction of the converter
and to the displacement of point C. In other words, electrical
current depends on the damping of the horn and specimen.
The damper is installed on the converter. In the generator,
an interface called J2 has been set up, which has a plug with
0  10 volt (DC tension) corresponding to 0  100% of vibra-
tion amplitude of the converter. This output is calibrated with
the displacement of the horn end (point B), to determine the
stress in the specimen using a computer that acquires this
64 Chapter 3

DC tension. The stress can be calculated by the following

equation
V
s ¼ Eks kh UC100%
10
where ks, is a factor of the specimen depending on the geome-
trical form, kh is the ratio of amplitude amplification, UC100%
is the maximum amplitude at point C, which is constant, and
V is DC tension acquired by the computer. According to this
formula, the test stress for a certain specimen can be altered
not only by changing output power but also by replacing the
horn.
For calibration, a simple cylindrical specimen was used
with a gauge mounted in the middle. The strain measured
by this gauge and displacement of horn end (point B) UB is
given by the relation below.
rffiffiffiffi
r
e ¼ 2p f UB
E

where r is mass density. When the DC output is calibrated

according to this measure, a comparison between measured
strain in liquid nitrogen and strain calculated by control com-
puter for different power is presented in Figure 3.12. It shows
a good linear relationship between measured and calculated
strains.
Another group of calibration tests was made with an
optical sensor that measures displacement of the specimen
end at room temperatures. It is possible to apply a correction
from room temperature to low temperature, since the amplifi-
cation ratio is known for different temperatures. The results
were also satisfactory.
Furthermore, interface J2 installed in the command box
(Figure 3.13) makes the computer control possible. In this
connector, plug 2 supplies a DC tension of 10 volts with which
we can use a potentiometer to control the equipment manu-
ally. Plug 8 can be loaded with a DC tension of 0  10 volts
corresponding to 50% and 100% of the vibration amplitude.
Plug 9 also gives a DC power of 0  10 volts proportional to
Testing Machines and Their Performance 65

Figure 3.12 Comparison of measured and calculated values of

strain at 77K .

the vibration amplitude regardless of the magnitude of the

mechanical excitation. Plugs 3 and 10 indicate the power
and frequency, respectively. In general, direct control for
20 kHz is very difficult. Thus, it relies on the use of d=c signal
proportional to the amplitude of the alternation current

Figure 3.13 Command box and interface J2.

66 Chapter 3

signal (Wu, 1992). A normal A=D and D=A converter card con-
necting connector J2 and a PC (Figure 3.1) can achieve a com-
puter-controlled test at 20 kHz. A control program has been
written with Turbo Cþþ language, which calculates the vibra-
tion stress in the specimen. The test starts by giving a test
stress and the real stress rises within 85 ms to the expected
level without overloading. Then, the stress is held constant
with a control accuracy  3 MPa. When a crack appears the
testing system stops automatically because its intrinsic fre-
quency decreases and it gives the fatigue life. With this soft-
ware, a fatigue test between 105 to 1010 cycles can be
performed. A generator with a converter composed of six
piezo-ceramics is chosen to provide vibration energy. It has
a maximum power of 2 kW.
The computer testing system described above has the
following advantages:

1. The output signal to the power amplifier from the

D=A converter does not correspond directly to the
input of the A=D converter; that is, the adjustment
is done according to the difference between the input
signal and expected amplitude. If the electronic drift
and main system error are caused by the tempera-
ture in the vibrator or in the power amplifier, the
computer control system compensates for those
errors so that the experiments are accurate.
2. The program can easily compensate for the non-
linearity of the transducer-rectifier–filter after the
input and output relationship is calibrated.
3. This system possesses a better convergence ten-
dency, a rapid convergence velocity, and a steady
amplitude error within 5%.
4. The system applies a block spectrum with gradual
change for ultrasonic fatigue endurance tests, thus
making programming easy.
5. When normal frequency drops off and the mechani-
cal system stops vibrating, the computer can auto-
matically stop the power amplifier to protect the
equipment.
Testing Machines and Their Performance 67

6. The design of the computer control system respects

the integrity of the original machine. The test
machine can run without the computer.
7. The change of test parameters is continuous rather
than stepwise in some multi-stage control systems,
such as that of (Stanzl, 1981).

3.5. HIGH TEMPERATURE TEST EQUIPMENT

Figure 3.14 is a diagram of a system for high temperature

tests, where the temperature in the specimen is constant
along 5 to 6 mm. Figure 3.15 is a photograph of this same sys-
tem. The test equipment consists of a heating device in the
middle, a capacity transducer above, and a video camera with
an enlargement factor of 200 on the right. The television
images can be recorded on videocassette during the tests.
The fatigue crack growth rate can be determined easily up

Figure 3.14 Evolution of the specimen temperature for a Ni base

alloy.
68 Chapter 3

Figure 3.15 Specimen installed in the machine including measur-

ing and heating devices.

to the order of 10
9 mm=cycle. For the experiments at ele-

vated temperatures, a high-frequency inductor made by
CELES is used; test temperatures can then reach 1000 C
without problems. The computer system has a thermocouple
board that can send the temperature analogue signal to an
A=D converter, so that the computer is able to control heating
of the specimen by using a disjunctor.
Because the Young’s modulus decreases at high tempera-
tures, the resonance length of ultrasonic fatigue specimen will
be shorter than that at ambient temperature. For tempera-
ture sensitive materials, this change must be taken into
account. For example, in a crack growth experiment, we
Testing Machines and Their Performance 69

usually start crack initiation at ambient temperature with

adequate resonance length. Then, by cutting off both ends,
we can obtain the resonance length at high temperature.
An early design of ultrasonic fatigue equipment at high
temperature can be found in reports by Ebara (1994). The
system described there is capable of studying fatigue crack
growth rate and fatigue thresholds at 22 kHz, at elevated
temperatures of 200  500 C in an argon environment in a
heat chamber, and at 20 C using water as a coolant.

3.6. LOW TEMPERATURE TEST EQUIPMENT

Let us now discuss the possibility of testing materials at low

temperatures. A system for ultrasonic fatigue tests at cryo-
genic temperatures has also been developed in Bathias’s
laboratory (Tao, 1996). In the laboratory, liquid nitrogen,
liquid hydrogen, and liquid helium are used to create a
cryogenic temperature atmosphere. Liquefied gasses are
costly, especially liquid helium. If conventional fatigue testing
were employed, the fatigue tests at very low temperatures for
titanium alloys used in space rockets would require a large
amount of liquefied gas because the tests would take a very
long time. This is another advantage of the ultrasonic fatigue
method, which substantially reduces testing time. The
machine with a computer control system works at 20 kHz
and at cryogenic temperatures (77 K and 20 K) for studying
fatigue behaviors of the titanium alloys used in rocket engines.
The device consists of three parts: a cryostat, a mechan-
ical vibrator, and a controlled power generator. Figure 3.16
shows the principal aspects of this machine, which is simpler
than a conventional hydraulic machine. The function of the
converter and the horn are the same as in other ultrasonic
fatigue apparatus: The converter changes an electronic signal
into a mechanical vibration and the horn plays the role of dis-
placement amplifier. A dewar cryostat contains liquefied
gasses to keep the testing temperature constant.
Ultrasonic fatigue tests at cryogenic temperatures for
load ratio R >
1 are also possible by adding a second horn
to the other end of the specimen (Figure 3.16).
70 Chapter 3

Figure 3.16 Low temperature and high frequency fatigue testing

machine.

Another example of low temperature test system is

that of Stanzl’s laboratory (Buchinger, 1984), where a
temperature environment of 77 K is guaranteed by liquid
nitrogen.
Testing Machines and Their Performance 71

3.7. THIN SHEET TEST EQUIPMENT

The geometry of ultrasonic fatigue specimens is usually either

a cylindrical or plane form with a reduced section in the cen-
tral part to form a higher stress area to accelerate the test
process.
Theoretically, as the excitation frequency of the machine
coincides with one of the resonance frequencies of the speci-
men, the specimen will vibrate in that frequency. To put this
theory into practice, however, is not an easy task. On the one
hand, the finishing of specimens precisely to the design
requires delicate work. On the other hand, the horn where
the specimen is installed may change the real frequency of
vibration of the system if the connection is not well designed.
This is the case especially for load ratio R >
1 when another
horn is necessary or when plane specimens are very thin. It
was indicated long ago (Ebara, 1994) that the use of a two
horn system and a positive constant mean stress is favored
for avoiding transverse vibration.
For most plane specimens of ultrasonic fatigue, the ratio
of the thickness to the largest dimension i.e, w=2ðL1 þ L2 Þ, is
about 6%  8%. Our experience shows that a thickness–length
ratio in this range does not pose severe problems to the
machine and control system in maintaining the desired fre-
quency of about 20 kHz. But, as the thickness–length ratio
of a plane specimen decreases by about one order of magni-
tude, say to 0.7%, the perturbation of vibration frequency of
the system becomes so great that the test could not be
performed at all if special measures were not taken.
A series of thin sheet tests of ferrous materials (Wang,
1996) has been conducted. The purpose is to determine the
fatigue strength (or S-N curve) at 109 cycles and the threshold
of cracking at a small propagation speed of 10
12 m=cycle,
both with R ¼ 0.1. Figures 3.17 and 3.18 present dimensions
of two types of specimen. The ratio of thickness to length is
0.7%, an order of magnitude smaller than that of ordinary
specimens.
The key to the execution of the test is how to fix the thin
specimen to the ends of amplifying horns. We cannot use the
72 Chapter 3

Figure 3.17 Fatigue life specimen of thin sheet.

same kind of set screw as that used in the attachment of

ordinary specimens (i.e., specimens with the thickness–length
ratio of 6%  8%). Among the methods for linking two steel
components are riveting, bolt jointing, welding, and gluing.
We have tried different methods and secured a special type
of screw and structural glue for the thin sheet specimens.
The glue is soluble in acetone, which reduces the number of
screws required.
In this way, we minimize the influence on the frequency
of the test machine used for thin sheet specimens that have
small transverse rigidity, and tests are therefore successfully
performed. Figure 3.19 presents the geometry of the special
screw and the connection with the horn.
Testing Machines and Their Performance 73

Figure 3.18 Crack growth specimen of thin sheet.

Figure 3.20 shows the test system. The experimental

results will be discussed in the next chapter.

3.8. HIGH PRESSURE PIEZO-ELECTRIC

FATIGUE MACHINE

It is well known that it is difficult to conduct a fatigue test

under high pressure with a conventional machine. The pro-
blem stems from the displacement of an actuator through
the wall of an autoclave. The use of a piezo-electric fatigue
system eliminates this problem because it is easy to get zero
displacement at the location where the sonotrode under pres-
sure crosses the wall of the autoclave.
74 Chapter 3

Figure 3.19 Special screw and the connection.

Figure 3.20 Test system of thin sheet.

Testing Machines and Their Performance 75

Figure 3.21 Autoclave description.

Figure 3.22 Wöhler Curve–INCONEL 718:Effect of hydrogen

pressure.
76 Chapter 3

A high pressure piezo-electric fatigue machine that

works under a pressure up to 300 bar has been built in Bath-
ias’s laboratory. The design is shown in Figure 3.21.
With this device, it has been shown that hydrogen under
a pressure of 100 bars has an effect on the S-N curve of IN 718
at room temperature. In Figure 3.22, S-N data in hydrogen
and in helium are compared to show the effect of hydrogen
between 106 and 109 cycles.

3.9. NON-AXIAL TEST EQUIPMENT

3.9.1. Ultrasonic Fretting Fatigue Testing
Fretting fatigue is generally promoted by high frequency low
amplitude vibratory motions and commonly occurs in
clamped joints and shrunk-on components (Lindley, 1997).
The surface damage produced by fretting can take the form
of fretting wear or fretting fatigue where the material’s fati-
gue properties can be seriously degraded. Some practical
examples of fretting fatigue failures are observed in wheel
shaft, steam and gas turbines, bolted plates, wire ropes, and
springs.
Fretting fatigue is a combination of fretting friction and
the fatigue process and involves a number of factors including
the magnitude and distribution of contact pressure, the
amplitude of relative slip, friction forces, surface conditions,
contact materials, cyclic frequency, and environment. Great
efforts have been made to quantify fretting fatigue in terms
of these factors, but limited success has been achieved. More
often, fretting fatigue characteristics are studied in the
laboratory experimentally by using a contact pad clamped to
a fatigue specimen in order to determine S-N curves with fret-
ting and thereby to establish the fatigue strength reduction
factor for a particular material. But these studies, generally
performed on the conventional tension–compression fatigue
machine with a low frequency, have some drawbacks:

1. The slip amplitude of fretting fatigue is usually

coupled with the fatigue stress, and to change the
Testing Machines and Their Performance 77

slip amplitude, pads with different gauge lengths are

needed.
2. The frequency is low and is not appropriate to simu-
late the high frequency small elastic vibration cycles
of mechanical, acoustic, or aerodynamic origin. On
the other hand, in some industries such as the auto-
mobile and railway industries, the determination of
high cyclic fretting fatigue properties up to 108 or
even 109 cycles is necessary. This experiment is
bound to be time-consuming and uneconomical.

In Bathias’s laboratory, an ultrasonic fretting fatigue

test technique at a frequency of 20 kHz has been developed,
in which the fretting slip amplitude can be changed without
changing the fretting pads. Experiments were performed on
a high strength steel and the results were analyzed.
The fretting pad has a cylindrical gauge profile. It is made
of the same materials as the specimen. A pair of opposing
pads are held on the sides of the specimen by springs.
Figure 3.23 shows a schematic diagram of an experi-
mental set-up, consisting of two parts. The first part is the
ultrasonic fatigue test machine that has been widely used in
fatigue tests for both endurance and crack propagation. Each
element in the machine is designed to have a resonant
frequency of about 20 kHz and an automatic unit maintains
the whole system operating at the resonant frequency. The
second part is a fixture to hold the two cylinder pads pressed
against the specimen by two springs. The normal contact
force is measured and controlled by the displacement of the
springs. Moreover, the use of the springs eliminates a
discernible changes in load should wear occur. The whole
experimental system is controlled by a PC.
The specimen of ultrasonic fretting fatigue has a cylind-
rical form with uniform section and is longitudinally asym-
metric to amplify the fatigue stress in the gauge length (see
the distribution of the vibration displacement and stress in
Figure 3.24). The specific length L is determined by the
requirement that the resonance frequency of the specimen
in the first mode of longitudinal vibration is 20 kHz:
78 Chapter 3

Figure 3.23 Schematic experimental system for ultrasonic fret-

ting fatigue.

p  L1 L2 
L ¼ X1 þ X2 ¼ S þ
k S1 S2
pffiffiffiffiffiffiffiffiffiffiffi
where k is a material constant, k ¼ 2pf r=Ed , S is the section
area of the cylinder. Figure 3.25 shows the details of the fret-
ting system.
The test system has the functions of regulating test para-
meters and recording the relative slip amplitude, normal
force, and stress. By changing the position of pads along the
specimen axis, the desired value of relative slip amplitude
can be obtained. The stress of fretting fatigue test is deter-
mined by the position of the pads for the load ratio R ¼
1.
For R >
1, the total stress is the superposition of static
Testing Machines and Their Performance 79

Figure 3.24 Pad and specimen.

80 Chapter 3

Figure 3.25 Fretting system for piezo-electric machine.

stress applied by the traction machine and the dynamic stress

of vibration.
The geometry of the pad and a typical specimen are
illustrated in Figure 3.25. With this machine, we can choose

Figure 3.26 Fretting system.

Testing Machines and Their Performance 81

independently the fatigue displacement and stress in a single

specimen. According to the position of pads in the specimen
axis, the displacement that provokes the fretting can be
chosen between 0.1 mm and tens of microns. By a gauge
mounted on the specimen at point xi ¼ 0, vibration deforma-
tion of the specimen can be measured.
Figure 3.26 shows another proposed fretting system.
Here, the normal force is applied by a ring and two screws.
Two small spheres are used to avoid the tipping of the pads
with the screws and to maintain a consistent distribution of
the contact pressure on the specimen.

3.9.2. Fretting Wear Testing

A study of fretting wear (Mason, 1982) is an early example of
the use of ultrasonic fatigue techniques. The purpose was to
explore the possibilities of using ultrasonic techniques as a
means of achieving accelerated fretting wear testing condi-
tions and to study how the increased severity of the contact
conditions would affect the fretting. As described in (Mason,
1982), during testing the vibrating specimen is clamped
between two stationary specimens. The upper one is mounted
on a traveling yoke that slides on two vertical rails. The
desired normal load is applied by simply mounting dead
weights on the yoke. As for the lower stationary specimen,
it is fixed in position and used only as a support for avoiding
high bending stress in the vibrating specimen.
It is worth recalling that the reason for Mason’s pioneer
work on the ultrasonic fatigue machine was to study fretting
wear (Mason, 1982).

3.9.3. Torsion Fatigue Testing

In a review article, Stanzl indicated that recently a new tech-
nique and equipment have been developed (Stanzl, 1986) that
allow one to perform torsion fatigue testing at ultrasonic fre-
quencies. The mechanical parts of the equipment must be
designed so that the torsion resonance vibration can be gener-
ated (Figure 3.27). Because the shear modulus is smaller than
Young’s modulus, all vibrating parts, including the speci-
82 Chapter 3

Figure 3.27 Ultrasonic torsion test equipment in Bathias’s

laboratory.

mens, must be smaller in order to obtain the resonance.

Besides this difference, the other experimental details such
as amplitude measurement and control are much the same
as those for the axial ultrasound fatigue loading. The super-
position of axial load is possible and has been investigated
in experiments on ceramic materials. Superposition of small
compressive loads leads to a lifetime twice as long as that
of pure cyclic 20 kHz torsion loads because of the increased
friction forces (Mayer, 1994).

3.9.4. Three-Point Bending Fatigue Testing

The three-point bending ultrasonic fatigue testing system
developed in Bathias’s laboratory is illustrated in Figure
3.28 (Bathias, 2002).
This system was developed for testing certain aluminium
alloy-based metal-matrix composites used in the automobile
industry. Figure 3.29 shows the variation of the displacement
amplitude from the converter to the specimen, and Figure
Testing Machines and Their Performance 83

Figure 3.28 System for ultrasonic fatigue experiments in three-

point bending.

Figure 3.29 Variation of displacement amplitude along the acous-

tic wave train.
84 Chapter 3

Figure 3.30 Specimen dimension and the first bending vibra-

tional mode.

3.30 shows the geometry of the specimen and the first bending
vibrational mode.
The solution procedure for the eigenvalue problem (the
resonance length L for bending the specimen of Figure 3.30)
is given below.
The free flexural wave equation for a beam of uniform
section is

@ 4 uðx; tÞ @ 2 uðx; tÞ
EI þ rhb ¼0 ð3:1Þ
@x4 @t2
EI is the flexural rigidity of the beam, and
bh3
I¼ ð3:2Þ
12
Separating variables by
uðx; tÞ ¼ UðxÞ sinðotÞ ð3:3Þ
we have the equation for U(x)
@ 4 UðxÞ

k4 UðxÞ ¼ 0 ð3:4Þ
@x4
where
 1=4
12o2 r
k¼ ð3:5Þ
Eh2

The general solution of Eq. 3.4 takes the form

Testing Machines and Their Performance 85

UðxÞ ¼ C1 sinðkxÞ þ C2 cosðkxÞ þ C3 sinhðkxÞ þ C4 coshðkxÞ

ð3:6Þ
Considering boundary conditions

UðxÞ ¼ Uð
xÞðsymmetryÞ ð3:7aÞ

ðU 00 ðxÞÞx¼L ¼ 0 ðzero moment at free endÞ ð3:7bÞ

ðU 000 ðxÞÞx¼L ¼ 0 ðzero shear force at free endÞ ð3:7cÞ

We can determine constants C1, C3

C1 ¼ C3 ¼ 0 ð3:8aÞ

and have the relations

cosðkLÞ
C4 ¼ C2 ð3:8bÞ
coshðkLÞ

tanðkLÞ þ tanhðkLÞ ¼ 0 ð3:8cÞ

So, we have from Eq. 3.6

 
cosðkLÞ
UðxÞ ¼ C2 cosðkxÞ þ coshðkxÞ ð3:9Þ
coshðkLÞ

Other conditions the vibration mode must satisfy are

Uð0Þ ¼ A0 ð3:10aÞ

UðL0 Þ ¼ 0 ð3:10bÞ

These give
 
coshðkLÞ
C 2 ¼ A0 ð3:11aÞ
cosðkLÞ þ coshðkLÞ
86 Chapter 3

Table 3.1 Dimensions of Ultrasonic Fatigue Specimen of Three-

point Bending
h  b (mm2) 2L0 (mm) 2L (mm) E (GPa) r (kg=m3)

SiCp=2124 4 7 21 38 131 2800

4 10 21 38
8 5 30 54
SiCw=AC4CH 4 7 20 36 101 2800
4 10 20 36
8 5 28 50
SiCw=AC8C 4 7 20 36 110 2800
4 10 20 36
8 5 28 51
Al2O3=AS7G06 4 7 19 35 93 2750
4 10 19 35
8 5 27 49

cosðkLÞ
cosðkL0 Þ þ coshðkL0 Þ ¼ 0 ð3:11bÞ
coshðkLÞ
Substituting Eq. 3.11a into 3.9 we find the solution
 
coshðkLÞ
UðxÞ ¼ A0
cosðkLÞ þ coshðkLÞ
 
cosðkLÞ
 cosðkxÞ þ coshðkxÞ ð3:12Þ
coshðkLÞ

Resolution of transcendental Eqs. 3.8c and 3.11b give the

resonance length of the specimen
 2 1=4
Eh
2L ¼ 0:506925 ð3:13aÞ
rf 2
and
 2 1=4
Eh
2L0 ¼ 0:27966 ð3:13bÞ
rf 2
Table 3.1 presents the dimensions of three-point bending
ultrasonic fatigue specimen.
4
S-N Curve and Fatigue Strength

4.1. INTRODUCTION

The ultrasonic fatigue technique can be used in traditional

fatigue testing, often more economically and efficiently than
other techniques. However, only ultrasonic technique is
practical for reaching very high number of cycles of fatigue
load, for example in the gigacycle regime.
Chapters 4 and 5 give some experimental and numerical
results obtained by ultrasonic fatigue technique. They are
mostly related to the S-N curve and fatigue limit as well as
the crack growth and threshold in terms of the stress inten-
sity factor (SIF). Some results are compiled to provide a data-
base for practical and industrial use.
The materials studied by ultrasonic fatigue (typically
20 kHz) include:

Ferrous materials 4240U, 4240R, SGI52, Cr-V, Cr-Si,
steel 304, steel 17-4PH, mild steel

Titanium alloys Ti6246, T6A4V

Nickel alloys Udimet 500, Inconel 706, N18

87
88 Chapter 4

Aluminium alloys AC4CH, AC8C, Al2024, AlSiII,

Al6061-T6, Al-Li8090

Polycrystalline copper.

These high-performance materials with a variety of

microstructures are widely used in many industries such as
aeronautics, aerospace, automotive, and railway. They are
important materials in manufacturing key equipment such
as helicopter cyclic trays, turbine engines, cryogenic pumps,
disks, and blades, some of which do operate in ultrasonic
vibration conditions.
Test environments vary greatly. The load ratio R ranges
from –1 to 0.9, and the temperature from as low as 20 K to as
high as 700 C. The necessity for testing in such wide ranges
of environment has yielded not only the state-of-the-art of
ultrasonic fatigue technique, but also a practical guide for
researchers and engineers.
Safe-life design based on the infinite-life criterion was
initially developed in the 1800s through the early 1900s, an
example being the stress-life or S-N approach related to
the asymptotic behavior of steels. Many materials display
an apparent fatigue limit or ‘‘endurance’’ limit at a high num-
ber of cycles (typically >106). Other materials do not exhibit
such a limit, but instead display a continuously decreasing
stress-life S-N curve, even at a great number of cycles (106
to 109). Therefore, each point of an S-N curve is more appro-
priately designated as a fatigue strength at a given number of
cycles (Figure 4.1).
Time and cost constraints usually rule out the use of con-
ventional fatigue tests for more than 107 cycles to evaluate
structural materials. In contrast to conventional fatigue tests
that require a long duration of test to reach such high num-
bers of load cycles with low frequencies (typically <50 Hz),
the piezoelectric fatigue technique described herein operates
at much higher frequency (about 20 kHz). Therefore the time
required for measurements in the high cycle fatigue range is
reduced to a small fraction of that required for a conventional
test.
S-N Curve and Fatigue Strength 89

Figure 4.1 ASM handbook S-N data and fatigue limit modeling.

The form of the S-N curve between 106 and 1010 cycles is
another method that can be used to help in the prediction of
risk in fatigue cracking.
Since Wohler’s work, the S-N curve has been convention-
ally represented by a hyperbola or one of its modifications as
indicated below:
Hyperbola: ln Nf ¼ ln a  ln sa
Wohler: ln Nf ¼ a  bsa
Basquin: ln Nf ¼ a  b ln sa
Stromeyer: ln Nf ¼ a  b ln (sa  c)

The extrapolation of the life range between 106 and 109 cycles
has created safer modeling.
For the reasons stated above, it is necessary to apply
an accelerated fatigue testing method or ultrasonic fatigue
technique to investigate the full behavior of the S-N curve.
We give the results of gigacycle fatigue of several typical
alloys in the following discussions.
90 Chapter 4

4.2. FERROUS MATERIALS

The S-N curves of steel are said to be asymptotic after 106. In

fact, the decrease between megacycles (106) and gigacycles
(109) continues up to 109 cycles; there is no proof to date that
this will not also happen beyond 1010 cycles, even for carbon
steels. To investigate the fatigue behavior in this range,
several ferrous materials have been tested by the ultrasonic
fatigue method.
High strength steels and spring steels with tensile
strengths ranging from 1000 MPa to 1800 MPa were tested
between 105 and 109 cycles through ultrasonic fatigue test
at 20 kHz. The S-N curves for the steels were obtained, indi-
cating that the specimens continued to fail over 107, even
108 stress cycles.
The mechanisms of fatigue cracks have been determined
by microfractographic analyses with the scanning electron
microscope. The fatigue strength is also estimated by the
Murakami model, which has been applied to the fatigue of
many high strength steels with nonmetallic inclusions or
small defects. The model does not fit the data well for the
gigacyclic fatigue; therefore, modified empirical formula has
been proposed to predict the high cycle fatigue life of high
strength steels (Bathias, 2001).
Investigations on long fatigue lives (>107 cycles) are
relatively rare. The reason is obvious; the time and costs are
prohibitive to perform the fatigue tests over 108 cycles using a
conventional testing machine.
The experimental results have shown that when fatigue
fracture does occur beyond 107 cycles in the steels, the origin
of this fracture is not at the surface but at the interior of the
specimen.
Indeed, in the high and low cycle regimes, the sites of fati-
gue crack initiation are different. In the HCF (>107 cycles)
regime, the initiation sites were found at non-metallic inclu-
sions located in the interior of the specimen. The initiation sites
were found at the surface for higher stress low cycle fatigue.
A modified Murakami model, which evaluates the effects
of non-metallic inclusions and small defects as well as the
S-N Curve and Fatigue Strength 91

Vickers hardness for a specified number of cycles, was proposed

in this study. This model predicts the fatigue strength of high
strength steels more accurately (see Sections 4.2.1 and 7.5.2).

4.2.1. High Strength Steel

Two low-alloy–high-strength steels, 4240U and 4240R, have
been studied (Wang, 1998). Specimens are characterized by
the differences in the S-content, viz 0.024 wt% for 4240U,
and 0.087 wt% for 4240R, and in the tempering temperature,
viz 600 C for group B and 425 C for group C. The details are
listed in Tables 4.1 through 4.3.
The specimens were tested at ultrasonic fatigue frequ-
ency 20 kHz with a stress ratio R ¼ 1 under load control. The
samples were polished by using #500, 1200, 2400, and 4000
papers. The central part of the specimen was cooled by com-
pressed air and the temperature was kept at about 70 C. The
fatigue results obtained are presented in Figure 4.2 and 4.3.

Table 4.1 Chemical Compositions of Materials (wt%)

C Mn P S Si Al Ni Cr Cu Mo

4240U 0.428 0.827 0.012 0.024 0.254 0.023 0.173 1.026 0.21 0.224
4240R 0.412 0.836 0.015 0.087 0.242 0.023 0.186 1.032 0.209 0.164

Table 4.2 Heat Treatments

Rep B Austentization: 950 C; Oil quenching; Temper: 600 C
Rep C Austentization: 950 C; Oil quenching; Temper: 425 C

Table 4.3 Mechanical Properties

E (Gpa) r (Kg=m3) sm (MPa) HV (30)

4240U-Rep B 211 7820 1100 345

4240R-Rep B1 211 7820 1040 320
4240U-Rep C 216 7870 1530 465
4240R-Rep C1 216 7870 1485 450

sm ¼ yield strength corresponding to upper yield point.

92 Chapter 4

Figure 4.2 S-N curves of 4240U.

The experimental results show that the fatigue failure of

high strength steels may occur beyond 107 cycles. There is
apparently no horizontal asymptote between 106 and 109
cycles, where the fatigue limit decreases by 60 MPa. However,
the scatter of the results seems very large. In fact, two dis-
tinct initiation mechanisms are acting: up to 106 cycles the
crack initiation occurs at the surface; beyond 108 cycles, the
initiation takes place at the interior. The fatigue life can be
substantially different depending on the mechanisms (Wang,
1998).
A typical subsurface crack initiation site in the 4240 low
alloy steel is shown in Figure 4.4. The stages of crack initia-
tion, stable crack propagation, unstable crack propagation,
and final failure are well defined. Fracture surfaces at all
the subsurface crack initiation sites appeared flat and
smooth. The fracture origin was identified by use of energy
S-N Curve and Fatigue Strength 93

Figure 4.3 S-N curves of 4240R.

dispersive analysis. In the high-cycle regime (>107 cycles), all

the initiation sites were found at non-metallic inclusions
located in the interior of the specimen. The chemical compo-
sition of most inclusions was sulphide. The sizes of the
inclusions range from 10 to 40 mm.

Figure 4.4 Subsurface initiation in 4240 steel.

94 Chapter 4

There are few models that can predict the effect of non-
metallic inclusions on fatigue strength. This may be because
adequate and reliable quantitative data on non-metallic inclu-
sions are hard to obtain. Murakami and co-workers (1994,
2002) have investigated the effects of defects, inclusions,
and inhomogeneities on fatigue strength of high strength
steels, and expressed the fatigue limit as a function of Vickers
hardness and the square root of the projection area of an
inclusion or small defect. Their formula is
 
CðHv þ 120Þ 1  R a
sw ¼ pffiffiffiffiffiffiffiffiffiffi 1=6
ð areaÞ 2
pffiffiffiffiffiffiffiffiffiffi
where Hv is Vickers hardness (in Kgf=mm2), area in mm,
C ¼ 1.43 for surface inclusion or defect, 1.56 for interior inclu-
sion or defect, 1.41 for inclusion or defect just below surface,
and a ¼ 0.026 þ Hv104.
This model does not specify the effect of number of cycles.
Some materials, such as most non-ferrous and aluminium
alloys, do not exhibit a fatigue limit. Instead their S-N curves
continue to drop at a slow rate at a high number of cycles. For
these materials, the fatigue strength rather than the fatigue
limit should be reported. We propose below an empirical for-
mula that does include the number of cycle to estimate high
cycles fatigue life of high strength steels
 
bðHv þ 120Þ 1  R a
sw ¼ pffiffiffiffiffiffiffiffiffiffi 1=6
ð areaÞ 2

where b ¼ 3.09–0.12 ln Nf for interior inclusion or defect and

b ¼ 2.79–0.108 ln Nf for surface inclusion or defect.
Table 4.4 compares the fatigue limits predicted by
Murakami model and the modified model with experimental
data for high strength steels and nickel base alloys.

4.2.2. Spring Steels

Two spring steels are tested to obtain their gigacycle fatigue
strength. Their chemical compositions and mechanical prop-
erties are listed in Tables 4.5 and 4.6.
Table 4.4 Comparison of Predicted Fatigue Strength with Experimental Data
SUP9TM1 SUP10M SUP10M
4240C-5 4240C-3 4240C-11 4240B-10 4240C1-7 [8] 3[8] 6[8] N18 [9] Cr–Si

Hv 465 465 465 345 450 445 550 554 445 500
S-N Curve and Fatigue Strength

p
Nfffiffiffiffiffiffiffiffiffiffi 5.75e8 8.76e7 7.12e8 4.92e5 2.59e5 4.5e5 2.0e7 1.63e6 1.45e7 1.7e8
area 20 16 13 25 20 60.1 14.1 28.9 53 25
h (mm) 900 135 25 0 0 0 0 240 350 650
sexp 760 740 750 630 760 588 862 883 550 780
sw (1) 555 575 595 390 495 408 673 589 417 566
sw (2) 724 787 775 592 763 621 862 902 588 762
Err(1)% 27 22 21 38 35 30.6 21.9 33.3 26 27
Err(2)% 4.7 6.4 3.3 6.0 0.4 5.6 0 2.2 6.9 2.3

sexp ¼ experimental fatigue strength (MPa); sw(1), sw(2) ¼ fatigue strengths estimated by Murakami model and modified model;
Err% ¼ (sw  sexp)=sexp.
95
96 Chapter 4

Table 4.5 Chemical Compositions of Spring Steels (wt%)

C Si Cr V Mn S P

Cr–V 0.48–0.53 0.1–0.4 0.8–1.1 0.15 min 0.7–1.0 0.04 max 0.035 max
Cr–Si 0.51–0.58 1.2–1.6 0.6–0.8 0.6–0.8 0.04 max 0.035 max

Table 4.6 Mechanical Properties of Spring Steels

UTS (MPa) E (Gpa) r (kg=m3) Ef % Hv

Cr–V 1800 210 7850 435

Cr–Si 1800 210 7850 35 500

Specimens are in the form of normalized hot-cooled

6.5 mm diameter wire. The Cr-V and Cr-Si wires are suitable
for service under shock loads at moderately elevated tempera-
tures, and the latter has better relaxation resistance and can
work at temperatures as high as 245 C. Tests are performed
at 20 kHz frequency with a stress ratio R ¼ 1 under load con-
trol and at ambient temperature. During testing, the middle
section of the specimen is cooled by compressed air and the
temperature is kept at about 70 C. The fatigue strength is
determined in the life range of 106  109 cycles as shown in
Figures 4.5 and 4.6.

Figure 4.5 S-N data of Cr-V steel at R ¼ 1.

S-N Curve and Fatigue Strength 97

Figure 4.6 S-N data of a spring steel Cr-Si steel at R ¼ 1.

Fatigue crack initiations are observed at the sites of

internal defects for fatigue life beyond 107 cycles. Using the
Paris law, the number of cycles to propagate a crack from
the defect to the surface (Np) can be estimated, assuming
the stress intensity factor to be:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffi
DK ¼ 0:5Ds p area

around an internal non-circular defect and

2 pffiffiffiffiffiffi
DK ¼ Ds pa
p
for a circular crack, the number of cycles to the initiation, Ni,
is estimated as:

Ni ¼ Nf  N p

The portion of fatigue life contributing to crack initiation is

estimated to be greater than 90% in the high cycle regime
for these steels (Wang, 1998).
Fatigue fractures beyond 107 cycles were observed. The
fatigue strength at 107 cycles is 860 MPa for Cr–Si steel,
and better than 810 MPa for Cr–V steel. At 109 cycles, the
98 Chapter 4

strength is about 800 MPa for Cr–V steel and 770 MPa for
Cr–Si steel.
Fatigue life experiments on ferrous materials show an
important difference in crack initiation between the gigacyclic
fatigue and the low cycle fatigue. In the latter, crack initiation
is the result of local plastification around surface disconti-
nuities.The local plastic deformation brings about the multi-
plication of dislocations due to cyclic hardening of the
material. Contrary to this, in the gigacyclic fatigue, the site
of crack initiation is observed at the interior rather than at
the surface. An explanation may be that, in the gigacyclic
fatigue, the applied stress is too small to provoke a cyclic
plastic zone localized at the surface, and the internal defects
are likely to become the main sources of crack initiation. In
the case of ferrous materials, the crack initiation from
inclusions takes place 50 mm  1000 mm under the surface,
which is followed by nucleation and circular propagation of
micro-cracks. The diameter of the initiation zone is about
50 mm  100 mm. A macro-crack is formed and propagates
circularly until this zone approaches the specimen surface.
From these initial observations, we concluded that the
mechanism of gigacyclic fatigue needs to be further
explored.
For additional comparison of test results, the fatigue life
experiments of other ferrous alloy (i.e., steel 17-4PH and
12%Cr) are discussed in the next section.

4.2.3. Martensitic Stainless Steels

Steel 17-4PH is also tested at 20 kHz and R ¼ 1, with the
maximum strain controlled constant in the middle section of
the specimens (Bathias, 2001). Some samples are used for
conventional fatigue tests with a loading frequency of
20 Hz  50 Hz under a push–pull stress alternating cycle.
Another martensitic stainless steel, 12% Cr steel, is
tested with the same conditions in order to compare the
results with 17-4PH steel data. Chemical composition and
mechanical properties after quenching and tempering
(620 C for 2H) are given in Tables 4.7 and 4.8.
S-N Curve and Fatigue Strength 99

Table 4.7 Chemical Composition of 12% Cr Steel (wt%)

C Si Mn P S Cr Mo Ni Al

0.11 0.11 0.77 0.01 0.002 11.4 1.47 2.53 0.064

Table 4.8 Mechanical Properties 12% Cr Steel

E (Gpa) UTS (MPa) sy (MPa) Ef Ra

216 958 829 64% 18%

Figure 4.7 gives the gigacycle fatigue S-N curves for

these two martensitic stainless steels. It is shown that failure
can occur between 109 and 1010 cycles in both of these materi-
als. The location of the initiation is on the surface up to 107
cycles and in the subsurface beyond this life. The results
obtained here do seem to suggest that there is no lower limit
of fatigue strength, even for ferrous alloys, and that they do

Figure 4.7 Experimental results of 17-4PH and 12% Cr steels.

100 Chapter 4

Figure 4.8 Gigacycle initiation in 17-4 PH steel.

undergo rupture even after a large number of cycles at low

fatigue stresses.

4.2.4. Bearing Steels

Two bearing steels were tested in gigacycle fatigue; their
chemical compositions and mechanicals properties are given
in Tables 4.9 and 4.10.
Specimens were tested at 20 kHz with a load ratio
R ¼ 1. The number of maximum testing cycles is limited to
1010 so that, when cycles pass 1010, the testing machine will
stop automatically. Since heat is induced in the specimen by
absorption of ultrasonic energy, during the testing process

Table 4.9 Chemical Compositions of SUJ2 and 100C6 (wt%)

C Si Mn P S Cr Cu Ni Mo

SUJ2 1.01 0.23 0.96 0.012 0.007 1.45 0.06 0.04 0.02
100C6 0.35–1.1 0.15–0.35 0.20–0.4 20.025 20.015 1.35–1.60 20.10

Table 4.10 Mechanical Properties of SUJ2 and 100C6

E UTS (MPa) r Hv30

SUJ2 210 2316 7.86 778

100C6 213 2500 7.45
S-N Curve and Fatigue Strength 101

Figure 4.9 Fatigue results for 100C6 and SUJ2 bearing steels up
to 1010 cycles.

the middle of the specimen is cooled by compressed air and

the temperature is kept at 70 C in the megacycle regime
and at 25 C in the gigacycle regime.
The stress concentration effect was studied with notched
specimens. ANSYS was used to calculate the stress field at
the root of the notch. The diameters of the notch specimen
are 9.2 mm and 6.4 mm. The results are given in Figure 4.9.
Several specimens failed between 109 and 1010 cycles
with crack initiation at an internal inclusion. A notch effect
is observed in the gigacycle regime. The fatigue strength of
100C6 steel at 1010 cycles is 800 MPa without notch and
600 MPa with notch. Also a large scatter of the data is
observed at these high cycles.

4.2.5. Low Carbon Ferritic Steel Thin Sheets

Our piezo-electric fatigue machine was adapted to test speci-
mens of a small thickness, less than 1 millimeter. Results for
a low carbon steel are given here (Wang, 1993).
The chemical composition and the mechanical properties
of this low carbon steel is given in Tables 4.11 and 4.12.
102 Chapter 4

Table 4.11 Chemical Composition of Low Carbon Steel (wt%)

Co C Mn Si P S Al

0.14 0.08 0.4 0.1 0.025 0.025 0.02

Table 4.12 Mechanical Properties of Low Carbon Steels

E (GPa) sy (MPa) UTS (MPa) Ef r Hv5

203 225 340 36 7.83 95

Figure 4.10 S-N curve for a low carbon steel at 20 kHz, R ¼ 0.1.

The specimens, after polishing, are tested with a load

ratio R ¼ 0.1. The temperature of the specimen during the test
is kept between 50 C and 70 C. The results in Figure 4.10
show that a low carbon fenitic steel can fail at 5  108 cycles
and do not confirm an asymptotic nature of S-N curve beyond
109 cycles. The fatigue strength at 109 cycles is close to
220 MPa.

4.2.6. Austenitic Stainless Steel

In addition to a ferritic steel, an austenitic 304 stainless steel
was tested in the gigacycle regime. In the test, the specimens
were cooled with tap water to prevent temperature elevation.
S-N Curve and Fatigue Strength 103

Figure 4.11 Quasi-asymptotic S-N data for 304 stainless steel.

The results presented in Figure 4.11 show that there is

an apparent asymptote between 106 and 109 cycles. Some spe-
cimens failed beyond 108. However, since the surface of the
specimens was not polished, the scatter is large. The fatigue
strength at 109 cycles is 198 MPa. For this steel, the initiation
was always observed at the surface.

4.2.7. Spheroidal Graphite Cast Iron

In the automotive industry, the designed fatigue life of compo-
nents often exceeds 109 cycles. Spheroidal graphite iron or
ductile cast iron is a favored material for fabrication of some
of these components because of its exceptional combination
of high strength and ductility. In the literature, few data on
the S-N curve of spheroidal graphite cast iron have been
obtained beyond 107 cycles, as the test time and cost to
perform fatigue tests of over 108 cycles using a conventional
fatigue machine are extremely high (Wang, 1998).
The chemical composition and mechanical properties are
listed in Tables 4.13 and 4.14.
High cycle fatigue S-N data of SGI52 at R ¼ 1 and R ¼ 0
are presented in Figures 4.12 and 4.13.
For zero mean stress R ¼ 1, the results show no notice-
able frequency effect on the fatigue behavior between 25 Hz
and 20 kHz. The ultrasonic fatigue data closely match the
104 Chapter 4

Table 4.13 Chemical Composition of Spheroidal Graphite Cast

Iron SGI52 (wt%)
C Si Mn S P Mg Cu Ni Mo Ti Cr Sn

3.45 3.21 0.13 0.019 0.031 0.031 0.024 0.59 0.013 0.043 0.02 0.030

Table 4.14 Mechanical Properties of Spheroidal Graphite Cast

Iron SGI52
E (GPa) sy (MPa) UTS (MPa) Ra r (kg=m3) Hv

179 380 510 14.5 7100 184

sy ¼ yield strength corresponding to 0.2% offset; Ra ¼ fraction of reduction in area

from a tensile test.

conventional fatigue data. At R ¼ 0, however, fatigue strength

in ultrasonic fatigue tests seems to be slightly higher than
that in conventional fatigue tests. It is evident that fatigue
failure can occur over 107 cycles, and the maximum fatigue
stress smax continues to drop with the increasing number of
cycles between 106 and 109.
It is also interesting to investigate the temperature of the
specimens during the test. Heating in the specimens is caused
by absorption of ultrasonic energy. During the testing pro-
cess, the temperature in the middle section of the specimen

Figure 4.12 S-N data of SGI52 (R ¼ 1).

S-N Curve and Fatigue Strength 105

Figure 4.13 S-N data of SGI52 (R ¼ 0).

is controlled between 50 C and 90 C for tests at R ¼ 0, and

70 C and 120 C for tests at R ¼ 1. Temperature evolution
(Figure 4.14) at R ¼ 0 consists of two periods:

Period 1: There is a steep rise near 106 cycles and the

maximum temperature depends on the amplitude of ultra-
sonic fatigue loading.

Period 2: There is a horizontal curve over certain num-

bers of cycles. This may be interpreted as an equilibrium state
between the dissipation of ultrasonic energy due to interior
crack nucleation and the heat induced by interior friction of
the material.
After the maximum temperature is reached, no signifi-
cant decrease is observed over certain numbers of cycles when
the initiation of crack takes place on the surface. When the
initiation occurs in the interior of the specimen, the maximum
temperature probably corresponds to the nucleation; this is
an illustration of gigacycle fatigue. Compare the temperature
changes for smax ¼ 305 and 360 MPa, in Figure 4.14.

4.3. ALUMINIUM MATRIX COMPOSITE

Generally, for fatigue life up to 107 cycles, the aluminium

matrix composite materials reinforced by fibers have a higher
106 Chapter 4

Figure 4.14 Temperature effect.

fatigue limit than those reinforced by particles. Comparing

composites with alloys, the resistance gained can reach
20% when the processing is done with powder metallurgy.
Some results (Bathias, 1996) show that the composite
S-N Curve and Fatigue Strength 107

2080=15%SiCp has a better fatigue resistance than alloy

7075T73 used for the manufacturing of helicopter cyclic
trays; this is illustrated in Figure 4.15. However, when the
S-N curves of Figure 4.15 are compared, we would expect
that these curves cross each other in the range of 106 to
107 cycles; it would be interesting to explore the gigacyclic
fatigue life.
Particular attention must be given to the form of fatigue
curves. A lot of fatigue tests are limited to 106 cycles because
machining specimens of Metal Matrix Composites (MMC) are
very expensive. Some tests by Jones (1991) carried out up to
107 cycles show the evidence that the S-N curve of 7075 is
much flatter than that of its alloy. The quasi-hyperbolic
shape in the semi-logarithmic plot of the S-N curve for alu-
minium alloys does not fit the S-N curve of the MMC. In
order to verify the forms of the S-N curves, Japanese
researchers performed fatigue tests on some MMCs until
108 cycles in rotating bending (Masuda, 1994). They showed
that no horizontal asymptote could be determined between

Figure 4.15 Comparison of S-N curves: (A) 7075 alloy (full curve);
(B) 2080=15%SiCp composite (filled circles).
108 Chapter 4

106 and 108 cycles, and that some ruptures occured between
107 and 108 cycles.
Our laboratory performed experiments on some alumi-
nium composites reinforced with 17% of SiC by squeeze
casting (Bathias, 1994). The compositions of the matrices
and mechanical properties of the materials are listed in
Tables 4.15 and 4.16.
The fatigue tests were carried out on prismatic bars in
three-point bending at 20 kHz frequency, ambient tempera-
ture, and R ¼ 0.1. The results presented in Figure 4.16 show
that some specimens were broken in fatigue between 108
and 109 cycles. Evidently, it is not possible to draw a
horizontal asymptote between 106 and 109 cycles. The fatigue
limit defined between 106 and 107 in the conventional
standard does not seem to exist in gigacyclic fatigue tests.
When the S-N curve of alloys is compared to the S-N curve
of composites, it is observed again that the resistance of alloys
is higher in low cycle fatigue and lower in gigacyclic fatigue.
From these observations we propose an empirical for-
mula for a good estimate of the equation of the S-N curves
for R ¼ 0.1 in the (following) form

smax ¼ UTS  Nfc

Table 4.15 Compositions of the Matrices of 17%SiC

Cu Si Mg Zn Mn Ti

AC4CH 0.113 7.124 0.338 0.007 0.008 0.132

AC8C 2.98 10.048 1.126 0.023 0.017 0.002

Table 4.16 Mechanical Properties of 17% SiC

UTS (MPa) E (GPa) Hv

AC4CH 762–789 101 150–160

AC8C 814–840 110 210–226

UTS ¼ ultimate tensile strength; Hv ¼ Vickers hardness.

S-N Curve and Fatigue Strength 109

Figure 4.16 Fatigue curves for MMC, R ¼ 0.1 (Masuda, 1994;

Bathias, 1996).

4.4. NON-FERROUS ALLOYS

To explore the gigacyclic fatigue behaviors of other metallic

alloys, several titanium and nickel alloys were selected as
examples.

4.4.1. Titanium Alloys

Titanium alloys play an important role in the aerospace
industry. It is generally accepted that titanium alloys behave
like steels in gigacyclic fatigue. This section examines struc-
ture–fatigue properties in titanium alloys (Bathias, 1994;
Jago, 1996, 1998) in which high cycle fatigue behaviors have
been shown to be significantly affected by microstructure.
Microstructures that have a small probability of low-stress
crack initiation, as in b-processed microstructures, generally
yield the best high cycle fatigue limit and tensile resistance.
110 Chapter 4

In this section we examine the effects of four thermo-mechan-

ical processes on stage I fracture mode in the Ti6246 (Ti-6Al-
2Sn-4Zr-6Mo) alloy—an a þ b titanium alloy used for
compressor disks and blades. The fatigue test programs are
so arranged that fatigue properties are examined in the range
of 107  109 cycles with R ¼ 1. The characteristics of fracture
mechanism in this alloy are examined by SEM (scanning
electron microscope) observations of fracture surface of each
broken fatigue-limit specimen. Microstructure features are
characterized by quantitative examination in the maximum
stress plane and the plane perpendicular to it.
Chemical compositions and microstructures obtained
by different thermal processes are given in Tables 4.17 and
4.18.
Quantification of the morphological aspects has been
performed to provide a comprehensive description of various
microstructures. Two orthogonal metallographic surfaces
are examined. The number of whole particles detected is
more than 2000. A global image analysis measures primary
a-phase volume fraction, total a-phase volume fraction,
thickness of primary a platelets, and mode distance between
coarser particles. Size and shape measurements are ana-
lyzed individually. This procedure provides, for example,
the perimeter, the area, and the longest dimension of each
particle.
Tensile tests have been performed in each TP condition.
The strain rate is equal to 8.4  105 s1 for all tests. Finally,
the yield strength, ultimate tensile strength, elongation, and
reduction of area have been measured.

Table 4.17 Chemical Composition of Ti6246

TP Al Sn Zr Mo C Cu Si Fe O H N
(No.) (wt%) (wt%) (wt%) (wt%) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm)

1(1) 5.76 1.97 4.08 3.97 90 <50 <50 400 930 44 80

1(2), 2
and 3 5.68 1.96 4.08 3.92 83 <50 <50 300 1100 28 70

TP ¼ Thermo-mechanical process.
S-N Curve and Fatigue Strength 111

Table 4.18 Thermo-mechanical Processes and Microstructures of

Ti6246
Final
TP Cast number Forging Thermal microstructure
(No.) (RMI) condition treatment type

1(1) 972886 955 C=WQ 935 C=2H=WQ þ a platelets and

and (Tb þ 10 C) 905 C=1H=Air þ b-transformed
1(2) 595 C=8H=Air matrix
2 982560 955 C=WQ 935 C=2H= Coarse platelets
(Tb þ 10 C) Slow cool Room T. and b-transformed
þ 595 C=8H=Air matrix
3 982560 905 C=Air 935 C=2H=WQ Bi-modal structure
(Tb  40 C) þ 905 C=1H=Air (a platelets and
þ 595 C=8H=Air a nodular) and
b-transformed
matrix

WQ ¼ Water quench.

All histograms of data exhibit the typical lognormal

distribution. Average and standard deviations for each
microstructure feature are given in Table 4.19.
Despite standard deviations, it is seen, on one hand, that
primary a platelets are significantly coarser in TP2 specimens
than in TP1(1) and TP1(2) specimens. That is, the average
particle thickness, longest dimension, and area are larger in
TP2 microstructure. In addition, the TP2 process increases
the primary a volume fraction and consequently decreases
the mean length between primary a particles. On the other
hand, the mean particle perimeter is longer in the TP1(2)
microstructure because of an a platelet shape difference with
TP1(1) and TP2 in lamellar phases. The bi-modal TP3 micro-
structure corresponds to a mixing of coarse nodular primary
a-phase (mean width, 3.1 mm; volume fraction, 20%)
and fine lamellar primary a-phase in b-transformed matrix
(Figure 4.17).
As for high cycle fatigue resistance at room temperature,
the role of the secondary a volume fraction seems to manifest
as an influence on yield strength without any differences
between lamellar and duplex structures in contrast with the
 112

Table 4.19 Quantitative Microstructure Features of Ti6246 Alloy

fap(%) dap (mm) db (mm) Pap=Sap (mm1)

TP condition (No.) lam glob fas(%) lam glob lam glob lam glob

1(1) 54  2 13  4 1.7  0.2 0.8  0.2 2.9  0.9

Axial view: 1(1) 37  2 15.8  4 1.7  0.2 2.5  0.2 2.6  0.8
Radial view: 1(1) 42  4 14.8  8 1.7  0.4 1.9  0.4 2.7  1.7
Volume: 1(2) 43.9  4 14.1  10 1.9  0.4 2.4  0.4 2.2  1.5
2 66.2  4 7.3  10 2.2  0.4 0.74  0.4 1.7  1.6
3 20  4 27.3  4 20  10 1.0  0.4 3.1  0.4 1.6  0.4 4.3  0.4 1.16  0.8 4.5  3.2

fap ¼ primary a-phase volume fraction; fas ¼ secondary a-phase volume fraction; dap ¼ thickness of primary a platelets; db ¼ mode dis-
tance between coarser particles; Pap ¼ perimeter of particle; Sap ¼ area of particle; lam ¼ lamellar microstructure; glob ¼ globular
microstructure.
Chapter 4
S-N Curve and Fatigue Strength 113

Figure 4.17 Microstructure of 6246 Ti alloy.

results of 77 K tests (that will be presented in Section 4.2.4.1).

At low temperatures, the presence of a coarse nodular phase
significantly reduces the fatigue limit resistance. Thus, at
cryogenic temperature, fas cannot be considered to be the sole
influence in a þ b processed materials.
Figure 4.18 presents high cycle fatigue results of Ti6246
at room temperature, which demonstrates a significant differ-
ence in S-N curves among different TP conditions.
The TP3 material has the highest fatigue resistance of
510 MPa at 109 cycles, while the TP1(1) shows a slightly lower
fatigue limit that is estimated as 490 MPa, and TP1(2) sam-
ples give a much lower limit of only 400 MPa . TP2 material
exhibits the lowest fatigue limit of 325 MPa.
At room temperature, the 0.2% offset yield stress and the
fatigue limit stress mainly depend on the secondary a-phase
volume fraction. The final aging process, producing the sec-
ondary a precipitation, is the same for all TP conditions and
gives an identical average width for the small needles. The
experimental results also show a greater frictional stress for
the primary a-phase than for b-phase. In comparison with
114 Chapter 4

Figure 4.18 S-N curves of Ti6246 at room temperature for differ-

ent TP conditions.

the yield stress, precipitation strengthening has a more

significant effect on fatigue limit resistance.
In addition to the above conclusions, the Ti6246 test
data, like gigacyclic fatigue results of other materials, show
that:

Some fatigue ruptures occur at least up to 109 cycles

There is no asymptote

Fatigue limit at 109 cycles is much lower than the
conventional fatigue limit at 106 cycles.
Another titanium alloy T6A4V was also tested at 20 kHz
and R ¼ 1 with the maximum strain kept constant in the
middle section of the specimens (Bathias, 1994). Chemical
compositions, heat treatments, and mechanical properties
of this material are included in Tables 4.20 through 4.22.
Some samples were used for conventional fatigue tests
S-N Curve and Fatigue Strength 115

Table 4.20 Compositions of U500 and T6A4V (wt%)

C Si Mn Al Cr Cu Fe Mo Ti Zr Co Ni

U500 0.12 0.75 0.75 3.0 18 0.7 4.0 4.2 3.0 0.06 19 Remains

V Fe O Ni H Y C Al Ti

T6A4V 3.5 0.25 0.13 0.05 0.01 0.005 0.08 5.5 Remains
4.5 6.5

Table 4.21 Heat Treatments of U500 and T6A4V

U500 Annealing 5 mn to 1080 C, quenching in oil, aging 24 h to
845 C, cooling in air 16 h to 760 C, cooling in air

T6A4V Quenching 1 h to 950  970 C, cooling in water, annealing 2 h to

700  10 C, cooling in air, stablizing (during machining) 2 h to
700  10 C

under push–pull stress alternating cycles at 20 Hz  50 Hz.

Figure 4.19 gives the ultrasonic fatigue test results in
comparison with test data obtained by other laboratories.
We notice that for 17-4PH and T6A4V fatigue resistance
behavior in ultrasonic fatigue regime seems to be slightly bet-
ter than that observed in traditional fatigue loading. That is,
materials seem to respond differently to ultrasonic fatigue
and conventional fatigue. Similar observations have been
mentioned in the literature (Bathias, 1994) for other materi-
als. At the time of this investigation, the fatigue behavior of
materials was considered to depend upon both the loading
regime and mechanical characteristics of the materials. In
ultrasonic fatigue tests, the specimen is excited at constant

Table 4.22 Mechanical Properties of U500 and T6A4V

sy UTS E Ed r
(MPa) (MPa) (MPa) (GPa) ef Ra (kg=m3) n ef

U500 826 1780 192 214 23.5 19 8020 0.29 0.2107

T6A4V 989 1190 108 110 15 42.5 4420 0.3 0.5516
116 Chapter 4

Figure 4.19 Experimental results of T6A4V.

strain amplitude. As is well known, most engineering materi-

als exhibit their maximum cyclic hysteresis properties when
excited in push–pull fatigue. There will be softening or hard-
ening according to the mechanical properties of the material.
One discovery in the literature (Laird, 1982) is that, in most
cases, softening occurs for materials in which the ratio of yield
strength sm to the elastic limit s0.2 is less than 1.2, and hard-
ening occurs when this ratio is greater than 1.4. Here we have
sm=s0.2 < 1.2 for alloy T6A4V. We may assume that materials
loaded in an ultrasonic fatigue regime exhibit the same cyclic
hardening characteristics as in conventional fatigue excita-
tion. We will find, in the next section, that alloy U500 with
sm=s0.2 > 1.4 hardens under constant strain amplitude and
the peak stress level increases. For 17-4PH and T6A4V, the
inverse effect occurs; that is, the peak stress level decreases
under constant strain cycling.
We then examined the fracture surfaces of specimens that
failed in ultrasonic and conventional fatigue tests by means of
SEM. Figure 4.20 shows some examples of crack initiation and
propagation configurations obtained in ultrasonic fatigue load-
ing. It is observed that cracks initiate from either a single
S-N Curve and Fatigue Strength 117

Figure 4.20 Crack initiation in ultrasonic fatigue loading.

surface point or a mass defect beneath the surface, then propa-

gate in a fan-shaped fatigue failure configuration. It is evident
that the surface condition of specimens tested in ultrasonic
fatigue loading influences experimental results due to high
exciting frequency and high strain rate. In this study, the spe-
cimens used in high frequency tests had been polished
mechanically to eliminate micro-cracks on the surface.

4.4.2. Udimet 500

A large family of nickel alloys used in turbine engines is no
exception to the rule: some fatigue ruptures occur in the
gigacyclic domain and the drop between the megacyclic and
gigacyclic fatigue limits is large.
Udimet 500 is a typical nickel super alloy whose chemical
composition, heat treatment, and mechanical properties are
listed in Tables 4.20 through 4.22. (These tables also include
the data for T6A4V discussed earlier.)
The experimental results (Ni, 1991) at 20 kHz and
R ¼ 1 are illustrated in Figure 4.21. The data indicate a
decrease of fatigue limit down to 200 MPa.

4.4.3. Aluminium Alloys

To close the review of the gigacycle fatigue behaviors of the
main alloy families, the cases of cast and wrought light alloys
118 Chapter 4

Figure 4.21 S-N data of alloy Udimet 500 (Dentsoras, 1983).

are presented here. As has long been assumed, there is no

asymptotic S-N curve endurance limit for aluminium alloys.
For 2219 and 6061 alloys, there exists about 100 MPa differ-
ence between the two fatigue strengths at 106 cycles and 108
cycles (Tao, 1996) for R ¼ 1. The difference is as large as
200 MPa for 2024–T351 alloy (Stanzl et al., 1993b) at 106
cycles and 109 cycles.
It is of interest to note that for cast aluminium and mag-
nesium, this difference is smaller and less than 100 MPa
(Figure 4.22) in spite of the presence of porosity in casting.
In the gigacycle regime, the initiation site is located
inside the specimens for titanium, nickel alloys, and high
strength steels. However, the initiation takes place at the
surface of specimens in aluminium and magnesium alloys,
especially in cast alloys where the void formed near or at
the surface is the main location for crack formation.

4.5. ALLOYS AT CRYOGENIC TEMPERATURES

We now examine the influence of temperature on fatigue

properties of turbine engines or cryogenic pumps. In this
S-N Curve and Fatigue Strength 119

Figure 4.22 Aluminium and magnesium alloys (A291) S-N data

with R ¼ 1 at 20 kHz (Ds ¼ 50 to 100 MPa) (Stanzl, 1996).

section, some alloys used for manufacturing engines of Ariane

pumps are discussed in the domain of the gigacyclic fatigue at
the temperatures of 77 K and 20 K.

4.5.1. Titanium Alloys

Titanium alloy Ti6246 is tested at 77 K, after the base test
data is obtained at room temperature (Jago, 1996; Tao,
1996). In these experiments, four microstructures are tested
to determine the effect of microstructure on the fatigue limit
at cryogenic temperature. Chemical compositions and differ-
ent microstructures obtained by different thermal processes
are given in Tables 4.17 through 4.19. In liquid nitrogen, high
cycle fatigue resistance depends mainly on the primary a
grain size and decreases greatly with the presence of coarse
globular primary a-phase. The resistance of gigacyclic fatigue
is distinctly higher at 77 K than that at room temperature:
640 MPa vs. 490 MPa for the best microstructure (Figure
4.23). There appears to be a significant effect of thermal pro-
cessing. The lowest fatigue limit of the material TP2 is
120 Chapter 4

Figure 4.23 S-N curves of Ti6246 in liquid nitrogen until 109

cycles (T ¼ 77 K, R ¼ 1).

explained by the presence of large primary a platelets due to

slow cooling after solution treatment. The highest fatigue
limit at 77 K is obtained with a fine microstructure. But it
is difficult to explain the difference between materials
TP1(1) and TP1(2). TP3 process which yields a duplex
structure, does not greatly increase the high cycle fatigue
resistance; the difference is only 50 MPa between those at
room and liquid nitrogen temperatures.
However, we must emphasize that the S-N curves do not
have a horizontal asymptote at cryogenic temperatures; some
fatigue ruptures are observed beyond 108 cycles depending on
the alloy microstructure.
We observe similar results in the T6A4V alloy. (Tao,
1996) gives S-N curves of T6A4V in gigacyclic range at
300 K, 77 K, and 20 K. Some fatigue tests in liquid hydrogen
S-N Curve and Fatigue Strength 121

Figure 4.24 Fatigue results of T6A4V in liquid H2 and He at

20 K.

carried out up to 108 cycles were extremely difficult. Regard-

less of the temperature, the environment (air, nitrogen,
helium, or hydrogen) and the load ratio (R ¼
1 or
R ¼
0.1), the fatigue ruptures can occur between 106 and
109 cycles (Figure 4.24) (Tao, 1996).

4.5.2. Aluminium-Lithium Alloys

The main reason for the replacement of conventional alumi-
nium alloys by aluminium-lithium alloys is to obtain the
lightweight structures demanded by the aeronautical and
space industries. For example, aluminium-lithium alloys
have been used in space vehicles for reservoirs of liquid
oxygen and hydrogen fuel.
The material studied is Al-Li8090. Tables 4.23 and 4.24
list the chemical composition and mechanical properties of
this alloy.

Table 4.23 Chemical Composition of AL-Li8090 (wt%)

Li Cu Fe Si Mn Cr Zn Ti Mg Zr Al

2.2  2.7 1.0  1.6 0.30 0.20 0.10 0.10 0.25 0.10 0.6  1.3 0.04  0.16 Remains
122 Chapter 4

Table 4.24 Mechanical Properties of Al-Li8090

T ( K) E (GPa) sy (MPa) UTS (MPa) ef r (kg=m3)

300 81 455 500 7 2350

77 86 2350

Tests were carried out at 77 K in liquid nitrogen and at

20 kHz (R ¼  1) (Tao, 1996). Figure 4.25 gives the S-N curve
in comparison with some literature data. There exists a differ-
ence greater than 100 MPa between the fatigue strengths at
106 and 108 cycles.
The crack initiation at 20 kHz and 77 K takes place
mainly under the surface, at the internal flaws of the mate-
rial. The initiation occurs at the surface as usual if the stress
is very high and the fatigue life is less than 106 cycles.
The failure surface of an Al-Li8090 specimen tested
at high frequency and low temperature shows a very brittle
rupture by cleavage, which results from the cryogenic

Figure 4.25 S-N curve of Al-Li8090 in comparison with other

results.
S-N Curve and Fatigue Strength 123

temperature and elongated microstructure developed by

tension during heat treatment. This forms very significant
textures.

4.6. N18 ALLOY AT HIGH TEMPERATURE

As the final part of this survey, we discuss the case of nickel

alloys chosen for manufacturing turbine disks that must be
resistant to the low cycle and megacyclic fatigue on different
portions of the disks. The N18 high temperature fatigue
seems to indicate that nickel base alloys can crack at 109
cycles (Bonis, 1997). Powder N18 is tried—with and without
seeding of inclusions—to determine the effect of the inclu-
sions in the gigacyclic domain.
The microstructure of turbine disks in powder N18 is
very homogeneous in the center with grain size of 7  10 mm.
In the periphery the grain is coarser (10  15 mm). The powder
is produced by atomization in argon, then sifted to 75 mm.
Strengthening is achieved by extruding the powder container.
This technique guarantees the high quality, reproducibility
and stability of the desired properties of the material. The iso-
thermal forging at 1120 C follows as the formation process.
The inclusions in seeded (or polluted) N18 are ceramic parti-
cles. There are 30,000 ceramic inclusions of Al2O3 and MgO
sifted to 70  80 mm in one kilogram of the alloy. In com-
parison, the standard N18 has fewer than 20 inclusions per
kilogram.
The chemical composition, mechanical properties, and
heat treatments of the superalloy N18 are given in
Tables 4.25 through 4.27.

Table 4.25 Composition of Superalloy N18

Cr Co Mo Al Ti Nb Hf C B Zr
(%) (%) (%) (%) (%) (%) (%) (ppm) (ppm) (ppm)

11.5 15.5 6.5 4.3 4.3 <1.5 0.5 200 150 300
124 Chapter 4

Table 4.26 Mechanical Properties of Superalloy N18

T ( C) E (GPa) n r (kg=m3)

0 220 0.31 8000

800 170 0.29 8000

Table 4.27 Heat Treatments of Superalloy N18

Solution 4 h at 1165 C; quenching in oil 90 s; aging 24 h at 700 C, cooling in
air; stabilizing 4 h at 800 C, cooling in air

From the test data in Figure 4.26, several points must be

emphasized to characterize the fatigue between 106 and 109
cycles.

The effect of inclusions is sometimes obscured by the
effect of porosity when R ¼ –1 or R ¼ 0. The scatter
of the results for R ¼ 0 in N18 seeded with inclusions
is more pronounced than that for standard N18.

It is very remarkable that, when R ¼ 0 or R ¼ –1, the
resistance to the gigacyclic fatigue at 450 C is
250 MPa for the N18 with or without inclusions.

Figure 4.26 S-N data of N18 nickel based alloy at 450 C between
106 and 1010 cycles.
S-N Curve and Fatigue Strength 125

On the other hand, when the static strain of the fati-

gue cycle is very high, (R ¼ 0.8), the effect of inclusions
becomes significant. Without inclusions, the N18
fatigue limit at 450 C is 155 MPa at 109 cycles. How-
ever, with inclusions it is 125 MPa . In practice, such
fatigue cycles occur in turbine disks.

Finally, between 106 and 109 cycles, the fatigue limit
can decrease by 30% depending on the load ratio R.

4.7. ROTATING-BENDING INTERNAL CRACK

STRESS CORRECTION

At this point, it is of interest to compare the fatigue curves in

rotating-bending and in tension-compression. In Japanese
literature, many results had been given for JIS SUJ2, by
Sakai (1999), Murakami (2002), and others. It has been found
by these Japanese researchers that the occurrence of the
internal initiation follows the appearance of a plateau in
rotating-bending loads (Figure 4.27). Similar curves have
been observed by Nishijima (1999) in other steels tested in
rotating-bending loads.

Figure 4.27 S-N curve of SUJ2 bearing steel (R ¼ 1) rotating-

bending test (Sakai, 1999).
126 Chapter 4

Although the SUJ2 is not equivalent to 100C6 bearing

steel, it seems that the step wise S-N curve is related more
to the rotating-bending behavior than to the steel itself.
In fact, when the SUJ2 was tested in Bathias’s labora-
tory in tension-compression at 20 kHz, no step was observed
in the S-N curve (Figure 4.28). Thus, it may be assumed that
the appearance of this step results from the use of the
nominal maximal stress in the S-N plot for the subsurface
initiation data.
Since the maximum stress in rotating-bending is located
at the surface of the specimen, a correction of the stress
should be made to the data for the specimens with internal
crack initiation accounting for its radial distance to the
surface (Figure 4.29).
Figure 4.27 shows the curve originally obtained by
Sakai (1999), where two straight lines may be drawn: one
in the conventional fatigue regime and one in the high cycle
fatigue regime, separated by a plateau. A correction to the
stress is made for the distance from the inclusion to
the surface, which yields a continually decreasing curve.
The magnitudes of corrections are observed in the comparison

Figure 4.28 SUJ2 tested in tension-compression at 20 kHz

(R ¼ 1).
S-N Curve and Fatigue Strength 127

Figure 4.29 Relationships between dinc and inclusion stress

correction Ds.

shown in Figure 4.30, where we see corrections up to

110 MPa.
Finally, the corrected values may be considered to form
part of the S-N curve in tension-compression for the same
steel.
The aim of this correction is to demonstrate that, when
correctly analyzed and plotted, the fatigue data would yield
essentially the same S-N curves regardless of the difference
in the loading method (tension-compression or rotating-
bending, for example).

4.8. Ti-Al INTERMETALLIC ALLOYS

The last example given is the gigacycle bending fatigue of

Ti-Al intermetallic. A piezo-electric fatigue machine working
in three point bending is useful when the metals are not
available in large amounts, are expensive, or are difficult
to machine.
128 Chapter 4

Figure 4.30 S-N curve for JIS SUJ2 steel tested at rotating
bending fatigue system [plot of original (Sakai, 1999) results with
corrected results].

4.8.1. Material
The material tested in the investigation is cast Ti-Al alloy.
Since Ti-Al alloy has excellent castability, studies have now
been undertaken to replace the existing machined or forged
parts with cast products of the Ti-Al alloy. However, the most
serious problem in using cast Ti-Al alloy parts is its poor relia-
bility and deterioration in mechanical properties. In order to
reduce the effect of pores, Hot Isostatic Pressing (HIP) tech-
nology has been applied to the material. In the study, the
Ti-Al alloy have been treated by HIP at 1200 C, 100 MPa
for 4 h, followed by homogenization treatment in vacuum at
S-N Curve and Fatigue Strength 129

1000 C for 20–24 h. The composite of Ti-Al alloy is shown in

Table 4.28.
The specimens were machined out from a 100 mm dia-
meter bar by electro discharge machining (EDM) followed
by grinding. The specimens were mechanically polished to
remove surface scratches before testing. The dimensions
of the rectangular specimens were 36  8  4 mm, and the
surface roughness was below 0.4 mm.

4.8.2. Mechanical Properties

To determine the mechanical properties of the Ti-Al alloy, sta-
tic tension tests and static bending tests were conducted;
Tables 4.29 and 4.30 show the results of these tests.

4.8.3. Fatigue Test

High-cycle three-point bending fatigue tests were performed
on a resonance fatigue testing machine at a frequency of

Table 4.28 Chemical Composition (wt%) of

the Material
Al Nb W Mo B

At(%) 45 1.5 0.05 0.1 0.05

Table 4.29 Static Tensile Test Results

Stress Modulus Load
(max) (MPa) (MPa) E (kN)

500.544 160668.578 5.835

Table 4.30 Static Bending Test Results

Load (kN) Stress (MPa)

3702 851.61
130 Chapter 4

20 kHz and stress ratios of 0.1 and 0.5. All the tests were car-
ried out at an ambient temperature, so as to determine the
fatigue strength at the life of 1010 cycles. The static preload
of fatigue tests was controlled on an INSTRON machine.
The dynamic stress applied to the specimen is controlled
by the maximum vibration displacement in the center of the
specimen; that is, the vibration displacement at the end of
the horn must agree with the value of input from the compu-
ter. Before testing, the specimen must be calibrated, to ensure
that the error between the actual displacement amplitude
and that given by the control unit is under 1%.
The microstructure of the Ti-Al alloy specimens was
observed and the microstructural characterization performed
on materials from fractured test bars with samples prepared
through standard grinding, polishing, and etching methods.
Microstructure of the samples was analyzed in order to
determine whether the fatigue test had altered the structure
in any way. The fracture surface of fatigue specimens was
examined by SEM to determine the microscopic fracture
mode.

4.8.4. High-Cycle Fatigue Properties

The fatigue S-N properties are shown in Figure 4.31 (Xue,
2004), which plots the elastic nominal peak stress at the outer
fiber of the beam versus the number of cycles to failure. Speci-
mens which did not fail are marked with open circles. When
R ¼ 0.1, higher dynamic loads were applied, so it has lower
peak stress compared to R ¼ 0.5. From the S-N plot we
observe that fracture can occur between 107 and 1010 cycles.
The asymptote of the S-N curve is inclined slightly, but it is
not horizontal. The fatigue strength of the Ti-Al alloy was cal-
culated at about 400 MPa for R ¼ 0.1 and 580 MPa for R ¼ 0.5.
The S-N curve for R ¼ 0.5 has a higher slope.
From the S-N curve we see that fatigue fracture occurred
between 107 and 108 cycles for R ¼ 0.1, and many specimens
cracked beyond 108 cycles for R ¼ 0.5. The asymptote of S-N
curve is inclined gently and no fatigue life limit is observed
for materials with R ¼ 0.5. Under the same nominal peak
S-N Curve and Fatigue Strength 131

Figure 4.31 Three-point bending high cycle fatigue test results

for Ti-Al intermetallic alloy.

stress, the scatter in fatigue lifetimes remain large for two dif-
ferent test conditions (R ¼ 0.1 and R ¼ 0.5). The large scatter
may be related to the fatigue crack mode for the Ti-Al alloy
with the nearly lamellar structure.
From the S-N curves obtained with R ¼ 0.1 and R ¼ 0.5,
interesting observations can be made. Most specimens failed
beyond 107 cycles when stress ratio is R ¼ 0.5; in fact, many
of them failed beyond 108 cycles. In contrast, few of the speci-
mens survived 108 cycles with R ¼ 0.1.
The fatigue life is mainly decided by the crack initiation.
Failure is initiated at weak spots such as large lamellar colo-
nies oriented perpendicular to the loading direction, and large
g-grains near the surface of the tested specimen. There is no
apparent stable crack growth for the near lamellar material.
Most fatigue crack initiations are from the interlamellas
or the boundaries between lamellar colonies on the surface of
the specimen. However, for specimens with very high fatigue
life (over 107 cycles), crack initiation is often from the subsur-
face of the specimen and, even if crack initiation takes place
at the surface, it may remain as a very small surface fatigue
crack. More than one crack initiation can be found in many
specimens.
132 Chapter 4

Figure 4.32 Fatigue fracture surface (smax ¼ 630 MPa; R ¼ 0.5;

Nf ¼ 1.8  109): (a) surface fatigue initiation; (b) intergranular fati-
gue initiation on surface.

The heterogeneity and anisotropy of the lamellar colo-

nies may account for the large scatter of the high fatigue test
results (Figure 4.32).
5
Crack Growth and Threshold

Using a piezo-electric fatigue machine, it is possible to deter-

mine the fatigue crack growth curve, DK versus da=dN,
including the threshold regime to growth as low as 10
12
m=cycle. It will be noted that the fatigue thresholds are basi-
cally the same in both conventional and resonant fatigue if
the computation of the stress intensity factor K is correct.
One objective of using the piezo-electric fatigue machine to
determine DKth is to save time and money, but it is also of
interest to note that a piezo-electric fatigue machine is very
effective both for a wide range of cyclic load ratios—R ¼
1
in tension-compression and R > 0.8 with a high mean stress.
Prior to developing the data by piezo-electric fatigue
machine methods, it is appropriate to review the history of
the analysis of crack growth and its threshold. Mechanical
fatigue testing machines (McEvily, 1958) developed the first
very wide range of crack growth rates in aluminium alloy
7075T-6 (and 2024T-3) shown in Figure 5.1. This plot is the
first (Paris, 1962) to employ the range of the crack tip stress
intensity factor, DK, vs. the crack growth rate, da=dN, as
the primary parameters on a double logarithmic basis. At that
133
134 Chapter 5

Figure 5.1 Fatigue crack growth data (McEvily, 1958) for

7075 T6 (R ¼ 0)

time a straight line on that plot was noted to illustrate the

broad trend as:
da
¼ CðDKÞm
dN
with the constant, C, dependent on the load ratio, R. It was
understood at the time that, for cyclic loading, a cyclic plastic
zone occurs caused by DK at the crack tip. Within a static
maximum load, Kmax, a plastic zone about 4 times larger,
depending upon R and cyclic hardening or softening, results
(Figure 5.2) (Paris, 1963). It is acknowledged that this
static=cyclic plastic zone concept originated with McClintock.
McEvily’s data (Figure 5.1) were noted to be close to one
Burger’s vector, b, per cycle at the low rate end, so it was nat-
ural to wish for even slower data to see how far that straight
line extended. Therefore (Lindner, 1965), using an Amsler
Vibrofore machine which could produce 150 Hz load rates,
developed the data given on Figure 5.3 with several
months of continuous testing. The flattening of the curve in
Crack Growth and Threshold 135

Figure 5.2 Succession of zones of plasticity near a crack tip.

Figure 5.3 was initially thought to be a threshold for fatigue

crack growth for rates below one Burger’s vector per cycle.
However, it was felt that more observations should be made
prior to accepting that the threshold indeed did exist.
(Paris, 1970), via tests of 9310 gear steel in a servo-
hydraulic test machine modified to produce 200 Hz by
R. Churchill and H. R. Hartmann, verified the threshold con-
cept and showed the strong effect of load ratio, R (Figure 5.4).
At about this time, (Elber, 1970, 1971) demonstrated that
crack closure occurs during cyclic loading. Shortly thereafter,
(Schmidt, 1973) produced the first extensive data for load
ratio effects on threshold using an electro-dynamic shaker
to produce up to 700 Hz cyclic loads. Some of his data are
plotted in Figures 5.5 and 5.6. He noted that the load ratio
effects can be explained with Elber’s crack closure concept.
The data demonstrate that at high load ratio, R, no crack
136 Chapter 5

Figure 5.3 First fatigue crack threshold data on 7075-T6

(Lindner, 1965) (R ¼ 0).

closure occurs, hence in that range the DKth for threshold is

constant. However, at low load ratio, R, crack closure occurs
at a K-level, which is quite independent of the minimum load
and results in a range in which the Kmax for the threshold is
basically constant. This is because it was assumed that the
effective stress intensity range, DKeff, extended from the
crack opening load to the maximum load. This general form
of behavior of threshold is shown in Figure 5.7.
More recently (Donald, 1997 and later) has provided
extensive data for a wide variety of load ratios near and above
threshold and analyzed these data using various concepts
such as closure to explore load ratio effects. Typical results
are presented in Figures 5.8 through 5.10 to illustrate these
concepts as simply as possible.
All three figures plot the same data on 7055 aluminum
alloy for load ratios, R ¼
1, 0.1, 0.3, 0.5, and 0.7, where in
Figure 5.8, the actual applied loads are used to compute DK
with no accounting for crack closure. Figure 5.9, the crack
Crack Growth and Threshold 137

Figure 5.4 Verification of threshold behavior in 9310 gear steel

for R ¼ 0 and 0.9 (Paris, 1970).

opening load is determined from the load displacement

records becoming linear (by the ASTM method) and is desig-
nated as:
DKeff ¼ DKop ¼ Kmax
Kopen
in Figure 5.10 the partial closure model (Paris, 1999) is used
where, for simplicity, the form adopted is:
2
DKeff ¼ Kmax
Kopen
p
Incidentally, when the minimum load is above the crack
opening load the applied load range should be used since
there is no closure. For R ¼ 0.7 and higher this is usually
the case.
138 Chapter 5

Figure 5.5 Threshold data on T-1 steel with load ratio effects
(Schmidt, 1973).

Figure 5.6 Threshold data for A533 steel (Schmidt, 1973).

Crack Growth and Threshold 139

Figure 5.7 Form of closure effects on threshold (Schmidt, 1973).

The partial closure model assumes that closure does not

occur all the way to the crack tip and it is noted to improve the
data correlation in the near-threshold regime, as well as above
threshold. The adjusted compliance ratio of Donald (1997)
denoted ACR is equally able to improve the data correlation here,
as well as other methods he has developed. This ability to better
correlate the data shows that more than simply first crack
closure is of influence, especially in the near-threshold regime.
Further, for the high load ratio regime where closure
does not occur, Donald (1997, 1999) has demonstrated a sen-
sitivity of the data to Kmax, as is shown in Figures 5.11 and
5.12 by improving the data correlation further using the
correlating parameter: DK 0:875  Kmax 0:125
. These correlations
140 Chapter 5

Figure 5.8 Near-threshold data of Donald (1997) using applied

stress intensity for 7055 aluminum (Paris, 1999).

Figure 5.9 The same 7055 data using the opening load method
(Paris, 1999).
Crack Growth and Threshold 141

Figure 5.10 The same 7055 data using the partial closure 2=Pi0
method (Paris, 1999).

Figure 5.11 High maximum stress intensity data comparison on

2024T-3 (with no closure) (Donald, 1997,1999).
142 Chapter 5

Figure 5.12 The same 2024T-3 data given in Figure 5.11 adjusted
by including the maximum stress intensity in the load parameter
(Donald, 1997, 1999).

are presented to show that there is some understanding of

load ratio effects and maximum load effects.
However, for practical circumstances, measurements or
predictions of opening loads etc. are not possible. Therefore
data for practical use simply are developed in terms of the
applied stress intensity factor range disregarding closure for
the load ratios involved. The piezo-electric ultra-sonic data
to be presented later will be given on that basis, using simply
the DK applied.
Before going on to the ultra-sonic data, another strong data
trend of a different nature is worthy of discussion here. In the
early 1960s, W. E. Anderson (Donaldson, 1961) observed that
the basic data on fatigue crack growth of many alloy base mate-
rials and alloys of different strength levels of each base material
can be normalized by dividing DK by the elastic modulus, E.
This was observed after analyzing the data in Figure 5.13.
In addition (Hertzberg, 1997) had noted in the 1960s that
Lindner’s first data on threshold flattened at a growth rate of
one Burger’s vector, b, per cycle. He also noted that the
Crack Growth and Threshold 143

Figure 5.13 The comparison of data for a variety of base metals

(Donaldson, 1961).

normalization with modulus is dimensional

pffiffiffi and tried the
non-dimensional parameter, DK=E b with good success.
Further Paris (1999) noted that this parameter was improved
further by using the effective stress intensity from the prece-
ding discussion.
Consistent with Hertzberg’s analysis plus our modifica-
tion, the threshold corner is predicted to be at
da DKeff
¼b and pffiffiffi ¼ 1
dN E b
and this can be used as the starting point to predict above
threshold growth rates for all metal alloys that all follow
the same relationship
 
da Keff 3
¼ b pffiffiffi
dN E b
This threshold corner and line are plotted in Figures 5.14 and
5.15 to illustrate its effectiveness as a predictor of the
144 Chapter 5

Figure 5.14 The predicted threshold corner and slope using the
partial closure model for 2024T-3 (Paris, 1999); see also (Hertzberg,
1996, 1997).

Figure 5.15 The partial closure predicted threshold corner for

6061 (data provided by Donald, 1997).
Crack Growth and Threshold 145

data. Hertzberg (1996,1997) presents considerable supporting

data for this prediction formula on many different metal
alloys.
Small crack behavior, which differs from long cracks by
being absent of crack closure, is known to be bounded in growth
rates by high load ratio long crack behavior. The high load ratio
long crack behavior is also closure free and consequently is also
at a higher maximum K, which accounts for the bounding effect.
Therefore, in the ultrasonic data that follows, the high load ratio
data with other considerations will suffice to cover this area of
small crack behavior in a minimal way.
Let us indicate that the computation of load ratio R is dif-
ferent in ultrasonic fatigue and in conventional fatigue. In
conventional fatigue tests, force can be measured. The nom-
inal force and nominal stress are constant at each cross
section of the specimen and the stress intensity factor is
proportional to the nominal stress. So, the ratio R will be
easily determined from

Fmin smin Kmin

R¼ ¼ ¼
Fmax smax Kmax

where F is the applied force, s the nominal stress, and K the

stress intensity factor.
In ultrasonic fatigue, however, the total stress is the
superposition of static stress and dynamic stress. Because
there exists an inertia force in vibration, the nominal stress
will not be constant in the cross sections of the specimen.
Since the nominal force is impossible to measure, the nominal
stress cannot be determined during tests. In this case, R
cannot be defined by using F and s and the following formula
is perhaps the only choice

Kmin
R¼ and it is noted : DK ¼ Kmax
Kmin
Kmax

Therefore in ultrasonic fatigue, the calculation of K is

necessary not only to determine DK, the crack growth rate
curve, but also to obtain R.
146 Chapter 5

5.1. TITANIUM ALLOYS

The material to be tested is T6A4V, whose chemical composi-

tion, heat treatment, and mechanical properties are listed in
Tables 4.20 to 4.22. (The calculation of stress intensity factor
K has been discussed elsewhere.)

5.1.1. Fatigue Crack Growth Rate

The experiments are performed for FCG (fatigue crack
growth) rates between 10
7 m=cycle and 10
11 m=cycle and
with stress ratios R ¼
1 to 0.9. The fatigue threshold DKth
is determined for a rate near 10
11 m=cycle by increasing pro-
gressively the vibration amplitude from 5% to 10% and the
static load about 10%.
The experimental results (Bathias, 1997) are given with
curves da=dN versus stress intensity amplitude DK ¼ Kmax

Kmin. Recall that Figure 5.1 showed that the crack propaga-
tion rate da=dN increases with R, when the latter is positive.

Figure 5.16 Stress ratio influence on T6A4V fatigue crack growth.

Crack Growth and Threshold 147

In conventional fatigue, when R becomes relatively high, the

crack closure effect is suppressed because the cycle minimum
load is above the opening load according to (Elber, 1971). This
model has already explained the influence of ratio R on the
effect of the loading cycle and on the FCG rate. In ultrasonic
fatigue with mean load equal to zero (R ¼
1), the closure
effect is very small when we consider only the tensile part
of the cycle (Kmin ¼ 0). Consequently, the results obtained
for R ¼
1, 0.7, and 0.9 are close to one another (Schmidt,
1973).
However, the tests show that between R ¼ 0.7 and
R ¼ 0.9, there is a very small DKth variation. Therefore, the
closure effect seems to remain even for the highest values of
ratio R but to the contrary this is Kmax effect. pffiffiffiffiffi
For R ¼ 0, the threshold equals 5.66 MPa m, a value
significantly higher than the others. This result is comparable
to that obtained at conventional fatigue machine speed with
at frequency of 2 Hz, and with a stress ratio R equal to 0.03.

5.1.2. Fractography
Fractographic observations have been made for specimens
tested in ultrasonic fatigue at 20 kHz, at room temperature
and with R ¼ 0,
1, 0.8, and 0.9. The observation can be
summarized as follows:
 For R ¼ 0.8 and 0.9, the fracture surfaces are the
same: Near the threshold, the fracture mode is very
crystallographic with a strong influence of the micro-
structure. When FCG rate da=dN becomes greater
than 10
8 m=cycle (intermediate regime), the fracture
surface becomes smoother and less crystallographic.
 For R ¼
1, as for R ¼ 0.8 and 0.9, the propagation
mode is very crystallographic at the threshold because
of the low value of DK. When da=dN becomes consider-
ably higher, some fatigue striations have been identi-
pffiffiffiffi
ffi
fied with some difficulty, for DK¼ 8MPa m and
da=dN ¼ 1.510
8 m=cycle.
 For R ¼ 0, the main observation that characterizes the
fracture surface is the presence of fatigue striations
148 Chapter 5

Figure 5.17 Striations (T6A4V, 20 kHz).

pffiffiffiffiffi
for DK ¼ 13.62 MPa m and da=dN ¼ 210
8 m=cycle
(Figure 5.17). These striations are clear and distribu-
ted in a regular way, and sometimes separated by sec-
ondary cracks.

The results of ultrasonic fatigue show that for high load

ratio R (0.8 and 0.9), striations are absent whereas for R ¼ 0
with a significant crack closure effect, the striation mecha-
nism is the same as in conventional fatigue. However, at
da=dN rates higher than 10
8 m=cycle, in both cases, stria-
tions are absent.
To summarize, the ultrasonic fatigue data of T6A4V
show that:
 DKth decreases with the increase of R from 0 to 0.9.
Thus, the effect of ratio R exists in ultrasonic fatigue
as in conventional fatigue, and the closure effect
remains despite the possibility of a reduction of
residual stress by vibration.
 At room temperature and when the environment is
not aggressive, the crack propagation mechanisms
are apparently the same in ultrasonic fatigue and in
conventional fatigue (Bathias, 1997).
Crack Growth and Threshold 149

5.1.3. Cumulation of Ultrasonic Fatigue

with Slow Fatigue
This part studies the damage cumulation during crack propa-
gation of specimen T6A4V when a 20 kHz vibration combines
with a low fatigue cycle. Thus, the aim is to determine the
conditions under which the vibration becomes significantly
damaging with regard to the low cycle.
During the test, a crack
pffiffiffiffiffi is first propagated at low cyclic
rates with Kmax ¼ 20 MPa m and R ¼ 0. Then, the specimen
is subjected to a superimposed 20 kHz loading. When the
static load reaches its maximum value, a small amplitude
vibration is superposed at a frequency of 20 kHz. This
sequence was repeated threeptimes, ffiffiffiffiffi while the same static
SIF Kmax was kept at 20 MPa m (Figure 5.18).
In particular after applying 2122ppropagation
ffiffiffiffiffi cycles in
slow cyclic fatigue with Kmax ¼ 20 MPa m, a 20 kHz vibration
6
pffiffiffiffiffi additional blocks (5.4  10 small
is superposed during three
cycles) at DK ¼ 2.6 MPa m and with R ¼ 0.89. Analysis shows
that this vibration has no significant effect on the behavior of
low fatigue crack propagation. With the vibration pffiffiffiffiffi amplitude
having slightly been increased (DK ¼ 2.8 MPa m), the crack
propagated by 0.2 mm (after three cumulation blocks). There-
fore, there is a threshold for which a superposed vibration is
damaging when it is applied to a slow cyclic fatigue. Note that
the crack propagation rate per block (6  10
5 m=cycle) under
combined charge is much more important than that in slow

Figure 5.18 Cumulation fatigue test with a superposed SIF

value.
150 Chapter 5

Figure 5.19 Fracture aspects of cumulative fatigue: (a) Three

sequences corresponding to three cumulation blocks; (b) transition
zone from combined fatigue to slow fatigue; (c) transition zone from
slow fatigue to combined fatigue.

fatigue without superimposed vibration (2  10
7 m=cycle)

although the vibration amplitude is very small (Bathias,
1997).
When the experiment was completed, we performed a
metallographic and microfractographic analysis in the scan-
ning electron microscope. Figure 5.19 allows us to identify
the three blocks of cumulation corresponding to the three
superposition sequences. It shows that between two consecu-
Crack Growth and Threshold 151

tive sequences of ultrasonic fatigue (two blocks) a second

crack appears, playing the role of fatigue striation, which
allows one to identify the low fatigue cycle.
Passing from slow fatigue to ultrasonic fatigue, we notice
a transition zone of a few microns in length depending on the
plastic zone radius at the crack tip as shown in Figure 5.19(c).
This zone does not exist in the passage from ultrasonic fatigue
to low fatigue Figure 5.19(b). Therefore, we think that the
passage from ultrasonic fatigue to slow fatigue is different
than the passage from slow fatigue to ultrasonic fatigue.
The plastic zone radius is proportional to (DK=sy)2 where
sy is the yield strength
pffiffiffiffiffi of the alloy. In slow fatigue,
DK ¼ Kmax ¼ 20 MPa
pffiffiffiffiffi m whereas in ultrasonic fatigue
DK ¼ 2.8 MPa m. Hence, the plastic zone radius at the crack
tip is more significant in the passage from slow fatigue to
ultrasonic fatigue than vice versa. That explains why the
combined fatigue threshold is slightly higher than the
one obtained in ultrasonic fatigue alone with R ¼ 0.9
(Figure 5.16).
The cumulation tests show that a light vibration at high
frequency superimposed on slow fatigue may accelerate the
propagation of a crack. This could be very dangerous for
mechanical structures such as turbine blades.

5.2. NICKEL-BASED ALLOYS

A series of nickel based alloys have been tested at 20 kHz in

the CNAM=ITMA Laboratory for many years. This section
will deal with three of them: Astroloy (Wu, 1992), N18 (Bonis,
1997), and Inconel 706 (Bonis, 1997).

5.2.1. Astroloy
The chemical composition and mechanical properties of
Astroloy are listed in Tables 5.1 through 5.3.
Figure 5.20 gives the two FCG rate curves, the threshold
corresponds to a growth rate of 10
9 mm=cycle at R ¼
1
where DK is taken to be Kmax. Again, it is necessary to emp-
hasize that in ultrasonic fatigue, a special method has to be
152 Chapter 5

Table 5.1 Composition of Astroloy (%wt)

C O S Mn Si Cr Zr Mo Co Ti Al Fe Ni

0.022 0.013 0.002 <0.02 <0.1 14.9 0.05 5.0 17.0 3.51 4.02 <0.1 Remainder

Table 5.2 Heat Treatment of Astroloy

Isothermal forged thread (70 C=min), oven vacuum, 1100 C – 4 h;
quenching in air; 650 C – 24 h – air, 760 C – 8 h – air

Table 5.3 Mechanical Properties of Astroloy

Temperature ( C) E (GPa) r (kg=m3) n

20 214 8000 0.30

400 200 8000 0.30

n ¼ Poisson’s ratio.

used to obtain the FCG rate curves. The appearance of an

initial crack from the notch requires a high load. In this case,
a large plastic zone exists at crack tip and it is necessary to
decrease the test load, i.e., the vibration amplitude. The
starter crack length of between 2 and 3 mm is generated
and then the FCG rate curves is obtain by increasing DK
tests. These curves at R ¼
1 are analogous to conventional
crack propagation curves at R  0.7, that is, without a crack
closure effect.

5.2.2. N18
The chemical composition, mechanical properties, and heat
treatment of alloy N18 were given in Tables 4.25 through
4.27. FCG tests of superalloy N18 are performed at high
temperatures. At 450 C, three values of R are used: R ¼
1,
0, and 0.8. At 700 C, there is one value of R: R ¼ 0.
Starter cracks have a minimum length of 1.5 mm after
pre-cracking at the notch of the specimen. The specimen’s
Crack Growth and Threshold 153

Figure 5.20 Crack growth rate of Astroloy at 20 kHz.

surface is finely polished to make observations of FCG easy at

high temperature. Tests at R ¼
1 do not need the addition of
static traction; this permits ultrasonic excitation to continue
during crack propagation. For other values of R > 0, it is
occasionally necessary to stop excitation in order to keep the
R ratio constant. In this case of R > 0, the stopping
time, even if minimized, brings some oxidation at the crack
tip. This phenomenon is also observed when a static traction
keeps the crack open continuously. This is, of course, one of
the difficulties for tests at R > 0 because of the persistent
oxidation at the crack tip during the test process.
Figure 5.21 groups experimental results of FCG of N18 at

450 C for R ¼
1, 0, and 0.8. The thresholds
pffiffiffiffiffi for these
pffiffiffiffiffi three
values of R
pffiffiffiffiffi are respectively 5.5 MPa m , 8 MPa m, and
4.5 MPa m. For R ¼
1 and 0.8, the effect of crack closure is
notably weak, which explains why the thresholds are similar.
Figure 5.22 illustrates the results of tests done in air at
400 , 450 , and 700 C and in vacuum at a temperature of
650 C. These results are also compared with the tests per-
formed at low frequency of 0.5 Hz and with an R ratio
approaching zero.
We notice that the curve at 650 C in ambient air is above
that of 700 C tested at 20 kHz. This difference can be
154 Chapter 5

Figure 5.21 Crack growth rate of N18 at 450 C and different R

values.

Figure 5.22 Mixed comparison of FCG rate curves of N18 up to

700 C.
Crack Growth and Threshold 155

explained by the fact that very low testing frequency facili-

tates oxidation in the crack between cycles. This conclusion
about oxidation is also confirmed by the good agreement of
the curves obtained at low frequency in a vacuum with the
results at ultrasonic fatigue. Another difference is observed
between the curves of conventional fatigue at 400 C and that
of ultrasonic fatigue at 450 C. At very high frequency, the air
has much less time to arrive at the crack tip, which limits the
oxidation.
On the other hand, the increase of temperature between
450 C and 700 C diminishes the threshold for cracking. We
believe the reason may be the thermal activation of oxidation
at the crack tip. This demonstrates that a test at 20 kHz is not
always similar to the one at low frequency.

5.2.3. Inconel 706

Inconel 706 is used to manufacture jet engine disks and tur-
bines blades. It is a superalloy mainly composed of nickel
and iron, with a precipitation hardening structure and large
grains of about 125 mm, and it is easy to weld. The precipitates
are g-Ni3 (Ti, Al) and b-Ni3Nb.
Tests were carried out at 20 kHz with a load ratio R ¼ 0.1
and at two temperatures of 20 C and 400 C. The chemical
composition, heat treatment, and mechanical properties are
listed in Tables 5.4 through 5.6.
With alloy N18, we also find an increase in crack
growth rate induced by oxidation at the crack tip when the

Table 5.4 Chemical Composition of Inconel 706 (wt%)

C Mn Si Cr Al Nb Ti Ni Fe

0.03 0.18 0.18 16 0.2 2.9 1.75 41.5 37.44

Table 5.5 Heat Treatment of Inconel 706

982 C 1 h, cooling in air; 843 C 3 h, cooling in air;
718 C 8 h, cooling to 621 C (38 C=min) and stable 8 h, cooling in air
156 Chapter 5

Table 5.6 Mechanical Properties of Inconel 706

s0.2 (MPa) E (GPa) r (kg=m3) su (MPa)

931 210 8050 1138

temperature is increased. However with Inconel 706 for a

lower FCG rate, an oxide film props the crack faces open
and reduces the cyclic effects. This side effect is particularly
notable with Inconel 706 and prevents the exploration of the
very slow speeds of propagation (Figure 5.23).
The results of ultrasonic fatigue tests are consistent with
those obtained at 20 Hz. We observe in Figure 5.24 that the
slopes of the upper FCG regime are parallel. At 20 kHz, the
environment has less time to interact with the metal at the
crack tip than at 20 Hz, which may explain the fact that the
curve of 20 kHz is slightly below that of 20 Hz.
The following test procedures were designed to better
distinguish the propagation acceleration effect and side effect
mentioned earlier that are both due to oxidation. A specimen

Figure 5.23 FCG rate of Inconel 706 at 20 C and 400 C.

Crack Growth and Threshold 157

Figure 5.24 FCG rate of Inconel 706 at 20 C: Comparison of

20 kHz and 20 Hz.

is tested at 20 C keeping the crack in propagation, and pro-

gressively decreasing the DK value until the complete arrest
of propagation. Always at ambient temperature, we impose
a DK value less than the threshold of propagation at 400 C.
Once the excitation is well established, the temperature rises
rapidly. Then the progression of the crack takes a much
higher speed than the one that can be extrapolated from the
curve of da=dN at 400 C. After a stop of several minutes at
400 C, the propagation does not recover the same value of
DK. However, by increasing the excitation, we return to the
curve of da=dN at 400 C.
The following experiment demonstrates the significance
of the side effect due to oxidation and explains its blockage
effect on crack propagation as the threshold is approached.
Because the crack is swept by a gaseous nitrogen jet, the
cracking speed is slightly lower and blockage to the propaga-
tion when approaching the threshold is delayed. This can be
seen in Figure 5.25.
Curves of FCG in ambient air with and without nitrogen
sweep do not differ greatly although the paths of cracking and
the crack surfaces are noticeably different.
158 Chapter 5

Figure 5.25 FCG rate of Inconel 706 at 400 C with and without
nitrogen sweep.

A comparison of da=dN curves of N18 at 450 C with that

of Inconel 706 at 400 C in Figure 5.26 indicates that N18 has
a slightly better behavior in cracking. Since N18 is a little less
sensitive to the phenomenon of blockage due to the side effect
of oxidation, the threshold is higher.

Figure 5.26 Comparison of FCG rate of N18 and Inconel 706.

Crack Growth and Threshold 159

5.3. ALUMINIUM ALLOYS

Stanzl’s group and co-researchers in Vienna, Austria have

studied a series of aluminium alloys. The testing conditions
cover a variety of loading: rapid load reduction, low amplitude
two-step loading, in-service loading, and so on. Results
regarding the properties of crack growth and threshold
mainly in gigacyclic regime can be found in (Mayer, 1991,
1992) and (Stanzl, 1991, 1993).

5.3.1. Al2024 Alloy After Rapid Load Reduction

(Mayer, 1991)
One effect of FCG with variable cyclic stress intensities
is crack growth retardation after overloads. This causes
the crack growth rate to be generally lower than the sum
of the single crack propagation rates expected from the
crack propagation curve obtained with constant load ampli-
tude. The retarded crack growth is due to several crack clo-
sure effects. For multi-stage loading of a homogeneous
metallic material, plastic deformation and crack surface
roughness, as well as oxidation, are regarded as the main
causes.
Retardation effects are important in designing struc-
tures because, in general, service load amplitudes are not
constant. Knowledge of the influence of previous load history
on crack propagation rate is therefore important. A compre-
hensive study of crack closure effect after load reduction in
the threshold regime has not yet been presented. This
regime is important to study since crack propagation beha-
vior and the crack closure effects cannot be extrapolated to
low Kmax values from measurements at higher stress inten-
sity values. The objective of this study was to begin to ana-
lyze quantitatively crack retardation effects in the threshold
regime.
The material tested was aluminium alloy Al2024 whose
chemical composition and mechanical properties are listed
in Tables 5.7 and 5.8.
160 Chapter 5

Table 5.7 Chemical Composition of Al2024 (wt%)

Cu Mg Si Fe Mn Cr Zn Ti Al

4.5 1.5 0.1 0.2 0.7 < 0.05 < 0.05 < 0.05 Remainder

Table 5.8 Mechanical Properties of Al2024

pffiffiffiffiffi
E (GPa) sm (MPa) s0.2 (MPa) A% KIC (MPa m)

72.5 460 352 18 35

The material was delivered as rolled sheets of 20 mm

thickness. The heat treatment procedures consisted of solu-
tion heat treatment at 495
C for 2 h, quenching, cold work
and age hardening at room temperature for more than 4 days.
For fatigue crack propagation measurements, rectangular
specimens with a circular tapered cross section in the center
and a single edge notch were used. As stated before, this type
of specimen has been commonly used for ultrasonic fatigue
testing and allows the strain and stress to form a standing
wave with the maximum of strain in the middle of the reduced
section.
Loading was performed with a 20 kHz ultrasonic fatigue
machine at R ¼ 1. The tests were carried out in two different
environments: ambient air and vacuum. The curves of da=dN
versus Kmax without load reductions are shown in Figure 5.27
(Stanzl, 1996; Mayer, 1999).
Before rapid reduction of the stress intensity factor
value, a fatigue crack is initially grown
pffiffiffiffiffiapproximately 1 mm
at the high cyclic SIF of K1 ¼ 7 MPa m and takes approxi-
mately 5  105 cycles. This fatigue crack extension at the high
stress intensity values K1 leads to well-defined starting condi-
tions for the rapid load reduction test. The crack surface
roughness is typical for K1; both the plastic zone in front of
the crack tip according to K1, and the oxide layer of character-
istic thickness on the crack surface are the same at beginning
of each load reduction test.
Crack Growth and Threshold 161

Figure 5.27 FCG curves of 2024-T3 in a vacuum and in humid air

(Stanzl, 1996).

After rapid load reduction, cycling with a defined lower

stress intensity value K2 follows until the crack has grown
0.10  0.15 mm. If no crack growth could be detected after
1109  1.5109 cycles after rapid load reduction, the experi-
ment is stopped and it is assumed that the threshold for this
type of loading is obtained. This stress intensity value is
called reduction threshold. Figure 5.28 (Mayer, 1992)
162 Chapter 5

Figure 5.28 Test sequence for delay time cycle measurements

after growing the p
crack
ffiffiffiffiffi by approximately 1 mm with a cyclic stress
intensity of 7MPa m ( approximately 5  105 cycles).

presents the test sequence; Figure 5.29 (Mayer, 1992) gives

crack propagation
pffiffiffiffiffi after constant amplitude cycling at
Kmax ¼ 7 MPa m and rapid load reduction.
Loading is performed at the reduced stress intensity
value until crack growth of 0.1–0.15 mm is detected. If no
further crack extension is observed, loading at the low stress
intensity is continued for at least 1  109 cycles (Stanz, 1996).

Figure 5.29 Delayed crack growth after rapid reduction of the

stress intensity value (Mayer, 1991).
Crack Growth and Threshold 163

Figure 5.30 Dependence of delay cycles after rapid reduction of

stress intensity to the lower stress intensity level (Mayer, 1991).

As a quantitative measure for retarded crack propaga-

tion after reduction of the Kmax value, the so-called delay-
cycles are determined. These are defined as the number of
cycles necessary for a crack to grow approximately 0.10 
0.15 mm at the lower SIF value after rapid load reduction,
minus the number of cycles needed for the same crack incre-
ment at low stress intensity amplitude without previous load-
ing at high stress intensity level.
Figure 5.30 (Mayer, 1991) shows the delay cycles for
rapid load reduction pffiffiffiffiof
ffi the cyclic stress intensity amplitude
from Kmax ¼ 7 MPa m to the value indicated at the ordinate.
The square dots show the measurements in ambient air, and
the circles characterize the measurements in vacuum. The
phenomena observed are as follows. The crack propagation
rate is rapidly slowed after rapid reduction of SIF. The retar-
dation is characterized by delay time cycles as defined above.
If Kmax is rapidly reduced, load reduction threshold SIF
amplitude is required to continue crack growth.pffiffiffiffiffi This reduc-
tion threshold is
pffiffiffiffiffi approximately 4.3 MPa m in ambient air
and 3.5 MPa m in vacuum. In both environments, this SIF
value is higher than the threshold determined in a usual
164 Chapter 5

crack propagation test with a SIF reduction of, at most, 5% 

7%. If the crack continues to grow after rapid reduction of the
SIF, the retardation number of cycles is lower in air than in a
vacuum. The reduction threshold, however, is higher than
that in a vacuum. The results show less scattering for the
vacuum than for air. In a vacuum, crackppropagation
ffiffiffiffiffi can be
observed at an SIF amplitude of 4 MPa m, although crack
arrest (delay) is observed during more than 5  107 cycles.
Detailed environmental influence on FCG and threshold
properties of Al2024 will be further discussed.

5.3.2. Al2024 Alloy Under Low Amplitude

Two-Step Loading
As indicated earlier, the increased threshold SIF value after a
reduction of Kmax is referred to as the reduction threshold.
The significance of such a reduction threshold for crack propa-
gation under variable amplitude loading is not clear, because
other considerable interaction effects can apparently occur
from a load amplitude reduction. This discussion deals with
the same material studied previously, but under low ampli-
tude two-step loading, as shown in Figure 5.31 (Mayer, 1992).
Particular interest is placed on whether cycles at or
below the previous reduction threshold will also affect crack
growth in such two-step load sequences. For practical reasons
it is worthwhile to know if numerous small cycles of an ultra-
sonic fatigue load spectrum contribute to fatigue crack
growth.
Tests were performed at room temperature, at 20 kHz
and R ¼
1. All p tests
ffiffiffiffiffi begin with a constant amplitude load
at Kmax ¼ 7 MPa m with a pre-crack length increase of about
1 mm. The plastic zone size for this Kmax value is about 40 mm.
This procedure guarantees well-defined starting conditions
for subsequent two-step loading tests. In the two-step tests,
the load level alternates periodically between the starting
level of Kmax and some lower level. The two-step loading con-
tinues until a stationary crack growth velocity is attained.
Then constant amplitude loading is applied again at the high
Kmax value. The fatigue load is essentially K-controlled.
Crack Growth and Threshold 165

Figure 5.31
pffiffiffiffi
ffi Tabulation of load sequences. For all tests Kmax,1 is
7 MPa m and N1 is 500 cycles (Mayer, 1992).

Each two-step
pffiffiffiffiffi test period consists of 500 cycles at
Kmax,1 ¼ 7 MPa m and a 100 times or 1000 times larger num-
ber of cycles at the lower level.
Some of the crack growth results are illustrated
pffiffiffiffiffi on Figure
5.32 (Mayer, 1992) with Kmax,2 ¼ 3.5 MPa m, N2 ¼ 50,000
cycles. The two predicted reference levels for the average crack
growth rate are also shown. The crack growth exhibits the fol-
lowing trends: Some crack extension does occur during the
two-step period before a stabilized average crack growth rate
is reached. After the two-step loading period is completed,
again some crack acceleration occurs during the subsequent
constant higher amplitude loading before the original crack
growth rate stablizes. In all cases, according to Miner’s pre-
diction the average crack growth rate during the two-step
loading starts at Miner’s prediction with increased threshold.
However, the crack growth rate immediately increases,
166 Chapter 5

pffiffiffiffiffi
Figure 5.32 Crack propagation behavior, Kmax,2 ¼ 3.5 MPa m,
N2 ¼ 50000 cycles. (A) Miner’s law crack propagation rate without
increased threshold; (B) Miner’s crack propagation rate, but with
increased threshold value (Mayer, 1992).

which implies that the increased threshold level does not

fully apply. In other words, either crack extension does occur
at the lower Kmax level, or the higher Kmax level has become
more damaging than in the constant amplitude tests (or both
apply). On the other hand, crack growth rate does not reach
the Miner’s prediction level without increased threshold.
Apparently, some retardation effect remains. Quantitatively,
differences between the test results are evident. Crack
growth rate during the two-step loading increases by a factor
of 1.8, well below Miner’s non-interaction prediction of a fac-
tor of 15. Other results (Mayer, 1992) indicate that for the
higher Kmax,2 value, the crack growth rate increased by fac-
tors of 15 and 300. This result comes much closer to the
non-interaction Miner prediction. For N2 ¼ 50,000 it remains
2.5 times below that level, while for N2 ¼ 500,000
Crack Growth and Threshold 167

the crack growth rate becomes equal to the non-interaction

level.
In conclusion, ultrasonic fatigue results show significant
retardation effects. Low SIF level does contribute to crack
growth. Delayed retardation explains this observation. Low
amplitude cycles may induce a smoother fracture surface
and increase crack growth rate at higher amplitude. Accurate
predictions are therefore problematic. For similar reasons,
omitting numerous low amplitude cycles in random load
experiments should be done with great care.

5.3.3. AlSi Alloy Under In-service Loading

(Stanzl, 1993)
Fatigue properties of structural materials are normally
reported for loads with constant amplitude. Simulations of
in-service loading conditions are sometimes attempted by
the use of randomly distributed amplitudes. The resulting
alternative S-N and lifetime curves are then often used to
select a material for a specific purpose. The study presented
here concerns the fatigue properties of two AlSi cast alloys
used for automobile wheels. Specimens are tested in the
regime of 108  1010 cycles at 20 kHz and R ¼
1. Table 5.9
gives the chemical compositions of the two materials. These
materials are nominally identical. In material 1, the H2 gas
content of the melt is artificially enhanced before casting, so
that its microporosity is also influenced. Material 2 is cast
under the same conditions as in normal production. Both
alloys are used without additional heat treatment.

Table 5.9 Chemical Compositions of AlSi Alloys (wt%)

Si Fe Cu Mg Zn Ti Mn Al

10.0 7 11.8 <0.18 <0.03 0.001  0.4 <0.07 <0.15 <0.05 for Remainder
mat. 1;
0.12  0.20
for mat. 2
168 Chapter 5

The stress amplitudes are applied as pulses. In-service

loading is simulated according to a cumulative frequency
curve obtained from actual load measurements. The stress
amplitudes of successive pulses are changed according to
this distribution during high frequency simulation of ran-
dom loading. Because of the resonance type loading, single
cycles within the pulses cannot be chosen individually, but
instead pulses of 1000 cycles are used. The sequence length
is 106 pulses; thus, the return period consists of 109 cycles.
The maximum applied stress amplitude smax occurs once
during 106 pulses and is 100 MPa. This value for smax is cho-
sen so that the expected number of cycles to failure is
108  109. The load sequence is stationary, which means that
the almost constant root mean square values of applied smax
prevail during the whole process. If a specimen has not
failed at the end of a full sequence, the experiments are
stopped.
Pauses between pulses allow for the dissipation of the
heat produced by internal friction in specimens during fatigue
loading. The pause length is chosen so that the heating of the
specimens is inhibited. In addition, a fan is used for cooling.
Figures 5.33 and 5.34 (Stanzl, 1993) give the results of
FCG experiments in the threshold regime. The threshold
values for R ¼
1 obtained by reducing
pffiffiffiffiffi the SIF valuespffiffiffiffiffi in
the steps by 5%  7% are 2.8 MPa m and 3.0 MPa m for
materials 1 and 2, respectively.
After determination of the threshold value, loading is
increased stepwise. The specimens are loaded for at least
2  107 cycles at each amplitude level if no crack growth can
be observed. The maximum cyclic p SIF,
ffiffiffiffiffi for which no crack
extension is
pffiffiffiffiffi detected, is 3.15 MPa m for material 1 and
3.7 MPa m for material 2. In this near-threshold
pffiffiffiffiffi region, with
R ¼
1 and Kmax between 3.7 and 5.5 MPa m, the FCG rates
are generally three times lower for material 2 than for
material 1.
Another important result is that cracks remain stopped
when Kmax is somewhat increased continuously after the
threshold value has been obtained in the tests. This may be
partially explained by crack closure effects.
Crack Growth and Threshold 169

Figure 5.33 FCG curve of AlSi cast alloy, material 1 (Stanzl, 1993).

To summarize, the ultrasonic fatigue tests demonstrate

the superiority of material 2 over material 1. This is attribu-
ted to higher mean crack initiation times due to the absence of
the large cast voids.

Figure 5.34 FCG curve of AlSi cast alloy, material 2 (Stanzl, 1993).
170 Chapter 5

5.3.4. Al6061-T6 Alloy Reinforced by Al2O3

Particle (Papakyriacou, 1995)
Higher stiffness than that of conventional aluminium alloys,
improved wear resistance, and superior high temperature
characteristics make aluminium alloys reinforced with SiC
or Al2O3 particles attractive as engineering material. They
are now used in the automobile industry for pistons or
sleeves. Various other applications are possible since produc-
tion and machining of such isotropic composite materials is
possible without too many problems. Knowledge of crack
initiation and propagation processes is necessary to guaran-
tee safe in-service behavior of components. Of special interest
are the fatigue properties at high numbers of cycles and the
FCG behavior at very low crack growth rates since, in auto-
motive applications, components often must endure lifetimes
of up to 108 cycles.
Tested materials were 6061-T6 aluminium alloys with
different contents and size distribution of Al2O3 particles.
Specifically, an alloy with a content of 15.0% volume of fine
particles (6061=Al2O3=15p), an alloy with a content of 21.1%
volume of more coarse Al2O3 particles (6061=Al2O3=
21p) and, for comparison, 6061-T6 without reinforcement
(6061) were tested. Table 5.10 lists the chemical compositions
of these three materials.
The production procedures for the particle reinforced
aluminium alloys are as follows. After extrusion of rods with
cross-sections of 17  17 mm, the reinforced materials are
solution heat treated at 560 C for 0.5 h, and then water
quenched and aged at 160 C for 8 h (heat treatment T6).
The unreinforced aluminium alloy is solution heat treated

Table 5.10 Chemical Compositions of Al6061 Alloys (volume %)

Material Mg Si Cu Fe Mn Cr Zn Ti

6061 0.88 0.69 0.43 0.45 0.13 0.17 0.01 0.07

6061=Al2O3=15p 0.96 0.64 0.26 0.15 0.004 0.105 0.012 0.01
6061=Al2O3=21p 0.95 0.63 0.27 0.08 0.004 0.10 0.009 0.01
Crack Growth and Threshold 171

at 525 C for 0.5 h after extrusion of rods with the same cross-
section dimensions, followed by water quenching and aging at
160 C for 24 h. A quantitative evaluation of mean particle size
and density is listed in Table 5.11, together with mechanical
properties of the materials.
Tests were performed at a frequency about 20 kHz and
load ratio R ¼
1. Cyclic loading is applied in pulses with a
pulse length of 1000 cycles. Pauses between the pulses help
to dissipate the heat caused by internal friction in the speci-
men during loading. The length of the pauses was chosen to
be approximately 20  30 ms to avoid increase of specimen
temperature.
To control cyclic loading the displacement amplitude U0
at the specimen end was used. The maximum SIF Kmax is
obtained from the formula
U0 E pffiffiffi
Kmax ¼ af ða=wÞ
w
To obtain da=dN versus Kmax curves, the load is lowered in
steps of 7% (close to the threshold, 5%) and then increased.
Crack increments of about 300 mm are evaluated for each data
point. At least 2  107 cycles are applied to the specimen in
the threshold regime if no crack advance can be detected.
From this and the optical resolution of the assembly of
7 mm, a crack growth rate below 3.5  10
13 m=cycle is guaran-
teed for characterization of the threshold in the crack growth
curve.
Test results are plotted in Figure 5.35 (Papakyriacou,
1995). The FCG properties of the reinforced alloys are

Table 5.11 Mechanical Properties of 6061 Alloys and Mean

Particle Size
E sy UTS Dcircle Dmax Particles
Material (GPa) (MPa) (MPa) e% (mm) (mm) per mm2

60612 70.4 335 375 14

6061=Al2O3=15p 90.5 340 385 6 8.0 13.3 3500
6061=Al2O3=21p 96.3 365 405 4 11.1 16.8 2100
172 Chapter 5

Figure 5.35 Fatigue crack growth rates vs. Kmax of: (a) unrein-
forced alloy 6061-T6; (b) 6061=Al2O3=15p; and (c) 6061=Al2O3=21p.
A comparison of fatigue crack growth rates vs. Kmax for all three
alloys is given in (d) [Stanzl as reported in (Papakyriacou, 1995)].
Crack Growth and Threshold 173

pffiffiffiffiffi
superior to pure 6061 for Kmax pffiffiffiffivalues
ffi below 5 MPa m for
6061=Al2O3=21p and 6.5 MPa m for 6061=Al2O3=15p, but
worse for high Kmax values. Alloy 6061=Al
pffiffiffiffiffi 2O3=15p shows the
highest threshold SIF of 5.4 MPa m. The threshold pffiffiffiffiffi stress
intensity of 6061=Al2O3=21p is lower (4.7 MPa m) than that
of 6061=Al2O3=15p, but the difference between the crack propa-
gation behavior of the two particle reinforced alloys becomes
smaller with increasing load. Unreinforced
pffiffiffiffiffi 6061 alloy has a
threshold value Kmax of 3.9 MPa m. The slope of the crack pro-
pagation curve of 6061, however, is smaller than for both rein-
forced alloys, thus leading to superior crack growth properties
of the unreinforced material at higher cyclic loads.
To conclude (Papakyriacou, 1995): The threshold values
of Kmax for fatigue crack propagation at load ratio R ¼
1
are increased by the addition of 15% volume fine as well as
21% volume coarse Al2O3 particles to the aluminium alloy
6061. These hard particles improve FCG properties by acting
as obstacles to crack propagation. Firstly, impediment of
crack growth by these obstacles, changes of the crack pro-
pagation direction, and branching of the crack path restrain
fatigue crack growth, thereby, increasing fracture surface
roughness and raising the modulus of elasticity, which is
beneficial. Secondly, Al2O3 particle reinforcement improves
the FCG properties more efficiently in the threshold regime
than at higher Kmax values. At higher loads, particle and
interface fracturing is more frequent, which results in higher
crack growth rates. Finally, the alloy with finer particles
shows better FCG properties than the alloy with coarser
particles. Particle fracture and cracking of the interface
between particles and matrix are both more frequent for
coarse reinforcing particles, which may explain this result.

5.4. MATERIALS OF B.C.C. AND F.C.C.

CRYSTALLINE STRUCTURE
(TSCHEGG, 1981)

(Tschegg, 1981) presents an early study at low temperature.

The aim is to investigate crack growth behavior of b.c.c. (body
174 Chapter 5

centered cubic) and f.c.c. (face centered cubic) metals in the

threshold regime and to correlate their fracture appearance
with crack growth rates in order to explain the differences
in low and high temperature behaviors.
Mild steel is chosen as an example for b.c.c. metals. Rods
are cold drawn from 3–7 mm diameter and cold rolled to bands
with the cross section 1 mm  4 mm and a length of 125 mm.
For f.c.c. metals, 99.9% commercial copper and austenitic
AISI type 304 stainless steel are used. At 77 K, steel 304 exhi-
bits rather large amounts of strain induced martensite.
Both materials are cold rolled to bands of cross section
1 mm  10 mm.
The chemical compositions, heat treatments, and
mechanical properties of these materials are listed in Tables
5.12 through 5.14.
To study the influence of the surrounding temperature
on crack propagation the samples are immersed in silicone
oil at 273 K (20 C) and liquid nitrogen at 77 K, respectively.
Silicone oil is a noncorrosive liquid and mainly considered as
replacing vacuum conditions. Experiments are repeated to
compare measurements in vacuum and silicon oil extensively.
In order to avoid temperature rises due to damping
effects at very high frequency, the stressing is done in a
pulsed manner. Pulse packets and intervals are chosen so
that the temperature rises at the sample surface are less than
5 C.
For DK calculation, only the tensile part of the stress is
used, assuming that the compressive load does not contribute

Table 5.12 Chemical Compositions of Mild Steel, Copper, and

Steel 304 (wt%)
C Si Mn P S Al N O
Mild steel 0.036 0.01 0.08 0.012 0.008 0.002 0.005 0.015

Cu
Copper 99.9

C Cr Ni Si Mn Mo
Steel 304 0.036 19.04 10.40 0.75 1.18 0.48
Crack Growth and Threshold 175

Table 5.13 Heat Treatments of Mild Steel, Copper, and Steel 304
Microstructure and
Heat treatment grain size

Mild steel 700 C=90 min furnace-cooling Ferrite: 15 lm

Copper 630 C=90 min furnace-cooling Grain size: 20 lm
Steel 304 1050 C=30 min water-quenched Austenite þ small
amounts ferrite

Table 5.14 Mechanical Properties of Mild Steel, Copper, and

Steel 304
sy UTS eu
(MPa) (MPa) (%)

273 K Mild steel 275 325 35

Copper 69 235 48
Steel 304 230 700 55
77 K Mild steel 840 6
Copper 83 365 60
Steel 304 420 1500 39

to crack propagation. This means that an eventual contribu-

tion of the compressive cycle toward DK in the near threshold
growth is not taken into account.
Tests are performed at about 20 kHz and symmetric load,
R ¼
1. The dependence of the FCG rate on stress intensities
at 293 K and 77 K are compared in Figures 5.36, 5.37, and
5.38 for f.c.c. copper, steel 304, and b.c.c. mild steel, respec-
tively.
It is emphasized that the scattering of the data points is
caused partly by the method of measuring; i.e., the stressing
amplitude is not kept constant during the tests but lowered
and raised repeatedly so as to attain very low Kmax values.
Therefore, crack growth rates are obtained from single crack
increments and not from a smoothed curve. A second reason
for the scatter of the results is discontinuous crack growth.
For calculating da=dN curves, crack increments of about
 XXXX-0 Bathias Ch05 R3 081204

176 Chapter 5

Figure 5.36 Influence of temperature (293 K silicone oil and 77 K

liquid nitrogen) on fatigue crack growth rates in polycrystalline
copper (21 kHz; R ¼
1) (Tschegg, 1981).

100 mm are used. Crack growth rates below about 10
10 m=

cycle are mean values of higher crack propagation rates and
crack stops. Thus it is understandable that crack propagation
rates of less than one Burgers vector per cycle on average can
occur. Data points with arrows refer to cracks that did not
propagate 10 mm of measurement accuracy at minimum.
The crack growth rates of these points are calculated by divid-
ing 10 mm by the applied number of cycles.
The main result shown in Figure 5.36 for polycrystalline
copper is that the crack growth curve is shifted to somewhat
higher stress intensities at 77 K compared to 293 K; that is,
the crack growth rates are a bit lower for identical stress
intensities. A threshold stress intensity seems to exist for
Crack Growth and Threshold 177

Figure 5.37 Influence of temperature (293 K silicone oil and 77 K

liquid nitrogen) on fatigue crack growth rates in steel 304 (21 kHz;
R ¼
1) (Tschegg, 1981).

pffiffiffiffiffi pffiffiffiffiffi
both temperatures; it is 2.7 MPa m and about 3 MPa m at
293 K and 77 K, respectively.
Crack propagation curves of steel 304 in Figure 5.37
show essentially the same result as copper. At 77 K some-
what higher stress intensities Kmax are necessary to gain
the same crack growth rates. Again, a threshold stressp inten-
ffiffiffiffiffi
sity seems to exist p for
ffiffiffiffi
ffi both temperatures; it is 7 MPa m at
293 K and 8.5 MPa m at 77 K, respectively; i.e., it is about
20% higher at 77 K for steel 304.
For crack propagation curves of mild steel in Figure 5.38,
the differences of crack growth rates at 77 K and 293 K are
more pronounced than in f.c.c. copper p and
ffiffiffiffiffi steel 304. The

thresholdpffiffiffiffistress
ffi intensity is 3.8 MPa m at 293 K and
5.3 MPa m at 77 K, respectively. The increase is about 40%.
178 Chapter 5

Figure 5.38 Influence of temperature (293 K silicone oil and

77 Kelvin liquid nitrogen) on fatigue crack growth rates in mild
steel (21 kHz; R ¼
1) (Tschegg, 1981).

To summarize (Tschegg, 1981), low temperature causes a

shift of FCG curves to higher stress intensity Kmax. This is
found for polycrystalline f.c.c. copper and steel 304 as well
as b.c.c. mild steel and might be explained by the increased
tensile properties. The fracture mode of mild steel pisffiffiffiffifficlea-
vaged completely at stress intensities Kmax  20 MPa m and
cleavage at crack growth rates da=dN > 10
8 m=cycle, but
by reversed plasticity at lower values. It pisffiffiffifficoncluded
ffi that
sometimes below intensity Kmax ¼ 20 MPa m the stress is
obviously too low to cause cleavage, but then crack propaga-
tion is still possible in this case by a reversed plasticity
mechanism.
Crack Growth and Threshold 179

5.5. LOW CARBON STEEL SHEET

The cold-rolled low carbon steel sheet is produced from

pickled hot rolled by cold reduction to the desired thickness.
Owing to its advantages of hardness and much thinner thick-
ness, cold rolled steel sheets are widely used in the auto-
motive industry, which in turn leads to a reduction in weight.
In the automotive industry, most components are required to
have a very long fatigue lifetime both for safety and economic
reasons, hence it is necessary to well describe the fatigue
crack growth behaviors in the near-threshold regime.
We discussed some special testing measures adapted to
this type of thin specimen in Chapter 3.
The test is conducted at a stress ratio R ¼ 0.1 at a fre-
quency of 20 KHz and at room temperature. The thickness
of the cold-rolled steel sheet is 0.75 mm. The chemical compo-
sition and mechanical properties obtained in static tests were
listed in Tables 5.15 and 5.16, respectively; the threshold of
this material is determined at a rate of 10
11 m=cycle.
In the test, the specimen is first pre-cracked without sta-
tic load, R ¼
1. Then the experiments are performed with a
constant stess ratio of R ¼ Kmin=Kmax ¼ 0.1 by decreasing pro-
gressively the static load and the vibration amplitude from
5% to 10%. When the threshold is reached, experiments con-
tinue by increasing the stress intensity factor.

Table 5.15 Composition of Cold-Rolled Low Carbon Steel

Sheet (%wt)
Cu C Mn Si P S Al

0.14 0.08 0.4 0.1 0.025 0.025 0.02

Table 5.16 Mechanical Properties of Cold-Rolled Low Carbon

Steel Sheet
sy (MPa) UTS (MPa) eu % Hv5 E(GPa) r (kg=m3)

225 340 36 95 203 7830

180 Chapter 5

The results of the tests are shown in Figure 5.39. The

thresholdp value is found at a stress intensity factor of
5.0 MPa m with a propagation rate of 7  10
12 m=cycle.
In the study, temperature is measured at the crack tip by
the infrared thermograph method. It was found that the rise
of temperature depended upon the material microstructure
and the amplitude of vibration loading. For the fatigue crack
growth test in the near-threshold regime, the loading ampli-
tude is very small and the temperature rise is not very
significant with the flow of compressed air (at most 60 C;
Figure 5.40). In this case, the absorption of ultrasonic vibra-
tion energy has no considerable effect on the fatigue crack
growth behavior.
Fractographic analyses of the fracture surface are
grouped in Figure 5.41. No evident fatigue striation is
observed on the whole fracture surface. This is probably
because of the small amplitude of vibration. At high DK, the
fraction of transgranular rupture is observed in Figure 5.41a.
For the moderate values of DK, the fracture surface exhibits a
mixed mode of transgranular and ductile intergranular
rupture, and several secondary cracks can be seen at

Figure 5.39 FCG rate of the thin steel sheet (R ¼ 0.1).

Crack Growth and Threshold 181

Figure 5.40 Temperature rise in the specimen of an ultrasonic

fatigue test.

intergranular locations, (Figure 5.41b). In the near threshold

region, the ductile intergranular rupture becomes predomi-
nant and the fracture surface is smoother (Figure 5.41c).

5.6. AUSTENITIC STAINLESS STEEL (SUN, 1999)

Test results for 304 steel are presented on Figure 5.42 as a

plot of fatigue crack growth rate da=dN vs stress intensity fac-
tor range, DK ¼ Kmax
Kmin. The propagation threshold was
determined at a very low growth rate down to
10
12 mm=cycle. For R ¼
1, Kmin is taken to be zero.

5.6.1. Fatigue Crack Growth Rates

and the Threshold
Ultrasonic fatigue endurance experiments have shown that,
for a great number of alloys, fatigue failure can still occur
at 109 cycles or beyond, and the difference of fatigue resis-
tance between 106 and 109 can reach 100 or even 200 MPa
182 Chapter 5

Figure 5.41 SEM of fracture surface of low carbon steel at 20 kHz.

(Bathias, 2003). But the ultrasonic fatigue crack growth test

results show that there is very small difference between the
thresholds determined at the rates of 10
9 and 10
7 mm=cycle.
In this study,
pffiffiffiffiffi this difference in DK threshold is only about
0.2 MPa m. Some other studies on different alloys, such as
nickel alloys (Bathias, 1993), titanium alloys (Bathias, 1994,
1997), aluminium alloys (Sun, 2001), and steels (Sun, 2001),
have arrived at much the same conclusions. This means that
the fatigue crack growth threshold determined by conven-
tional fatigue testing is reliable for the engineering design.
It is obvious that ultrasonic fatigue is time-saving and practi-
cal; for example, threshold measurements can be obtained
within a few hours of testing at 20 kHz. Moreover, the
Crack Growth and Threshold 183

Figure 5.42 Ultrasonic fatigue crack propagation test results for

304 steel at 20 kHz.

ultrasonic vibration fatigue technique is especially useful for

tests at very high stress ratios.

5.6.2. Effect of Stress Ratio, R

The effect of the stress ratio on the fatigue crack propagation
behavior at ultrasonic frequency is illustrated in Figures 5.42
and 5.43. As expected in conventional fatigue tests, high load
ratio induced lower threshold values and faster growth rates
at a given applied DK. This relation can be described as:
DKth ¼ f(R)DKth 0 R0 ¼ Kopen=DK0
with f(R) ¼ 1
R for 0 < R < R0
f(R) ¼ R0 for R0 < R < 1
f(R) ¼ 1 þ 0.3 R for
1 < R < 0

The influence of R-ratio on the fatigue crack growth behavior

can generally be explained by the crack closure effects, as pro-
184 Chapter 5

Figure 5.43 Dependence of threshold on R ratio (see also

Figures 5.5 to 5.7).

posed by (Elber, 1971) and (Schmidt, 1973). However, in ultra-

sonic fatigue with mean load equal to zero (R ¼
1), the closure
effect becomes very small when we consider only the tensile
part of the cycle load (Kmin ¼ 0). Figure 5.43 shows that the
results for R ¼
1 are close to those for R ¼ 0.5 and R ¼ 0.7.
As originally suggested by (Schmidt, 1973), if the varia-
tion of threshold with load ratio is simply due to the crack
closure, and if the requirement to induce crack closure is inde-
pendent of load ratio, a transition behavior from the threshold
at low R ratio to that at high R ratio could be expected at
Kmin,th ¼ Kcl. From
pffiffiffiffiffi Figure 5.44 this value was estimated as
Kcl ¼ 3–4 MPa m. (See also Figures 5.5 to 5.7.)

5.6.3. Fatigue Crack Growth Mechanisms at

Ultrasonic Frequency
The SEM observations of fracture surfaces permit one to note
the fatigue crack growth mechanisms for ultrasonic frequency
at different R ratios. For this stainless steel, at
threshold regime, the rupture appears transgranular and
crystallographic, but the fracture surfaces are smoother and
finer at R ¼ 0 and 0.1 because of the crack closure effect
(Figure 5.45). For a higher rate of propagation to 2 or
3  10
6 mm=cycle, some quasi-cleavage has been found
Crack Growth and Threshold 185

Figure 5.44 Combination of (a) Kmax,th vs. R; (b) DKth vs. R; and
(c) DKth vs. Kmax,th
186 Chapter 5

Figure 5.45 SEM observations at near threshold regime.

locally at high R ratios (Figure 5.46). For a even higher propa-

gation rate of 2  10
5 mm=cycle, fatigue striations can be
found for R ¼
1, 0.5, 0.7 (Figure 5.47) but they are difficult
to identify for R ¼ 0 and R ¼ 0.1.

5.7. SPHEROIDAL GRAPHITE CAST IRON (SGI)

Figure 5.48 shows the SGI results of FCP rates vs. stress
intensity factor. For R ¼
1 in ultrasonic fatigue, DK ¼ Kmax

Figure 5.46 SEM observations of quasi-cleavage at R ¼ 0.7;

da=dN ¼ 2  10
6 mm=cycle.
Crack Growth and Threshold 187

Figure 5.47 Fatigue striations at higher crack propagation rate.

Figure 5.48 Fatigue crack propagation test results at ultrasonic

frequency in cast iron.
188 Chapter 5

is assumed because only the tension part of the cycle contri-

butes to the propagation
p of crack. The threshold stress p inten-
sity is 3.8 MPa m for R ¼
1 and DKth ¼ 6.3 MPa m for
R ¼ 0.1. On the other hand, it can be seen from Figure 5.48
that there is very little difference for the rates above thresh-
old beyond 3  10
10 m=cycle.
The effect of frequency on FCP behavior has been studied
by a number of researchers for various materials, showing
that no significant effect of frequency was observed on the
threshold values tested in ambient environment. For this
SGI, a conventional FCP test has been conducted at a fre-
quency of 35 Hz by Nadot (1999) in both ambient environment
and vacuum conditions. The test results were compared in
Figure 5.49 with the ultrasonic fatigue
p tests. It is shown that
the threshold atp35 Hz (8.6 MPa m) is higher than that at
20 kHz (6.3 MPa m). In contrast, in the high propagation
rate regime, crack propagation rate is higher at 35 Hz than
at 20 kHz for a given DK value.

Figure 5.49 Effect of test frequency on fatigue crack growth in

cast iron.
Crack Growth and Threshold 189

After fatigue testing, some specimens were broken and

carefully cleaned in alcohol in an ultrasonic bath before taken
to the SEM examination. Other specimens were cut in the
mid-thickness location parallel to the specimen sides. The
crack profile, after etching in 3% nital, was observed under
SEM in that condition.
The fractographic observations indicated that at a high
propagation rate (da=dN > 4  10
9 m=cycle), this SGI speci-
men tested at ultrasonic frequency propagated in a trans-
granular mode with some quasi-cleavage facets (Figure 5.50).
At an intermediate propagation rate, the quasi-cleavage
facets disappeared and the rupture mode is mainly trans-
granular (Figure 5.51). In the threshold regime (below 3 
10
10 m=cycle), intergranular failure was observed. However,
the transgranular rupture was always present on the fracture
surface, even at very low propagation rates (Figure 5.52).

5.8. DATABASE OF THRESHOLD SIF DKth

Table 5.17 contains values of threshold SIF for some materials

and alloys for practical and industrial use obtained by 20 kHz
ultrasonic fatigue techniques. Test conditions are also reported

Figure 5.50 Transgranular

p quasi-cleavage at high propagation
rate DK ¼ 15.5 MPa m; da=dN ¼ 4  10
9 m=cycle.
190 Chapter 5

p
Figure 5.51 Transgranular rupture DK ¼ 8 MPa m; da=dN ¼
9.2  10
10 m=cycle.

where available. The data in Table 5.17 include the results dis-
cussed in this book and other sources. Some rather early data
(Stanzl, 1996) are also listed; all sources are cited and footnoted.
Readers should be aware that the data given Table 5.17
depend strongly on the material properties, heat treatment,

Figure 5.52 pMix rupture of transgranular and intergranular,

DK ¼ 6.5 MPa m; da=dN ¼ 2.0  10
10 m=cycle.
Table
p 5.17 Threshold Stress Intensity Factors Obtained by Ultrasonic Fatigue Techniques (DKth in
MPa m)
Tables (chemical
composition,
heat treatment,
da=dN Source da=dN and mechanical
Material R Environment Temperaturea DKth (m=cycle) (see footnote) Figure properties) Remarks

11
T6A4V 0 Air 5.66 9  10 (1) 4.22 4.14  4.16

1 Air 3.2 5  10
11 (1) 4.22
Astroloy
1 Air 20 C 6.0 1  10
9 (2) 4.26 4.22  4.24
Crack Growth and Threshold

1 400 C 5.0 1  10
9 (2) 4.26

N18
1 Air 450 C 5.5 1  10
9 (3) 4.27 4.19  4.21
0 Air 450 C 8 1  10
9 (3) 4.27
0 Air 700 C 7 1  10
7 (3) 4.28
0.8 Air 450 C 4.5 1  10
9 (3) 4.27
Inconel706 0.1 400 C 6.4 1  10
9 (3) 4.29 4.25  4.27
0.1 nitrogen 400 C 5.5 1  10
9 (3) 4.31
0.1 sweep 20 C 7 5  10
10 (3) 4.29
AlSiII-1
1 2.8 2  10
13 (4) 4.39 4.30
AlSiII-2
1 3.0 2  10
13 (4) 4.40 4.30
6061
1 3.9 3  10
13 (5) 4.41 4.31
6061=Al2O3=15p
1 5.4 3  10
13 (5) 4.41 4.32
6061=Al2O3=21p
1 4.7 3  10
13 (5) 4.41
f.c.c. Cu
1 silicon oil 20 C 2.7 7.4  10
14 (6) 4.42 4.33  4.35

1 liquid N2 77 K 3 9  10
14 (6) 4.42
f.c.c.steel 304
1 silicon oil 20 C 7 8.5  10
13 (6) 4.43 4.33  4.35

1 nitrogen 77 K 8.5 1  10
12 (6) 4.43
b.c.c. mild steel
1 silicon oil 20 C 3.8 8  10
14 (6) 4.44 4.33  4.35

1 nitrogen 77 K 5.3 1  10
13 4.44
Al
1 oil 20 C 1.0 8  10
13 (7)

1 air 20 C 1.33 1  10
13 (7) Grain size and cold
work effects
191

(Continued)
Table 5.17 (Continued ) 192

Tables (chemical
composition,
heat treatment,
da=dN Source da=dN and mechanical
a
Material R Environment Temperature DKth (m=cycle) (see footnote) Figure properties) Remarks

AlMg5
1 oil 20 C 1.4 3  10
13 (7)

AlZnMg1
1 oil 20 C 1.9 7  10
13 (7)
Comparison with
Cu
1 Oil 20 C 2.7 4  10
14 (7) NaCl-solutions

1 liquid N2 77 K 2.5 4  10
14 (7)

1 air 20 C 1.4  2.0 1  10
13 (7) Grain size effect;
1.4  2.6 1  10
15 (7) cold work effect;
1.8  2.3 1  10
13 (7) single crystals
Comparison with
Low C Steel
1 oil 20 C 3.8 6  10
14 (7) NaCl-solutions

1 liquid N2 77 K 5.3 9  10
14 (7)
pffiffiffiffiffi

1 liquid N2 20 K 7.2 1  10
10 (7) 4.1 MPa m for 70 Hz
AISI304
1 oil 20 C 7.0 5  10
13 (7) Comparison With

1 liquid N2 77 K 8.5 9  10
13 (7) NaCl-Solutions
X10Cr13
1 oil 23 C 6.7 6  10
14 (7)
X20Cr13
1 humid air 23 C 3.6 4.4  10
13 (8)
XXXX-0 Bathias Ch05 R3 081204

GGG 100-B
1 humid air 23 C 5.7 4.4  10
13 (8)

34CrMo4
1 air 20 C 2.45 1  10
12 (7)
Ck60
1 air 20 C 2.44 1  10
12 (7)
PM-Mo
1 air 20 C 4.8  5.2 1  10
13 (7) Grain size effect
PM-Mo-0.8W
1 air 20 C 5.8 1  10
13 (7)
PM-Mo-1.5W
1 air 20 C 6.6 1  10
13 (7)
PM-Mo-Ti-Zr
1 air 20 C 7.0 1  10
13
PM-Mb
1 air 40 C 1.8 1  10
13 (7)
PM-Ta
1 air 40 C 3.7 1  10
13 (7)
A286
1 20 C 13.0 3  10
12 (7)
Chapter 5
IN-738
1 air 20 C 3.13 1  10
12 (7)
IN-792
1 air 20 C 4.48 1  10
12
no threshold
U-700 s.c.
1 air 20 C down to 10–11 (7)
(7)
INCO.800 Transv. 0.1  0.3 air 20 C 4 1  10
13 (7)
no threshold
HASR.X Transv. 0.1 air 20 C down to 10–10 (7)
IN 600 Transv. 0.3 air 20 C 5 1  10
11 (7)
CSN 412013
1 water 20 C 7.7 1  10
11 (9)
Steel
1 argon 200 C 6 1  10
11 (9)

1 argon 250 C 5 1  10
11 (9)
Crack Growth and Threshold

1 argon 300 C 5 1  10
11 (9)

1 argon 400 C 4.2 1  10
11 (9)

1 argon 500 C 3.05 1  10
11
CSN 415313
1 water 20 C 5.3 1  10
11 (9)
Steel
1 argon 200 C 4.7 1  10
11 (9)

1 argon 250 C 4.4 1  10
11 (9)

1 argon 300 C 4.3 1  10
11 (9)

1 argon 400 C 3.85 1  10
11

1 argon 500 C 3.4 1  10
11
Cr–Si
1 air 20 C 7 1  10
11 (10)
U500
1 air 20 C 13 2  10
6 (11)

1 air 20 C 8 3  10
8 (11)
Al-Li8090
1 air 20 C 7 2  10
8 (11) 4.17, 4.18, 4.38
Al2024
1 vacuum 3.3 2  10
13 (12) 4.33 4.28

1 humid air 2.1 2  10
13 (12) 4.33 4.29

1 dry air 2.3 2  10
13 (13)
Mo-3
1 air 25 C 5.2  0.5 < 10
13 (14) Ed ¼ 322 GPa
Mo-0.8W
1 air 25 C 5.9  0.4 < 10
13 (14) Ed ¼ 318 GPa
Mo–Ti–Zr
1 air 25 C 7.0  0.5 < 10
13 (14) Ed ¼ 314 GPa

a
Ambient temperature where not indicated.
References: 1: Jago, 1993; 2: Wu, 1994; 3: Bonis, 1997; 4: Stanzl, 1995; 5: Papakyriacou, 1995; 6: Tschegg, 1981; 7: Stickler, 1982; 8:
193

Mayer, 1995; 9: Sun, 2001; 10: Kong, 1991; 11: Mayer, 1992; 12: Stanzl, 1991; 13: Weiss, 1982.
References: 1: Jago, 1993; 2: Wu, 1994; 3: Bonis, 1997; etc.
194 Chapter 5

test temperature, and environment. Therefore, before using

these ultrasonic fatigue threshold data, suitable references
should be consulted. For R ¼
1, the compression part of the
load cycle was neglected in computing the threshold.

5.9. OTHER APPLICATIONS: FRETTING

FATIGUE
5.9.1. Fretting Fatigue of Aluminium–Lithium
Alloy and Titanium
Fretting fatigue experiments have been performed on
aluminium–lithium alloy Al-Li8090 and titanium alloy T6A4V
(Tao, 1996). Chemical composition, heat treatment, and
mechanical properties of T6A4V can be found in Tables 4.20
through 4.22, while Tables 4.23 and 5.18 list those of Al-
Li8090. Some mechanical properties differ from those given
in Table 5.18, as the heat treatments are not the same.
Aluminium alloys of high strength, such as Al-Li8090,
are widely used in aeronautic industries. At the same time
the menace of fretting to the safety of aeroplane structures
is of great concern (Kuzmenko, 1984). More than twenty years
ago it was estimated that the existence of a fretting zone was
responsible for 90% of fatigue damage (Tein, 1975).
As has been indicated, titanium and its alloys like
T6A4V, are ideal materials for aeronautical and aerospace
equipment. But the sensitivity of these materials to fretting
is so high it is an obstacle to their utilization, particularly
the latter. Fretting can reduce the fatigue strength of tita-
nium alloys by 40% to 70% (Li, 1992; Lutynski, 1982).
Specimens and pads of the two materials are used in
order to form the following three types of experiments:

Table 5.18 Heat Treatment and Mechanical Properties of

Al-Li8090
Heat treatment sm (MPa) s0.2 (MPa) E (GPa) r (kg=m3)

2h30 in 106 C 434 352 80 2530

Crack Growth and Threshold 195

1. Specimen of Al-Li8090, pads of T6A4V

2. Specimen of Al-Li8090, pads of AL-Li8090
3. Specimen of T6A4V, pads of T6A4V

The geometry of the pads and specimens is given in

Figure 3.24 on p. 79.
All the tests are carried out at 20 kHz, ambient tempera-
ture of 25 C, and humidity of about 60%. Before each test the
position of pads is chosen and adjusted; after the test the posi-
tion of pads is measured. The relative slip amplitude and
stress si are re-calculated.
With the above three types of specimen–pad combina-
tions, we designed five groups of tests with conditions and
results listed in Table 5.19.
Test 1. After 1.2  107 cycles, we observe typical signs
of fretting. On the specimen surface the platelets and small
cracks due to fretting are evident. The debris and the surface
are oxidized.
At the same time fretting also takes place on the surface
of pads of T6A4V; there are a lot of platelets and debris.
The possibility of fretting at very high frequency is con-
firmed by these results, and alloy T6A4V seems sensitive to
fretting at this high frequency of 20 kHz.
Test 2. After 6  106 cycles, fretting occurs in the speci-
mens (Figure 5.53). The signs of fretting are larger and more
distinctive, because the amplitude is higher. From Ai ¼ 0.2 mm
to Ai ¼ 0.5 mm, in the border of the friction zone we can dis-
tinguish the signs of successive stages of fretting: stripes,
platelets, and small cracks. In some stripes we find thin plate-
lets. When Ai ¼ 1.2 mm, the fretting is significant and the sur-
face is covered with platelets and ejected oxidised debris.
The relative slip amplitude constitutes an important
factor in vibration fretting as in classical fretting.
Test 3. Fatigue lives of the fretting fatigue specimens
are Nf ¼ 4.8  106 and 3  106 cycles. The two specimens were
cracked in fatigue after the fretting where the stress
si ¼ 158 MPa and 236 MPa.
The specimens for comparison without fretting were
not broken after Ni ¼ 5  107 cycles for smax ¼ 161 MPa
Table 5.19 Conditions and Results of the Tests 196

Material FN A0 xi Ai smax si Ni or
Test (specimen–pads) (N) (mm) (mm) (mm) (MPa) (MPa) Nf (107) Notes

1 8090–T6A4V 30 15 35 7 26.8 19 Ni ¼ 1.2 Fretting only, in order to

confirm the possibility of
fretting at very high frequency.
2-1 8090–8090 15 3 3 0.2 5.36 5.35 Ni ¼ 0.6 Fretting only, in order to observe
2-2 15 3 5 0.34 5.36 5.35 Ni ¼ 0.6 the influence of the relative slip
2-3 15 3 7.5 0.5 5.36 5.29 Ni ¼ 0.6 amplitude in the case of Al-Li 8090.
2-4 15 3 18.5 1.2 5.36 4.91 Ni ¼ 0.6 Weak influence of stress is
not considered.
3-1 8090–8090 15 90 9 18 161 158 Nf ¼ 0.48 3-1 and 3-2: Fretting–fatigue,
3-2 15 135 9 27 241 236 Nf ¼ 0.3 in order to study the influence
of fretting at very high frequency
on the endurance in the case of
Al-Li 8090.
3-3 90 161 Ni ¼ 5 3-3 and 3-4: Same tests with same
3-4 135 241 Ni ¼ 5 parameters of vibration but without
the contact of pads, in order to
determine the importance of the
fretting effect.
Chapter 5
4-1 T6A4V–T6A4V 15 10 1.2 0.3 27.7 27.69 Nf ¼ 0.6 Fretting only, in order to observe
4-2 15 10 3.2 0.8 27.7 27.6 Nf ¼ 0.6 the influence of the relative slip
4-3 15 10 8.0 2 27.7 27.15 Ni ¼ 0.6 amplitude in the case of T6A4V.
4-4 15 10 19.5 4.7 27.7 24.3 Ni ¼ 0.6 Weak influence of stress is not
4-5 15 20 29.5 13.5 54.4 40.8 Ni ¼ 0.6 considered.
5-1 T6A4V–T6A4V 15 100 2.0 5 277 276 Nf ¼ 0.24 5-1: Fretting–fatigue, in order to
study the influence of fretting
at very high frequency on the
endurance in the case of
T6A4V.
Crack Growth and Threshold

5-2 100 277 Nf ¼ 0.6 5-2: Same tests with same

parameters of vibration but
without the contact of pads,
in order to determine the
importance of the fretting effect.

FN ¼ normal load on the pads; Ni ¼ number of cycles on the testing wear; Nf ¼ number of cycles on the life wear. See also Figure 3.24.
197
198 Chapter 5

Figure 5.53 Fretting scar on T6A4V (N ¼ 6106; Ai ¼ 0.2 m).

and 241 MPa, much higher than that of fretting fatigue

tests.
The conjugation of the platelets and small cracks form
large cracks that develop in the specimens and finally provoke
a rupture. Around the rupture it is seen that fretting develops
to a crack. In this way, at high frequency, the vibration fret-
ting initiates a crack earlier and reduces the fatigue strength
of the material, as in the case of classical fretting.
Test 4. The results in Figure 5.54 (Sun, 2001) are simi-
lar to those of Test 2. Again, the relative slip amplitude influ-
ences fretting.
Alloy T6A4V exhibits a sensitivity to fretting at high fre-
quency. Even for low relative slip amplitude of 0.3 mm, plate-
lets are produced. As the relative slip amplitude becomes
higher than 0.8 mm, fretting was much more pronounced.
Test 5. For the fretting fatigue specimen with fatigue
life Nf ¼ 2.4  106 cycles, crack initiation happens at the point
of fretting. Figure 5.55 (Sun, 2001) shows that fretting
appears in the crack.
The fatigue life of the comparison specimen without fret-
ting was Nf ¼ 6  106 cycles. This specimen was broken at
point xi ¼ 0 where the stress has the highest value.
Crack Growth and Threshold 199

Figure 5.54 Fretting scar on T6A4V (Ai ¼ 0.8 mm).

These preliminary results demonstrate that vibration

fretting can substantially decrease the fatigue strength in
the case of T6A4V. This material should be considered as very
sensitive to classical fretting.

Figure 5.55 Fretting scar and fatigue crack on T6A4V.

200 Chapter 5

In conclusion to results discussed above, the following

remarks are made:

 Fretting occurs under vibration friction at very high

frequency of 20 kHz.
 The relative slip amplitude is an important factor. The
larger the amplitude, the more typical and more
pronounced the signs of fretting become. In alloy
Al-Li 8090, the fretting is very detrimental when the
amplitude exceeds 1 mm.
 Fretting due to high frequency triggers earlier cracks
and leads to rupture, if the stress is high enough. Fati-
gue strength is reduced by vibration fretting as well as
by conventional fretting.
 Alloy T6A4V is very sensitive to vibration fretting as
well as to conventional fretting.
 The test method of vibration fretting is very rapid,
and can be performed with variable displace-
ment (0.1 m to a few dozen microns), and with or
without fatigue stress. With this method, studies on
very high cycle fretting fatigue strength can be
carried out.

5.9.2. Fretting Fatigue of High Strength Steel

The material studied was low alloy and high strength steel
42CrMo4U or 4240. (For chemical composition, heat treatment,
and mechanical properties, see Tables 4.1 through 4.3.)
The specimen for ultrasonic fretting fatigue had a cylind-
rical profile with different diameters and was asymmetrical to
amplify the fatigue stress in the gauge length. (See the
distribution of the vibration displacement and stress pre-
sented in Figure 3.24.) The specific length, L, is determined
for a specimen with a resonance frequency of the first longi-
tudinal vibration mode of 20 kHZ (Sun, 1999).
L ¼ X1 þ X2
 
p L1 L2
¼
S þ
k S1 S2
Crack Growth and Threshold 201

where k is a constant
rffiffiffiffiffiffi
r
k ¼ 2pf
Ed
and S is the section area of the cylinder.
In the test, a maximum displacement is achieved at free
ends while the maximum strain (stress) is obtained in the
gauge length of the specimen (Figure 5.56). In this test sys-
tem, the fretting slip amplitude and the fatigue stress are
the vibration displacement and vibration stress, respectively,
at the point on the gauge length of the specimen where the
pads are placed. They depend upon the position of the pad
and the maximum vibration amplitude of the specimen. The
latter is determined by the power of the generator and the
amplification of the horn. In our experiments, this can vary
from 3 mm to 95 mm. By regulating the position of the pads
along the specimen and by changing the power of the genera-
tor, either the slip amplitude or the fatigue stress or both can
be changed. As a result, these two parameters are dissociated
(Sun, 1999).
The fretting pad also has a cylindrical surface (Figure
5.57); it consists of the same materials as the specimen. The
pads are held on the specimen by two springs. The sides
normal contact force is Fn ¼ 30 N and the slip amplitude is
about 17 mm.
The conventional method to establish the important vari-
ables that can affect fretting fatigue is to generate S-N curves
with and without fretting, allowing fretting fatigue strength

Figure 5.56 Ultrasonic fatigue and the distribution of vibration

displacement and stress.
202 Chapter 5

Figure 5.57 Contact pad.

reduction factors to be evaluated. Such a curve is given for

two 4240 steels (B and C in Figure 5.58), which reveals that
fatigue strength is significantly reduced by fretting fatigue,
and the factor of reduction is of the order of 3 but varies with
the logarithm of the number of cycles in a linear relation
(Figure 5.59). Moreover it is found that the reduction is
greater on Rep-C (compared with Figure 4.2) with a higher
tensile strength. These tests results agree well with others
studies (Lindley, 1997; Li, 1992; Nakazawa, 1992).
Many studies have shown that fatigue initiation is likely
to occur at the surface. In ultrasonic fatigue, crack initiations

Figure 5.58 Fatigue S-N curve with and without fretting.

Crack Growth and Threshold 203

Figure 5.59 Fatigue strength reduction factor caused by fretting.

were observed at the surface when the number of cycles was

less than 107; otherwise, crack initiation occurs on the inter-
nal defects. In fretting fatigue, however, the initiation of fret-
ting fatigue cracks is caused by the surface stress resulting
from frictional forces, together with the bulk stress; they
always occurs at the contact surface (Figure 5.60). Therefore,
there is a consequent sharp reduction of fatigue strength in
the high cycle regime.
In conventional fretting fatigue tests it has been shown
that cracks initiate at a very early stage (5–10%) of fretting
fatigue life (Lindley, 1997; Li, 1992; Nakazawa, 1992). To
define well fretting fatigue crack initiation and propagation
conditions at ultrasonic frequency, a two-stage test is per-
formed in which the fretting contact pads are applied on the
specimen for a certain number of cycles in the fretting fatigue
test and then removed, and subsequently the specimen tested
in plain fatigue (fatigue without further fretting). The results
are shown in Table 5.20. In the first test with a very low fati-
gue stress, no crack was observed at the very early stage. This
204 Chapter 5

Figure 5.60 SEM of fretting fracture surface.

means that the fretting is a combined effect of the tangential

friction force and the fatigue bulk stress.
Both of these two parameters contribute to the crack
initiation in fretting fatigue. On the other hand, from the
results in Table 5.20, it can be seen that when the number
of cycles in the fretting period is lower than a certain limit,
failure does not occur in the following plain fatigue, even after
a very long time. In ultrasonic fretting fatigue, this limit is
higher (more than 50%), which implies that fretting fatigue
crack initiation takes a great fraction of the whole fatigue life,
as in the ultrasonic plain fatigue (Sun, 2001).
In conventional fretting fatigue, some analyses and
experiments show that the crack initiates at the contact edge

Table 5.20 Two-stage Fretting Fatigue Test for 4240 Rep-B

Fatigue
stress N1=Nf
Specimen (MPa) N1 N2 Nf (%) Observation

No. 1 172.9 4.4693  106 2.7782  109 2.2563  109 0.20 Nonrupture
No. 2 198.3 6.5792  107 2.5710  109 2.8659  108 23.0 Nonrupture
No. 3 221.9 2.0006  107 1.3757  109 3.9383  107 50.8 Nonrupture
Crack Growth and Threshold 205

where the tangential force and the slip amplitude are maxi-
mum (Wright, 1971; Realigns, 1972). But others (Waterhouse,
1971; Nakazawa, 1992) observed that a crack was initiated at
the boundary between an area of slip and an area of non-slip.
A possible interpretation is that if relative slip is large enough
to induce severe fretting wear, the rigidity of the surface con-
tact layer in the slip region will be significantly reduced
(Mutoh, 1995). Therefore in the whole slip region the contact
force and hence the tangential force are decreased. However,
a very high concentration of contact and tangential forces will
occur in the slip region near the boundary between slip and
non-slip regions. It is thought that the crack initiation point
depends upon the condition of contact. In this ultrasonic fret-
ting fatigue test where a particular contact of cylinder-on-
cylinder was used, the contact width is increased with the
number of cycles, and the crack initiation occurs at the edge
region of the initial contact (Figure 5.61). At the early stage
of fatigue life, a crack was initiated on a plane inclined to
the surface and then the propagation direction changed to
approximately perpendicular. This type of crack path is
very common in conventional fretting fatigue (Mutoh,
1995).
The experimental results in Figure 5.59 show that fati-
gue failure can occur after more than 107 cycles, and even
over 108 cycles, which reveals that for the fatigue design,
the fatigue limit usually determined at 107 cycles is not
always appropriate. This phenomenon indicates again the
importance of a high frequency fatigue test technique.
Fretting not only accelerates crack initiation but, also
increases the rate of crack propagation. Hence, fretting fati-
gue life is likely to be determined by the continued propaga-
tion of small fretting cracks. But there exists a threshold of
stress intensity factor in fretting fatigue below which a fret-
ting crack does not propagate. In this case, fretting scars
are larger in the form of an ellipse. Considerable fretting wear
is encountered over the entire contact area, at the surface of
both the specimen and the pad (Figure 5.60). The contact sur-
face increases with the number of stress cycles. Red oxide deb-
ris is observed at the contact surface and the examination of
206 Chapter 5

Figure 5.61 Fretting fatigue crack initiation point and propaga-

tion path.

fretting scars demonstrates some fine, but non-propagating,

cracks at the surface.
In conclusion, an ultrasonic fretting fatigue test per-
formed on high strength steel demonstrates that there is a
significant reduction in fatigue strength caused by fretting.
The factor of reduction is of the order of three times and
increases with the number of cycles. Fretting fatigue failure
is a combination of fretting slip and fretting stress. In ultraso-
nic fretting fatigue, crack initiation occurs at the edge region
of the initial contact area, and is present over a greater frac-
tion of the whole fatigue life. There is a minimum fatigue
stress below which fretting cracks do not propagate. In this
case, fretting scars are larger and fretting wear is consider-
able. For more information about fretting fatigue of high
strength steel at ultrasonic frequency, see (Sun, 2001).
6
Frequency and Environmental
Effects

6.1. FREQUENCY EFFECT

The ultrasonic fatigue technique is an accelerated testing

method with a frequency far beyond that of conventional fati-
gue experiments and applications. A problem is naturally
posed: Is there an important frequency influence on the
experimental results? If the frequency effect is too great,
the usability of the test data might be in question. Fortu-
nately, results in the literature and our own experience indi-
cate that for ultrasonic fatigue tests under low displacement
amplitude and small deformation conditions, the frequency
effect is small in most cases. Quite a few similarities between
ultrasonic and conventional fatigue data have already been
noted in previous chapters. The frequency effects are
discussed in this chapter in detail. We will show that the fre-
quency effect is much more significant between 10 Hz and
0.001 Hz than between 10 Hz and 20 kHz.

207
208 Chapter 6

6.1.1. S-N Curve and Fatigue Limit

There have been numerous studies of different materials and
microstructures considering the frequency effect on S-N curve
and fatigue limit. Some materials respond differently to ultra-
sonic fatigue loading than others. For example, for the nickel-
based alloy Udimet 500, the ultrasonic fatigue data closely
match the conventional fatigue tests results between 105
and 107 cycles. For this material, the fatigue limit diminished
continuously beyond 107 to 109 cycles when tested at 20 kHz.
(see Figure 4.21). Some materials even present slightly better
fatigue resistance behavior in ultrasonic fatigue regime as
shown in Figures 4.19 and 4.20 for titanium alloy T6A4V at
room temperature. (Kuzmenko, 1984) found that, for a num-
ber of steel and other alloys [a carbon steel–steel 45 (0.5C-
0.8Mn-0.25Si), an aluminium alloy D16T (4.5Cu-1.5Mg-
0.6Mn), and the titanium alloys OT4-1 (Ti-2Al-1.5Mn) and
VT22 (Ti-7.5Mo-2.5Al-1.0Cr-1.0Fe), for example], fatigue
strength increases monotonically as the loading frequency is
raised from 10 Hz to 20 kHz. But for a chromium-nickel alloy
E1612 (35Ni-15Cr-3.0W-1.5Mn-1.2Ti) and a steel X18N10T
(10Ni-18Cr-1.5Mn-0.6Si-0.5Ti), no noticeable frequency effect
on the fatigue behaviors and on the S-N curves has been
found for 20 kHz and 20 Hz tests. (Weiss, 1982) claimed that
the loading frequency effect diminishes as the strength of
the material becomes greater, and concluded quite correctly
that insufficient cooling may be the reason for these results.
That means that the reported differences are not intrinsic
but essentially temperature effects. (Yeske, 1982) reported
experimental fatigue life data on a nickel based super-alloy
MAR-M-246, tested over a frequency range of 3.2 Hz to
23.2 kHz, and a cast copper-nickel alloy tested at 30 Hz and
13.6 kHz. No appreciable frequency effect on fatigue proper-
ties was noted for those two materials. (Tschegg and Stanzl,
1981) found that, in general, f.c.c. materials are only ‘‘mildly’’
frequency sensitive but b.c.c. materials are highly frequency
dependent. The explanations of this phenomenon recalled
the fact that the dislocation sources and the slip systems for
f.c.c. materials remain active up to ultrasonic frequency,
Frequency and Environmental Effects 209

while for b.c.c. materials, the frequency effect results from the
increase in material strength as the strain rate or loading
frequency increases. Applying this logic to the fatigue situ-
ation, higher frequency displacement controlled fatigue
test data should then tend to shift toward higher fatigue
strength and longer fatigue lifetime when plotted on an S-N
diagram.
From a comparison of the S-N curves of a high purity
copper in the high cycle regime, a reasonable similarity of
fatigue data between bending tests of normal frequency and
ultrasonic tests is found, but the axial load control fatigue
data give a lower fatigue strength.
To sum up the effect of high frequency on the S-N curve
it appears that the data obtained are only slightly or not at all
affected between 10 Hz and 30 kHz. Figure 6.1 gives the best
examples.
According to published investigations and our own obser-
vations (Marines, 2003), no appreciable difference could be
found in the fatigue crack growing mechanism between ultra-
sonic and conventional loading regimes. Nevertheless, the
matting effect and the area reduction effect are hardly impor-
tant in facets of the specimens damaged in ultrasonic fatigue
tests, even with a cyclic stress ratio of R ¼
1. Figures 6.2 (for
T6A4V) and 6.3 (for 17-4PH) make a comparison of the rup-
ture feature of specimens loaded in the two different regimes
(tests are performed at room temperature). This is because in
ultrasonic fatigue tests, the specimen vibrates at longitudinal
resonant frequency and the vibrational energy is transmitted
by the remaining ligament and reflected on the crack facets,
resulting in a smaller closure effect at the crack tip and a
smaller plastic zone dimension by comparison with that
obtained in conventional fatigue loading. Additionally, it is
shown that for most engineering materials, the crack grows
at a lower rate da=dN, corresponding to a lower stress inten-
sity factor DK applied when loaded at ultrasonic frequency in
resonant vibration.
It seems that at cryogenic temperature, the frequency
effect decreases. This is the case for material Ti-6A4V. At
77 K, the results of vibratory fatigue and conventional
210 Chapter 6

Figure 6.1 Comparison of S-N curves at low and high frequency.

Frequency and Environmental Effects 211

Figure 6.2 Fatigue crack growth with striation formation for

T6A4V (R ¼
1) at 20 kHz.

fatigue are quite similar as shown in Figure 6.4 (Jago,

1998).
Investigations in Stanzl’s laboratory on a steel alloy and
several aluminum and magnesium alloys at different testing
frequencies also confirm that a real intrinsic frequency effect
cannot be detected, as long as elasticity dominates the
mechanical fields of the tested materials (Stanzl, 1996).
It must be pointed out that the old high-frequency
results must be considered with caution because it is always
difficult to control a high frequency fatigue machine. The
reader must know that a piezo-electric machine, working at
20 kHz, cannot be controlled without a computer processor
type 486 or better. This means the older data before 1990
are suspected to have errors.
212 Chapter 6

Figure 6.3 Fatigue crack growth with striation formation at high

and low frequency for 17-4PH.

6.1.2. FCG Rate and Threshold

Since crack propagation involves cyclic deformation at the
crack tip, similar frequency effects may be expected for crack
propagation itself. Therefore, one might expect a weak depen-
dence or none at all due to strain rate in f.c.c. materials and a
rather complex response of b.c.c. materials, resulting in the
possibility of inter-granular or cleavage fracture (Laird, 1982).
According to (Stanzl, 1996), the following experimental
results available thus far are:
 No influence of the testing frequency is observed for
13% chromium steel in tests in inert environment at
testing frequencies between 0.1 Hz and 20 kHz
(Speidel, 1980).
Frequency and Environmental Effects 213

Figure 6.4 Comparison of fatigue strengths of T6A4V at 20 kHz

and 33 Hz in liquid nitrogen.

 No influence is also observed for grey cast irons over a

frequency range of 5 Hz to 21 kHz, as well as for mild
steel, tested at 100 Hz and 20 kHz in inert oil at 20 C
(Stanzl, 1986).
 No measurable influence of frequency on FCG rates
between 2.3 Hz and 20 kHz at room temperature is
reported for three nickel alloys (Hastelloy-X, RA-333,
and IN800-H) (Hoffelner, 1982). It should be men-
tioned, however, that the range of scatter of the ultra-
sonic results is up to approximately 30% in all cases,
mainly due to lack of precise amplitude control.
 More recent measurements (Hoffelner, 1982) on
Armco iron at frequencies of 100 Hz and 0.1 Hz at
room temperature in air confirm the frequency
independence of the threshold value.
However, pronounced frequency effects have been found
for a carbon steel by (Puskar, 1986) with threshold values
increased by a factor of 1.8 but the crack propagation rates
decreased by a factor of 16 at 22 kHz from 70 Hz, and by
(Kuzmenko, 1984) for a titanium alloy VT-1 and an alumi-
nium alloy AK4. One might suspect, however, that their DK
values have not been accurately determined.
Meanwhile, recent experimental data on FCG rates and
threshold at ultrasonic frequency in our laboratory have little
214 Chapter 6

scatter and show only some weak frequency effect. For nickel
based alloy N18, the influence of frequency on FCG curves at
high temperature is due undoubtedly to different environ-
mental influences at ultrasonic and conventional frequencies
(see Section 5.2). A small frequency effect on FCG rate of
another nickel based alloy (Inconel 706) may be caused by
the difference in the interaction time between-environment
and the metal at crack tip. That is to say, an intrinsic fre-
quency effect on FCG rate and threshold is not observed.

6.1.3. Dislocation Structures and Plateau Stress

In this section we mainly cite remarks made by (Stanzl, 1996)
concerning the frequency effect on dislocation structures and
plateau stress.
In a fundamental consideration, (Laird, 1982) compares
the strain rate dependence of cyclic deformation and the dis-
location movement in f.c.c. and b.c.c. materials. In f.c.c.
metals, where the cyclic flow stress is weakly dependent on
strain rate, little frequency effect on fracture mechanism is
expected. In b.c.c. materials that are sensitive to high strain
rate, strain aging, dislocation glide, and shape changes can
be affected as well as the failure mechanism, by favoring
inter-granular failure during crack initiation and cleavage
during crack propagation.
Several extensive studies have been performed during
the last ten years on the dislocation structures and the
stress-strain response of single crystals. TEM studies on the
dislocation structure of ultrasonically loaded Cu single crys-
tals showed no significant differences in the dislocation struc-
tures below as well as at plateau stress compared with those
established under conventional fatigue (Stanzl, 1986). The
plateau stress, where PSBs (persistent slip bands) are formed,
is found at 26 MPa at room temperature, which is close to the
value of 28 MPa obtained under plastic strain controlled test-
ing conditions at conventional frequencies. This coincidence is
remarkable, since ultrasonic load control was not as accurate
in 1984 as it is today (about 95% accuracy). In addition, the
ladder spacing of the PSBs is approximately the same as
Frequency and Environmental Effects 215

observed for conventional frequencies. In (Stanzl, 1996), a

replica of a PSB is observed, which clearly shows how a micro
crack is formed in a PSB. In a similar study (Tschegg, 1981),
at 77 K the samples of copper single crystals of 99.9999% pur-
ity show a plateau stress of 47  2 MPa at 20 kHz, while the
value for the same material tested at low frequency is 48 MPa.
Ultrasonic fatigue loading at 77 K (Buchinger, 1984)
likewise causes similar dislocation arrangements as with con-
ventional frequencies. As an example, (Stanzl, 1996) shows a
typical fully developed LP (loop patch) structure. Further-
more, the ultrasonic fatigue of a polycrystalline Cu-16% Al
alloy also generates a dislocation structure similar to the
one caused by conventional frequencies in the mono-
crystalline alloy (Laird, 1986). However, one does find in
(Laird, 1986) that, for high frequency testing and in dense dis-
location structures that are essentially dipolar, the dipole
width is considerably smaller than that observed in low
frequency testing.
(Mayer, 1994) presents an experimental study on poly-
crystalline copper at different frequencies between 0.5 Hz
and 8 Hz under load and strain control, and finds that under
load control the formation of PSBs depends on the loading
frequency, but that no such influence is present for strain con-
trolled loading. As ultrasonic fatigue loading is essentially
strain controlled, this result explains very well the above
mentioned frequency independent formation of dislocation
structures, and is another verification of the non-existence
of a real inherent frequency effect. Furthermore, a frequency
change from 0.5 Hz to 2 Hz generally produces more signifi-
cant effects than one from 2 Hz to 8 Hz. This observation is
helpful in explaining little frequency effect at high frequency
for most materials.
The above results agree with another basic study on the
deformation mechanism, i.e., the so-called Blaha effect (influ-
ence of superposition of cyclic ultrasonic loading on unidirec-
tional loading) (Kirchner, 1984). There is no frequency effect
in the load response of several aluminium alloys at room
temperature. For specimens tested at frequencies of 0.5 Hz,
10 Hz, 50 Hz, and 20 kHz, the stress drop is caused by elastic
216 Chapter 6

relaxation, not by high frequency. On the other hand, discus-

sion of this effect was initiated again by (Kobayashi, 1991)
who did find a frequency influence. It may be suspected that
this might be caused more by the experimental procedure
than by frequency effect.

6.1.4. Fretting Wear

(Soderberg, 1986) observed a similar effect of frequency in
the fretting of mild steel at ultrasonic frequencies and at low
frequencies, with respect to both wear rates and wear
mechanisms.

6.2. HEAT EFFECT

High temperature ultrasonic fatigue testing of materials that

are designed for high temperature fatigue applications has
been developed by our group in recent years, after several
works on this subject had been published in the past.
The testing system is presented in Chapter 3. The
machine is designed to have 20 kHz resonant frequency.
The displacement amplitude of the specimen extremity is
measured by means of a capacitive transducer, which per-
mits evaluating the stress intensity factor with aid of FEM.
A heating solenoidal inductor is intalled at the specimen cen-
ter. The maximum temperature is at the center section with
a crack and the minimum temperature is at the ends; this
gradient must be taken into account in numerical calcula-
tion. So, the resonance length L2 is less in elevated tempera-
ture. A system of video-camera television has been used for
the detection of crack initiation and propagation, the system
resolves events to 1=25th of a second and magnifies the pic-
ture 200 times to obtain the threshold at 10
12 m=cycle
during 1 hour.
The performing of comparative fatigue crack growth
measurements on the nickel alloy Astroloy at 20 C and
400 C in an induction furnace by (Wu, 1994) shows that the
fatigue crack growth resistance decreased by about 30% at
400 C (Figure 6.5).
Frequency and Environmental Effects 217

Figure 6.5 FCGR of Astroloy.

It is seen that the threshold is determined for a low rate:

10
12 m=cycle for ultrasonic fatigue and 10
10 m=cycle for
conventional fatigue with CT specimen. At 20 kHz, the
threshold is lower at 400 C than at room temperature.
(Hoffelner, 1982) believes that there is no frequency
influence between 20 kHz and conventional fatigue. In Figure
6.5, two curves at 20 Hz at room temperature for R ¼ 0.05 and
R ¼ 0.7 are found. The first can be considered as the Kmax
curve because of Kmin ¼ 0. The second can be considered as
the Keff curve because for R ¼ 0.7 there is no closure effect.
Between them is the ultrasonic curve at 20 C, which is near
the curve at R ¼ 0.7. In the study by Hoffelner (1982), we find
the same phenomenon: The threshold at 20 kHz is less than at
60 Hz.
Figure 6.6 shows FCGR as a function of DK for N18 at
different temperatures in vibratory fatigue. At 20 kHz, the
threshold values are also determined at very low FCGR,
and they are lower at elevated temperatures than at room
temperature. However at 400 C, the threshold is lower than
218 Chapter 6

Figure 6.6 FCGR of N18 at high temperature.

at 650 C and 700 C. It is well known that oxidation can

decrease crack propagation under certain conditions. So, the
threshold increase at 400 C can be explained by oxidation
for N18. In fractographical observation, oxidation phenom-
enon is found in the fractures at 650 C and 700 C. (Hudak,
1988) has studied Astroloy in conventional fatigue. He also
found that the threshold was lower at 200 C than at 600 C.
Obviously, oxidation has a similar effect at 20 kHz and low
frequency.
Heat effect relates closely to frequency effect, because
under high frequency with alternating loading the internal
friction, or damping, of materials results in an elevation of
the temperature in the specimen, most notably at the crack
tip. This, in turn, changes the mechanical properties of the
tested material. At 20 kHz for high damping materials, the
specific energy input (heat source due to thermodynamic cou-
pling) at high stress may reach as high as several hundred
watts per cubic centimeter. This is another aspect of heating
effects that must be considered in some cases. Generally,
suitable cooling measures are required. The coolant may be
air or water, depending on the damping of material and=or
Frequency and Environmental Effects 219

the geometry of the specimen. For example, a compressed air

cooling system can maintain the temperature at the center of
a T6A4V specimen at 35 C, which effectively eliminates the
temperature rise caused by the ultrasonic frequency effect.
On the other hand, if cooling is introduced, environmental
effects such as corrosion fatigue must be considered in some
cases. Figure 6.7 shows the temperature rise at the center
of a cast iron endurance specimen, tested at different stress
levels (Wang, 1998a). If there were no cooling due to the
compressed air, the temperature rise would be considerable
even for low stress levels. For this kind of high damping
material, a thin sheet specimen is required. On the other
hand, titanium alloys exhibit little heat sensitivity at
20 kHz and at modest stresses, so there is no restriction in
the specimen geometry within temperature rise limitations
for these alloys.
Another possibility to avoid the heating of the specimens
is to perform loading in a pulsed manner with periodic inter-
ruptions. According to (Stanzl, 1981), usually a pulse length
of 500 to 1000 cycles is suitable with pauses between 50 and
1000 ms.

Figure 6.7 Temperature rise with stress levels for cast iron.
220 Chapter 6

Figure 6.8 Variation of Young’s modulus with temperature.

For highly heat sensitive materials, one must take the

variation of Young’s modulus into account in order to obtain
a correct mechanical field. Figure 6.8 gives the decrease of
Young’s modulus of Astroloy with increasing temperature
(Wu, 1992).

6.3. CRYOGENIC TEMPERATURE

Ultrasonic fatigue tests have been performed on three differ-

ent thermomechanical processes (TP) of the Ti-6246 alloy
usually used in high temperature conditions. Two b-forged
titanium alloys (TP1: fine lamellar primary a structure;
TP2: coarse lamellar primary a structure) and one a þ b
forged processed (TP3: coarse equiaxed and fine lamellar
primary a) have been studied (Jago, 1995) to compare
fatigue-limit differences at 109 cycles. Results in liquid nitro-
gen are grouped in Figure 6.9. For all conditions, fatigue-limit
stress resistance is higher at low temperature than at room
temperature. There is a cross-over of TP1 and TP3 S-N
curves: At cryogenic temperature, TP1(1) fatigue-limit value
(sd) is higher than that for TP3; this behavior is reversed at
Frequency and Environmental Effects 221

Figure 6.9 S-N curves for TP conditions of Ti-6246 alloy at 300 K

and 77 K.

room temperature. In addition, TP1(1) curves are signifi-

cantly above those for TP1(2). For both test temperatures,
the TP2 condition has the lowest fatigue-limit strength prop-
erties at 109 cycles.
Observations on the effect of cryogenic temperature at
20 kHz follow. Figure 6.10 presents a comparison of S-N
curves of alloy Ti-6A4V tested in liquid hydrogen and liq-
uid helium with conditions of 20 K and 20 kHz. These

Figure 6.10 S-N curves of Ti-6A4VPQ at liquid hydrogen and

liquid helium (20 K; R ¼
1).
222 Chapter 6

experiments were performed as described in (Tao, 1996). The

difference in fatigue strengths for the two environments is of
the order of several tens MPa. It is believed that this differ-
ence is caused by the difference in real temperatures around
the specimens. The temperature of a specimen in liquid
hydrogen is about 20 K, but for the tests in liquid helium,
the situation is different. Considering the place of entry of
the liquid helium fluid, the specimen, and the temperature
sensor, we find there is a temperature gradient in the cryo-
stat (Figure 6.11). The real temperature of the gas of liquid
helium around the specimen is lower than 20 K (between
10 K and 20 K). This drop of local temperature increases
the fatigue life of the specimen.
Another interesting result is given in Figure 6.12 which
shows the S-N curves obtained in liquid hydrogen for alumi-
nium alloys 6061 and 2219, tested at 20 kHz. For those alloys
the fatigue strength is much higher at 20 K than at 300 K.

Figure 6.11 S-N temperature gradient in the cryostat.

Frequency and Environmental Effects 223

Figure 6.12 S-N curves for Al alloys in liquid hydrogen.

Thus, the effect of cryogenic temperature is the same for tita-

nium and aluminium alloys, which is considered an effect of
microstructure on fatigue strength in the gigacycle fatigue
regime. For the Ti-6246 alloys the fatigue strength at 109
cycles is more than 200 Mpa higher at 200 K, depending on
the microstructure (Jago, 1996; Tao, 1996).

6.4. ENVIRONMENTAL EFFECTS

As indicated (Stanzl, 1996), the question of the influence of

frequency on the results obtained in ultrasonic fatigue experi-
ments is especially relevant when environmental influences
(i.e., time dependent processes) become effective. After
numerous research in the past, corrosion fatigue at ultrasonic
frequencies was further investigated mainly by (Ebara, 1994,
1987), where the influence of aggressive environments, such
as aqueous NaCl solutions on stainless steel 13Cr and NaOH
solutions on alloy Ti6A4V, was found. The fatigue strength of
13Cr at 1010 cycles in an aerated 3% NaCl aqueous solution
224 Chapter 6

was 74% lower than that in the atmosphere. This reduction is

almost equal to that of 75% at 3.7  107 cycles in the result of
the conventional rotating bending fatigue test at 60 Hz. This
observation confirms again the suitability of the ultrasonic
corrosion fatigue for this material. Besides the expected
reduction of fatigue strength, the experiments with Ti6A4V
in an NaOH solution revealed a pronounced additional decay
of fatigue strength at around 1010 cycles. The degradation of
the specimen due to corrosion pits is assumed to be most
likely responsible for this phenomenon. Because of the very
high number of cycles, this can only be investigated by ultra-
sonic fatigue techniques.
Stanzl’s group has explored the influence of humid and
dry air, as well as vacuum on the FCG behavior of the alumi-
nium alloy 2024-T3 (Stanzl, 1991). In humid air, a plateau-
like crack growth regime (Figure 5.12a in Stanzl, 1991) at
crack growth rates between 10
10 and 10
9 m=cycle is
observed, which is attributed to hydrogen effects. The results
show unambiguously that the plateau in the curve of thresh-
old is mainly determined by the environment and not by
micro-structural features. The increase in crack growth rates
depends on the water vapor content in air. Dry air experi-
ments lead to crack growth rates between those obtained in
humid air and those in vacuum. The plateau-like regime
occurs at fatigue crack growth rates of about an order of mag-
nitude lower than that reported in the literature for a loading
frequency of 35 Hz. This difference may be explained by the
mechanism of surface diffusion of water vapor molecules to
the crack tip being the time governing process. With the help
of SEM studies of the fracture surfaces, relevant crack growth
mechanisms are discussed by (Stanzl, 1991).
Figure 6.13 shows another approach to the environmental
fatigue limit at 109 cycles and at ultrasonic frequency for 17-
4PH and Ti6A4V, and stainless steel 403 in an aggressive envir-
onment (Willertz, 1982). Comparison of high and low frequency
corrosion fatigue tests by (Willertz, 1982) shows that adjoining
and overlapping fatigue curves are produced for these materials
in both aggressive and non-aggressive environments over the
range of frequencies examined (40 Hz to 20 kHz).
Frequency and Environmental Effects 225

Figure 6.13 Ranking of fatigue strength of three engineering

alloys tested at ultrasonic and conventional frequencies (Willertz,
1982).

(Roth, 1985) indicates that the degradation in fatigue

properties due to environment–fatigue interaction is obvious
by comparing fatigue limits in pure water with those in any
other environment. For example, the 109 cycle fatigue limit
of stainless steel 403 in air-saturated 22% sodium chloride
solution is 1=8 of the fatigue limit in pure water. Varying
degrees of degradation of the 109 cycle fatigue limit are
observed for the other combinations of materials and environ-
ments (Figure 6.14). It is also remarked by (Roth, 1985) that
the fatigue limit at ultrasonic frequencies for some engineer-
ing alloys exhibits sensitivity to the imposition of electro-
chemical potential during testing similar to that observed at
lower frequencies.
(Whitlow, 1982) investigated the frequency effect on cor-
rosion fatigue of Udimet 720, a nickel based alloy, at high
temperature of 704 C and at the frequencies of 48 Hz and
20 kHz. The tests are carried out in both air and an aggressive
molten salt environment. The conventional and ultrasonic
frequency S-N data in each environment are shown to be
essentially the same. In air, it is suggested that the alloy
226 Chapter 6

Figure 6.14 Effect of cathodic polarization on ultrasonic and

conventional-frequency corrosion fatigue of AISI 403 stainless steel
in aqueous chloride solution (Roth, 1985).

behavior is independent of strain rate in the range studied.

The observed degradation in life for the specimens tested in
salt solution as compared to those in air suggests that fatigue
resistance is reduced on a per cycle basis regardless of the
cycle duration. Whitlow further concludes that the use of
ultrasonic fatigue to extend the S-N curve to the gigacycle
regime may produce reliable long-term data.
(Yeske, 1982) reported the ultrasonic corrosion fatigue
properties of some surgical implant materials (tantalum, nio-
bium, and stainless steel) in the environments of pure water
and aggressive modified Early solution. The results show that
at the test frequency of 20 kHz, the fatigue life of stainless
steel is significantly reduced by the corrosive environment
Frequency and Environmental Effects 227

while no such effects could be observed for tantalum and

niobium specimens.
It is not easy to develop theories of mechanism for the
environmental effects on fatigue at 20 kHz, because the cycle
period is extremely small and the test data and environments
vary considerably. One attempt was made in (Weiss, 1982)
where a slip dissolution model for the 20 kHz corrosion
fatigue behavior of stainless steel 403 in aqueous chloride
solution was proposed. In that study, the authors indicate
similarity of the corrosion fatigue characteristics observed
for frequencies of 40 Hz and 20 kHz. The S-N data, while
not representing a complete overlapping of results from the
two frequencies, do in fact permit the data from both frequen-
cies to be comfortably attributed to a single S-N curve. The
morphology of the pit-initiated cracking appears to be the
same for tests at both frequencies, despite the wide
difference in the exposure times for tests with equivalent
accumulated stress cycles. The responses to changes in the
electrochemical conditions of the test environment were simi-
lar for the two frequencies investigated. However, in our
opinion, these apparent similarities can hardly be extended
to a general conclusion because of the different corrosion
fatigue and cracking mechanisms of different materials. Con-
sequently it is necessary to emphasize that developing a more
general model for the environmental effects on ultrasonic
fatigue is difficult. For example, experience indicates that
for many materials the site of crack initiation in the
gigacycle regime is internal rather than at the surface of
the specimen.

6.5. S-N CURVE AT ROOM TEMPERATURE AND

HIGH PRESSURE HYDROGEN FOR Ti-6A4V

Some high frequency fatigue tests have been carried out in

Industeel Industry Laboratory of Le Creusot with this
piezo-electric system for Ti-6A4V alloy at 300 bars hydrogen
pressure and room temperature for R ¼
1. At the same time,
other tests have been conducted for the same material and
228 Chapter 6

Figure 6.15 Ti-6A4V life curve at 20 kHz at 300 bars H2 and

atmospheric pressure for R ¼
1 20 kHz.

same conditions, at room temperature and air at atmospheric

pressure. The results are shown in Figure 6.15 together with
the results of fatigue performed at 20 kHz on the same batch
of materials by SNECMA.
The objective of this high frequency test is to reproduce
the same condition of environment as in a hydrogen turbine,
i.e, with the same fatigue stress level and exposure duration
at 300 bars pressure. In fact, the duration of exposure to
hydrogen, the temperature, the pressure, and the stress level
are also important parameters. Hydrogen diffusion is a func-
tion of the metallurgical structure and texture and therefore
has some effect.
7
Microstructural Aspects and
Damage to Materials in the
Gigacycle Regime

At the microscopic level, it seems that the initiation of a fati-

gue crack in gigacycle fatigue can be generally described in
terms of a local microstucturally irreversible portion of the
cumulative plastic cycle strain. In one way, there is no differ-
ence between fatigue mechanisms in mega- and gigacycle
regimes except for the surface strain localization in persistent
slip bands in the megacycle case. However, many other speci-
fic mechanisms can occur especially at high cycles internally
in the gigacycle fatigue regime.

7.1. GIGACYCLE S-N CURVE SHAPE

Generally, it is assumed that S-N curves for steel are different

from those for other materials. In order to obtain in a wide
overview of the gigacycle behavior, many alloys including
steels are considered here:
229
230 Chapter 7

Steels: Low carbon steel, 4240, spring steel, bearing

steels, 17-4 PH, rail steel, stainless steel 304

Sphero€
dal graphite cast iron

Ni based alloys: 718 and N18

Ti-alloys: Ti-6A4V, Ti-6246

Aluminium alloys: Al-Si and 2024

Magnesium alloy: AZ91
For fatigue S-N curves approaching 109 cycles, limited results
are available in the literature, with more than one-half of the
results from our laboratory. The other results are from Japa-
nese researchers such as (Nishijima, 1999), (Murakami,
2002), and (Sakai, 1999), and are often limited to 108 cycles.
Also, four S-N curves for light alloys are from the laboratory
of S. Stanzl-Tschegg and H.R. Mayer (Stanzl, 2001); these go

Table 7.1. Chemical Compositions of the Alloys Tested to

Gigacycle Fatigue
Alloy C Mn P S Si Al Ni Cr Cu Mo

4240U 0.428 0.827 0.012 0.024 0.254 0.023 0.173 1.026 0.210 0.224
4240R 0.412 0.836 0.015 0.087 0.242 0.023 0.186 1.032 0.209 0.164
54SC6 0.535 0.629 0.006 0.016 1.400 0.056 0.635
12Cr 0.11 0.77 0.011 0.002 0.064 2.53 11.4 1.47
17–4PH 0.07 1.0 0.04 0.03 5 17.5 0.5

Table 7.2 Mechanical Properties of the Alloys

Alloys E (GPa) r (kg=m3) UTS (MPa) Hv30

4240UrepB 200 7820 1100 345

4240RrepB1 200 7820 1040 320
4240UrepC 205 7870 1530 465
4240RrepC1 205 7870 1485 450
54SC6 210 7850 1692 510
12Cr 216 7760 1026 360
304 185 7978 599
100C6 210 7860 2316 778
Low Carbon Steel 203 7830 340 95
SG Cast Iron 178 7100 510 184
17-4PH 203 7800 1420
Microstructural Aspects and Damage 231

to 109 cycles. When available, the chemical composition and

the mechanical properties of alloys are given (Tables 7.1
and 7.2) order to document the results.

7.2. MECHANICAL ASPECTS OF INITIATION

BETWEEN 106 AND 109 CYCLES

Safe-life design based on infinite-life criteria was initially

developed from the Wöehler approach, which is the stress-life
or S-N curve related to the asymptotic behavior of steels.
Some materials display a fatigue limit (or ‘‘endurance’’ limit)
at a high number of cycles (typically >106). Most other
materials do not exhibit this response, instead displaying a
continuously decreasing stress-life response, even at a great
number of cycles (106–109), which is more correctly described
by a fatigue strength at a given number of cycles. Several
examples are given in the following discussion.

7.2.1. Gigacycle Fatigue of Steels and Cast Iron

Figures 7.1 to 7.10 present a large number of S-N curves for
steels and iron loaded in tension-compression or in tension-
tension where crack initiation can appear near and some-
times beyond 109 cycles.
It is noted that the shape of the S-N curve is not
the same from one steel to another. It is difficult to predict
the shape of an S-N curve between 106 and 109 cycles—
sometimes, the S-N curve is quite flat but the slope can be
steep for some metals. However, it is clear that for a first
class of alloys, the difference between the fatigue strength
at 106 cycles and 109, denoted by DsD, is only few mega-
pascals; that is to say, less than 50 MPa. Low carbon steel
(Figure 7.2), stainless steel 304 (Figure 7.3), 12Cr steel
(Figure 7.4), and also sphero€
dal graphite cast iron (Figure
7.5) present such behavior.
In contrast, there is a second class of steels for which the
difference between the fatigue strength DsD at 106 and 109
cycles ranges from 50 to 200 MPa. That means a higher slope
of the S-N curve in the gigacycle regime. This domain includes
Figure 7.1 Low carbon steel sheet S-N curve R ¼ 1 at 20 kHz
(difference between fatigue strength at 106 and 109, DsD ¼ 3 MPa).

Figure 7.2 Low carbon steel sheet S-N curve R ¼ 0.1; 20 kHz;
DsD ¼ 20 MPa.
Microstructural Aspects and Damage 233

Figure 7.3 Stainless steel 304 S-N curve R ¼ 1.

Figure 7.4 S-N data for 12% Cr steel R ¼ 1; 20 kHz; DsD ¼ 40 MPa.
234 Chapter 7

Figure 7.5 Spheroidal graphite cast iron S-N curve R ¼ 1 (up)

and R ¼ 0.1 (down); DsD ¼ 10 and 30 MPa.

4240 steel (Figure 7.6), bearing steel (Figure 7.7), rail steel
(Figure 7.8), spring steels (Figure 7.9), and martensitic stain-
less steels such as 17–4 PH (Figure 7.10). For technical appli-
cations a gap of 100 or 200 MPa on the fatigue strength cannot
be ignored. It seems that the higher the UTS of steels, the
Microstructural Aspects and Damage 235

Figure 7.6 High strength steel 4240 S-N curves R ¼ 1; 20 kHz;
DsD ¼ 30 and 80 MPa.
236 Chapter 7

Figure 7.7 Fatigue curve for 100C6 steel R ¼ 1; 20 kHz;

DsD¼ 180 MPa.

Figure 7.8 Rail steel S-N curve R ¼ 1; 20 kHz; DsD ¼ 100 MPa.
Microstructural Aspects and Damage 237

Figure 7.9 Spring steel S-N curves (55SiCr7; 54SiCr6) R ¼ 1;

20 kHz; DsD ¼ 200 and 150 MPa.

higher the S-N curve slope in the gigacycle fatigue regime.

Several mechanisms are involved in explaining crack initia-
tion in gigacycle fatigue as will be discussed in this chapter.
It can be pointed out that a progressive transition of mechan-
isms around 107–108 cycles induces a step in the S-N curve as
noted in (Murakami, 2002), (Nishijima, 1999), and (Sakai,
1999), and is shown in Figure 7.6 for 4240 steel. In this case
Figure 7.10 Martensitic stainless steel S-N curves (17–4 PH and
13–8 Mo) R ¼ 1 20 kHz; DsD ¼ 200 MPa.

Figure 7.11 N18 nickel base alloy S-N curves at 450 C between
106 to 1010 cycles.
Microstructural Aspects and Damage 239

the plateau is due to the large scatter of the results, or perhaps

due to rotating bending loading.

7.2.2. Gigacycle Fatigue of Ni Base Alloys

Several Ni base alloys were tested in the gigacycle regime (Ni,
1991) such as Udimet 500, Inco 718, and N18. For all these
alloys, the fatigue strength decreases by 150 to 200 MPa from
106 to 109cycles. Figure 7.11 gives our results for N18 alloy
tested at 450 C at R ¼ 0 and R ¼ 0.8. It is of interest to note
that, when the R ratio increases, the slope of the S-N curve
decreases. As expected, a high density of inclusions and pores
induced a large scatter in the number of cycles for initiation.
However, the fatigue strength at 109 cycles is only slightly
affected by the density of inclusions, since the controlling
feature is the size of the largest inclusion.

7.2.3. Gigacycle Fatigue of Titanium Alloys

Titanium alloys behave in the gigacycle fatigue regime in a
manner similar to Ni based alloys. For example, Figure 7.12
presents an S-N curve determined up to 109 cycles, at

Figure 7.12 Titanium alloy S-N curves between 106 and 109
cycles at Ti-6246 alloy. See Table 7.5 for thermal processing.
240 Chapter 7

20 kHz, 300 K, and R ¼ 1 for a Ti-6246 titanium alloy forged

with different heat treatments (Jago, 1998). Several features
should be pointed out:

Some fatigue initiations near 109 cycles

There is no asymptote

The fatigue strength sD at 109 cycles is much smaller
(by as much as 150 MPa) than the fatigue strength at
106 cycles

The forging process and the microstructure have a
marked influence on the high cycle fatigue life (sD
ranges from 325 to 490 MPa for 109 cycles)
Not having any inclusions or pores in this alloy enhances
the influence of processing and metallurgical transformation
in the Ti-6246 alloy. The worst fatigue strength at 109 cycles
for 6246 material is exhibited for heat treatments 1 and 2,
which give a larger volume fraction of secondary a and small
platelets of primary a. However, for heat treatment 3, the fati-
gue strength at 109 cycles reaches 490 MPa rather than only
325 MPa, and at 106 cycles it rises to 600 MPa. This means
the study of gigacycle fatigue is an important way to charac-
terize the alloys that are used for high technology applica-
tions such as jet engine turbine disks.

7.2.4. Gigacycle Fatigue of Aluminium Alloys

In this book, several gigacycle S-N curves for cast and
wrought aluminium alloys (2024, 2219, 6061, and A57) were
presented. These results come from Stanzl’s group (Stanzl,
2001), (Mayer, 1999), or the Bathias group (Bathias, 1998,
1999, 2001). For wrought aluminium alloys, there exists a dif-
ference of 100 MPa between the fatigue strengths, sD at 106
cycles and at 108 cycles. It is notable that for cast
aluminiums this difference is smaller and less than
100 MPa (Figure 7.13). The relative fatigue strength in the
gigacycle regime decreases 30% compared to the standard
fatigue limit at 107. Once again, it is shown in the Figure
7.13 that the effect of frequency is very small. It is pointed
out that one fatigue failure of AS5U3G is at 51010 cycles!
Microstructural Aspects and Damage 241

Figure 7.13 Aluminium alloy AS5U3G S-N curve R ¼ 1;

DsD ¼ 30 MPa.

In the gigacycle regime, the initiation site is located

inside the specimens for titanium, nickel alloys, and high
strength steels. However, fatigue initiation appears at the
surface of specimens in aluminium and magnesium alloys,
especially in castings where voids at the surface are the main
locations for crack formation.

7.3. INITIATION ZONE FOR LOW CYCLE TO

GIGACYCLE FAILURES

According to our own observations and those of the literature,

crack initiation in gigacycle fatigue seems usually to occur
inside the sample and not at the surface, if there are
inclusions or porosity inside the metal.
Three types of crack initiation occur in cylindrical samples
with a polished surface depending on whether it is low cycle
(104 cycles), megacyclic (106 cycles), or gigacyclic (109 cycles)
fatigue (Figure 7.14). For the smallest number of cycles to
242 Chapter 7

Figure 7.14 Mechanisms of fatigue initiation.

rupture, the initiation sites are multiple and on the surface,

according to the normal observation. However, at 106 cycles,
there is only one surface initiation site and, for a much higher
number of cycles to rupture, the initiation is located in an inter-
nal zone. What remains is to understand how and why some
fatigue cracks can initiate inside the metal in gigacycle fatigue.
Generally, but not always, the crack initiates from a defect,
inclusion, or pore. Sometimes the initiation is related to micro-
structure anomalies, for example, long platelletes or perlite
colonies.

7.4. INITIATION MECHANISMS AT 109 CYCLES

The full explanation of the gigacycle initiation phenomenon is

not clear. It seems that the triggering cyclic plastic deforma-
tion becomes very small in the gigacycle regime, in which
case, internal defects or large grain size can play a role creat-
ing plasticity, in competition with the surface damage. As the
initiation requirements for cyclic strain and maximum strain
to decrease, three main factors operate:
1. Anisotropy of metals: In the gigacycle regime, the
plastic strain is very small. Thus, the plasticity
appears only if the grain orientation and the grain
size are in agreement with dislocations sliding at
the surface or in the bulk of metals.
2. Stress concentration: It is suspected that stress
concentration due to metallurgical microstructure
Microstructural Aspects and Damage 243

misfit becomes an important factor when the applied

load is low. Inclusions, porosities, and grain size
effect are among efficient concentrators.
3. Statistical conditions: Statistically, the probability of
finding a sufficient stress concentration is more
likely in the bulk than in the surface of the metals
under high cycle low plastic strain requirement
conditions.
Thus, the probability of an offending stress concentrator
in the bulk of the metal is the best explanation for the loca-
lization of the initiation of the crack in the gigacycle
fatigue.

7.5. ROLE OF INCLUSIONS

In steel and nickel base alloys, inclusions can be significant

crack initiation sites especially if load ratio R is high. An
example of N18 alloy at 450 C is discussed in Section 7.5.2.
It is of great importance to understand and predict a fati-
gue life in terms of crack initiation and small crack propaga-
tion. It has been generally accepted that at high stress levels,
fatigue life is determined primarily by crack growth, while at
low stress levels, the life span is mainly consumed by the pro-
cess of crack initiation. Several authors demonstrated that
the portion of life attributed to crack nucleation is over 90%
in the high cycle regime (106 to 107 cycles) for steel, alumi-
nium, titanium, and nickel alloys. In the case when the crack
nucleates from a defect, such as an inclusion or pore, it is said
that a probable relation may exist between the fatigue limit
and the crack growth threshold.
However, the relation between crack growth and initia-
tion is not obvious for many reasons. First of all, it is not sure
that a fatigue crack grows at the very first cycle from a sharp
defect. Secondly, when a defect is small, a short crack does not
grow in the same manner as a long one. Thus, the relation
between DKth and sD is still to be studied.
Another important aspect is the concept of infinite
fatigue life. It is understood that below DKth and below sD
244 Chapter 7

the fatigue life is infinite. In fact, the fatigue limit sD is

usually determined for Nf ¼ 107 cycles. But fatigue failure
can occur up to 109 cycles and perhaps beyond 109. The fati-
gue strength difference at 107 and 109 cycles can be more
than 100 MPa. This means the relation sD versus DKth must
be established in the gigacycle regime if any relation exists.

7.5.1. Estimation of Crack Growth Life from

the da=dN Curve
To predict the number of cycles to initiate a fatigue crack from
an inclusion, several models are used with more or less suc-
cess; the integration of the Paris law, Murakami formula,
Tanaka model, and Kitagawa diagram are among the well
known approaches on the megacycle regime. In gigacycle
fatigue regime, the geometry of the fish eye initiation is a
circle that collapses at the surface of the specimen.
As discussed in the preceding section, the initiation of
crack growth consumes a dominant portion of the life in the
gigacycle fatigue range. The initiation from an inclusion or
other defect itself must be close to the total life, perhaps much
more than 99% of the life in many cases. This is made evident
by integrating the fatigue crack growth rates for small cracks
to estimate the possible extent of crack growth life.
In order to do this, one should refer to the general beha-
vior pattern of the crack growth rate curve as illustrated by
the equations on page 143. It is noted that small cracks such
as those growing from small inclusions do not exhibit crack
closure, therefore, in terms of DKeff, these equations apply
fairly well. They form the upper boundary on crack growth
rates for small cracks in ‘‘fish eye’’ range for which crack
closure is minimal.
Estimating the life of a crack of this type beginning just
above threshold, it is then appropriate to consider the growth
law as
 
da DKeff 3
¼b pffiffiffi
dN E b
Microstructural Aspects and Damage 245

where, for the circular crack growing in a ‘‘fish eye,’’ the stress
intensity factor formula is

2 pffiffiffiffiffiffi
DK ¼ Ds pa
p

The integration to determine the crack growth life begins

here, with the crack growth rate corner (as indicated by the
star on Figure 5.14), which we denote as DK0 corresponding
to an initial circular crack of radius a0. Substituting these into
the first formula we obtain

     3=2
da DK0 3 a 3=2 a
¼ b pffiffiffi ¼b
dN E b a0 a0

This equation can be integrated from a0 to afinal, which

gives
Z " #
af
ða0 Þ3=2 da 2ða0 Þ3=2 1 2a0
Np ¼
3=2
¼ 1=2
 small ffi
a0 b ðaÞ b ða0 Þ b

but it can be noted that

pffiffiffiffiffi
DK0 2Ds a0 pE2 b
1 ¼ pffiffiffi ¼ pffiffiffi pffiffiffi or a0 ¼
E b pE b 4ðDsÞ2

Then combining the last two expressions

pE2
Np ¼
2ðDsÞ2
This is the result for the approximate number of cycles
from threshold corner to failure from an initial crack size a0
as expressed above.
Now, if the example of 4240 steel (high strength) is taken
with a modulus, E, of about 215 GPa and a gigacycle fatigue
strength Ds of 500 to 750 MPa (from Table 4.3 and Figures
4.2 and 4.3), the result for the number of cycles, Np, of crack
246 Chapter 7

growth is about 150,000 to 290,000 cycles for starting crack

sizes, a0, of 37 to 72 mm. These numbers of cycles are not any-
where near the 109 cycles for these fatigue strengths for these
initial crack sizes close to the ‘‘fish eyes’’ observed.
Further, attempting to include integration of crack
growth rates below the threshold corner for these very small
initial cracks leads to smaller numbers of cycles added to the
numbers above.
Such calculations can be done in a manner similar to that
given above. Consequently, there seems to be strong evidence
that crack initiation is a dominant feature in gigacycle fatigue
life where ‘‘fish eyes’’ are observed.

7.5.2. Prediction of the Fatigue Limit at 109

Cycles Using the Murakami Formula
Another approach to predict the number of cycles at initiation
from a defect is to understand the notch effect phenomenon on
the fatigue strength. This approach is well known for the pre-
diction of the megacycle fatigue of steels where an empirical
relation is found between the ultimate strength (or hardness)
and the standard fatigue limit. In this respect, (Murakami,
2002) has proposed a parametric formula to predict the
fatigue limit of steel at 107cyles, depending on hardness
and defect size. Murakami has confirmed that the fatigue
limit concept of materials containing defects is essentially a
crack problem. In his approach the size of defect is given by
the expression (area)1=2, that is to say, the square root of
the initial defect projected area. From a great number of
experiments, Murakami and co-workers have determined a
strong correlation between the apparent fatigue threshold
at the tip of the defect and the size of the defect measured
as indicated above. They found the following relationship

DKth ¼ 3:3  103 ðHv þ 120Þ ðareaÞ1=6

According to Murakami, it seems the fatigue strength

at 109 cycles can be predicted using this formula with few
exceptions.
Microstructural Aspects and Damage 247

From our present data, we have verified this relation:

" #a
CðHv þ 120Þ ð1  RÞ3
sw ¼ pffiffiffiffiffiffiffiffiffiffi 1=6
ð area Þ 2
where
sw ¼ fatigue strength at 109 cycles MPa
C ¼ 1.78 for internal defects
Hv ¼ Vickers hardness
pffiffiffiffiffiffiffiffiffiffi
area ¼ projected defect surface (mm)
R ¼ load ratio
a ¼ 0.878 þ Hv104
An example of the application of the Murakami formula
in gigacycle fatigue of N18 nickel alloy is shown in Table 7.3.
(Bonis, 1997). The prediction of the model is always better
than 10% except in one case. It is of interest to note that inclu-
sions are mixed, sometimes, with porosities in this powder
metallurgy alloy.
The role of the inclusion is sometime hidden by the role of
porosities when the R load ratio is equal to –1 or to 0. Conver-
sely, when the mean stress of the fatigue cycle is very high,
for R ¼ 0.8, the role of inclusions becomes dominant. For all
R ratios, the Murakami formula gives a reasonable prediction
of gigacycle fatigue strength.
In order to adapt the validity of the Murakami approach
for different fatigue life Nf from 106 to 109, we have
introduced a correction factor b depending on Nf, as shown in
Figure 7.15 where the formula is applied for high strength

Table 7.3 Application of Murakami Model to N18 Nickel Alloy

R 1 1 0 0 0.8 0.8

Defect Mixed Inclusion Porosity Inclusion Porosity Inclusion

Localization
pffiffiffiffiffiffiffiffiffiffi Surface Internal Internal Internal Internal Internal
area (mm) 50 100 25 100 25 100
sw(MPa) 524 466 309 246 160 126
s experimental
(MPa) 525 400 280 270 160 130
Error % 0 þ14 þ10 9 0 3
248 Chapter 7

Figure 7.15 Murakami model for different fatigue lifes.

steels (Wang, 1998).

 
bðHv þ 120Þ ð1  RÞ a
sw ¼ pffiffiffiffiffiffiffiffiffiffi 1=6
ð areaÞ 2

7.5.3. Kitagawa Diagram at 109 Cycles

It is well known that the effect of inclusions is related to the
tensile strength of the metals in the megacycle fatigue regime.
According to the Kitagawa diagram (Kitagawa, 1976), the effect
of inclusions in gigacycle fatigue depends also of the UTS of
metals, especially for steels. From our own results, a Kitagawa
diagram is drawn for the gigacycle regime (Figure 7.16).
It is found that if the UTS of steels is high (1500 MPa and
greater) a small inclusion, less than 10 m, can be the initiation
site. Conversely, if the UTS of steels is low (500 MPa and less),
an inclusion of 100 m or more is need to initiate a fatigue crack
in the gigacycle cycle regime. This means that in low carbon
steel, the initiation is related to the grain size itself. In this
case, roughly speaking, sD =UTS ¼ 0:5. For high strength
steels, this ratio is less than 0.5 at 109 cycles.

7.6. GIGACYCLE FATIGUE OF ALLOYS

WITHOUT INCLUSIONS

What happens in alloys without inclusions in the gigacycle

fatigue regime? It is well known that in titanium alloys there
Microstructural Aspects and Damage 249

Figure 7.16 Relationship between the fatigue strength and

defects size (Kitagawa diagram).

is not any inclusion or porosity. Therefore, to answer this

question, titanium alloys were tested for initiation and propa-
gation. Since fatigue cracks cannot nucleate from internal
geometrical defects, the fatigue cracks must initiate from
metallurgical discontinuities.

7.6.1. Thermomechanical Processing of

Ti-6246 Alloys
The chemical composition of a Ti-6246 alloy, as supplied by
the RMI Company, is shown in Table 7.4. Four thermomecha-
nical processes (TP)—denoted TP1(1) TP1(2) TP(2) and
TP(3)—were used to produce forgings with different micro-
structures and attendant mechanical properties (Table 7.5).
The b-processed microstructures present similar lamella a-
phase morphology with different primary a volume fraction
and grain size in a transformed b matrix. The a þ b process
is characteristic of a bi-modal structure with duplex lamella
and globular primary alpha phase.
Quantification of the morphological aspects has been
performed to provide a complete description of various
microstructures.
 250

Table 7.4 Chemical Compositions of Ti-6246 Alloy Investigated (in wt%)

TP number Al Sn Zr Mo C (ppm) Cu (ppm) Si (ppm) Fe (ppm) O2 (ppm) H2 (ppm) N2 (ppm)

1(1) 5.76 1.97 4.08 3.97 90 <50 <50 400 930 44 80

1(2), 2 and 3 5.68 1.96 4.08 3.92 83 <50 <50 300 1100 28 70

Table 7.5 Thermomechanical Processes and Microstructures of Ti-6246

TP number Forging condition Thermal treatment Final microstructure type

935 C=2H=WQþ905 C=1H=

1(1) and 1(2) 955 C= WQ (Tb þ 10 C) Airþ595 C=8H=Air a platelets and b-transformed matrix
935 C=2H=Slow cool: Room T. Coarse a platelets and
2 955 C=WQ (Tb þ 10 C) þ595 C=8H=Air b-transformed matrix

935 C=2H=WQþ905 C=1H= Bi-modal structure (a platelets and
3 905 C=Air (Tb  40 C) Airþ595 C=8H=Air a nodular) and b-transformed matrix
Chapter 7
Microstructural Aspects and Damage 251

7.6.2. Fatigue Initiation in Ti-6246 Alloys

With this alloy and various thermomechanical processings it
is found that crack initiation and failure can occur up to 109
cycles though there is not any inclusion or pore. Figure 7.12
presents S-N curves showing dependency on the thermal
processing. At room temperature a significant differ ence
can be observed in S-N curves between the different TP con-
ditions. Thus, the TP3 material has comparatively the
higher fatigue resistance (510 MPa); the TP1(1) and TP1(2)
materials exhibit a lower response with a fatigue strength
for 109 cycles estimated respectively at 490 MPa and
400 MPa; and the TP2 material has the lowest fatigue-limit
resistance of only 325 MPa. In Figure 7.17, we notice that

Figure 7.17 Low fatigue crack growth rate of Ti-6246 alloy, for
different TP, at 20 kHz.
252 Chapter 7

the TP3 alloy has the lowest DK threshold with the best fati-
gue strength.
The SEM fractographic observations indicate that all
the TP1(1) broken samples have systematic near-surface
initiation (less than 40 mm from the external surface),
whereas TP1(2), TP2, and TP3 have systematic internal fati-
gue crack sites. In TP2 conditions, microstructure and more
specifically colonies of primary alpha phase a(P) show on
the fracture surface by backscattered electrons observations
and form a sort of facet (Figure 7.21). It can be seen that
the facets are oriented to the fracture plane, a feature com-
mon to all specimens.
In conclusion, it is emphasized that gigacycle fatigue
regime is not always correlated to defects such as inclusions
or pores. For Ti-6246, the gigacycle fatigue strength is
associated with transformed amount and secondary alpha
volume fraction. Internal fatigue initiation with quasi-
cleavage facets in primary alpha phase has been demon-
strated.
Under these conditions, it is very difficult to get a general
relation between DKth and DsD. A nucleation process must
exist. However, a linear relation is found between yield stress
and sD in the gigacycle regime for 6246 titanium alloy.

sD ¼ 1184 þ 1:6 sY

7.7. GENERAL DISCUSSION OF THE

GIGACYCLE FATIGUE MECHANISMS

With respect to the competition between surface and internal

initiations sites, gigacycle fatigue initiation mechanisms are
split into three cases, as follows.

7.7.1. A Critical Defect Exists at the Interior

of the Alloys
The gigacycle initiation mechanisms are strongly depen-
dent on the stress concentration due to inhomogeneous
Microstructural Aspects and Damage 253

Figure 7.18 Surface-subsurface transition in crack initiation

location in 4240 steel.

microstructure. They can be:

Mineral inclusions in high strength steels, martensitic
stainless steels, bearing steels, and spring steel.

Perlite nodules in rail steels

Platelets in titanium alloys

Carbides in cast irons
254 Chapter 7

Figure 7.19 Internal initiation in gigacycle fatigue regime of 17–4

steel. Initiation is related to a mineral inclusion.

Figure 7.20 Internal initiation in rail steel in gigacycle regime.

Initiation is related to a nodule of fine lamellar perlite shown on
the right side.

For all these alloys it was found that the initiation site
for fatigue cracks shifts from the external surface to subsur-
face at a particular lower stress range. The subsurface crack
initiation is dominant in the gigacycle range. A surface–
subsurface transition in crack initiation location has been
established at approximately 107 cycles. Sometimes there is
a plateau between 106 and 108 cycles on the S-N curve that
is not fully understood. The effect of microstructural details
Microstructural Aspects and Damage 255

Figure 7.21 Fatigue crack initiation on a primary phase in

Ti-6246 alloy.

is apparently very important in the gigacycle fatigue regime

(Figures 7.18 to 7.20).

7.7.2. Surface Initiation in Gigacycle Fatigue

Another special case is the gigacycle fatigue of Al-Si cast-
aluminium alloys for which the initiation location is at the
surface (Figure 7.22). It could be related to the high density
of pores always present on the specimen surface. In this case,
the S-N curve between 106 and 109 cycles is decreasing mono-
tonically without transition at 107 cycles (Figure 7.13).

Figure 7.22 Initiation of a fatigue crack under the surface in cast

aluminium alloy.
256 Chapter 7

Figure 7.23 Internal initiation in N18 alloy related with both

pores and inclusions.

7.7.3. Multiple Mechanisms

It seems that several mechanisms can operate in gigacycle
fatigue. This is the case of powder metallurgy alloys where
inclusions and pores are co-existing (N18 alloy). When inclu-
sions are present, the initiation site is in the interior. When
there are pores, the initiation is in the surface or in the inter-
ior (Figure 7.23).
For N18, the slope of the S-N curve is uniform between
10 and 1010 cycles but there is a transition in initiation sites
6

around 107 cycles, after which a competition appears between

the surface or interior initiation, depending on the load ratio
R. Inclusions are more dominant for high R tending to induce
initiation in the interior.

7.7.4. Conclusion
It has been shown that, for a large number of alloys, fatigue
crack initiation occurs beyond 107 cycles and that the differ-
ence in fatigue strength between 106 and 109 cycles often
decreases by 50 to 200 MPa. Obviously, the concept of infinite
fatigue life on an asymptotic S-N curve is not correct.
Very often there is a surface-to-subsurface initiation site
transition around 107 cycles when inclusions or microstruc-
Microstructural Aspects and Damage 257

tural defects are present. This mechanism is not unique since,

in the case of no critical damage in the interior, the initiation
will appear at the surface.
Appendix 1
Stress Calibration

For the ultrasonic fatigue machines to work correctly, stress

calibration is necessary. The first objective of the calibration
is to make the test system vibrate in resonance at ultrasonic
frequency of about 20 kHz. Among the important factors of
concern is the variation of Young’s modulus of the material
due to the high frequency or non-uniform temperatures.
Although some measures for calibration have been mentioned
in previous sections, it is necessary to present, in this appen-
dix, the detailed calibration procedures for the low tempera-
ture test system described earlier. The principles for the
calibration of machines operated at room temperature are
the same; those procedures will be simpler.
The mechanical system works in an elastic regime. The
relationships between displacement, strain, and stress are
therefore linear. The electrical voltage applied to the piezo-
ceramic is also linear and proportional to the displacement.
The electrical current density depends on the impedance,
the dynamic mechanical resistance of the equipment fixed in
the converter. For these reasons, we use a J2 type generator

259
260 Appendix 1

connector that permits us to know the displacement ampli-

tude and the output power.
The measures required at low temperatures are more
involved. One solution may be to use an accelerometer
installed at the end of the converter and to calibrate its signal
with an optical displacement sensor. Before using this signal,
we must convert the sinusoidal wave to a DC voltage. How-
ever, since we have the driving voltage to the piezo-ceramic
that is already a signal proportional to the displacement
amplitude, utilization of this signal will be a simpler and more
practical solution. This solution is better than the introduc-
tion of an accelerometer, especially since the signal is proper
and the additional installation of an apparatus (the acceler-
ometer) will be avoided. Furthermore, for the accelerometer,
the linear relation between this signal and the displacement
in the specimen’s head, is unknown because of the electrical
chain. Therefore, calibration is required.

A1.1. AMPLIFYING HORN

As the displacement amplitude of the converter is limited,

the role of the cone shape (amplifying horn) is very impor-
tant in order to increase the displacement of the specimen
and thus to raise its stress to the required level. There are
two antinodes and one node in the displacement distribution
of the cone (Figure A1.1). The amplification factor Fampl is
defined as




V1





V

2

Generally, a numerical analysis such as FEM is needed

to obtain the correct geometry of the horn incorporating the
effects of the temperature gradient and the variation of the
elastic modulus. The temperature of the horn may be reason-
ably assumed to vary linearly between two end values of 20
K
and T0, the temperatures of the cryostat and ambient air,
respectively, as shown in Figure A1.2.
Stress Calibration 261

Figure A1.1 Displacement distribution of amplifying horn (horn 1).

Figure A1.2 Temperature field of amplifying horn.

262 Appendix 1

Figure A1.3 Dimensions of the second horn (horn 2).

The amplification factor of the first horn (horn 1) shown

in Figure A1.2 is 3.52 at room temperature, 3.42 in liquid
nitrogen, and 3.40 in liquid helium. Horn 1 works well at
ambient temperature and in liquid nitrogen with or without
specimen. Meanwhile, the cone can only produce a maximum
stress of about 500 MPa. A second horn design is shown in
Figure A1.3, which is capable of developing stress between
650 MPa and 1300 MPa. Horn 2 gives an amplification factor
Fampl ¼ 8.93 in nitrogen and 8.98 in helium.

A1.2. FIRST CALIBRATION

The driving electrical voltage and power is supplied by a

Branson power source, and the first measurement is carried
out with an IBM PC computer.
Since horn 2 joined with the fatigue life specimen, does
not vibrate at ambient temperature, horn 1 must be used.
Figure A1.4 gives the measured signal in response to the
input signal; i.e., the voltage in plug 8. The open points are
the values measured from the optical sensor, and the full
points are those from plug 9. The linearity is practically
perfect.
Figure A1.5 presents the relation between the signal of
plug 9 in the connector J2 and the signal of the sensor
for horn 1 at ambient temperature. This relation can be
Stress Calibration 263

Figure A1.4 Response to input signal.

expressed by the equation

Uc1 ¼ 2:57 þ 4:146 Vm ðA1:1Þ
where Uc1 is the measured displacement of the first horn by
the sensor, and Vm is the voltage of plug 9. The symmetry
of the specimen, which implies identical displacement ampli-
tudes on two sides, has been taken into account. The voltage
applied to the ceramic imposes a displacement at the end of
the converter. For a given temperature, this displacement
must be constant. In a CNAM=ITMA test in liquid nitrogen,

Figure A1.5 Measured amplitude of signal.

264 Appendix 1

the temperature of the converter is above 0
C, because this

part does not freeze immediately after the removal of the
cryostat. Therefore, we consider that the displacement in
the converter is constant for a given electrical applied voltage,
whatever the mechanical load may be.
Based on the preceding analyses, the amplitudes of dis-
placement and stress are determined for horn 2 by the simple
calculation, as follows:
8:93
Uc2 ¼ Uc1 ðA1:2aÞ
3:52

s2 ¼ 16:2Uc2 ¼ 105:66 þ 170:54 Vm ðA1:2bÞ

Here, the value 8.93 is the amplification factor of horn 2 at
low temperatures, and 3.52 is that of horn 1 at room tempera-
ture. For material Ti6A4VPQ, one micron of displacement
corresponds to a stress of 16.2 MPa. Therefore the stress is
calculated by Eq. A1.2b using Eqs. A1.1 and A1.2. The con-
stant 105.66 in this equation comes from the electronics or
the adjustment of the A=D card. In other words, the initial
voltage is about 0.5 volt without vibration.
The curve of input stress-signal in Figure A1.6 is
employed in the control program to guarantee good conver-
gence.

Figure A1.6 Stress versus signal.

Stress Calibration 265

A1.3. SECOND CALIBRATION

With a computer PC486DX33 and a high-performance

Keithley card, we have carried out another calibration at
ambient temperature. The optical sensor measures the displa-
cement amplitude at the specimen’s head and a strain gauge
mounted on the specimen’s center measures the strain there.
The Ti6A4V specimen used is a uniform cylindrical bar with
124.72 mm in length. The two amplifying horns work at ambi-
ent temperature, which permits us to verify the stress for the
two horns. With the strain values of the specimen’s center
measured by the gauge and the displacement in the head
measured by the optical sensor, the strain along the specimen
can be calculated. The results are as follows.
Horn 1. The amplification factor Fampl is 3.52.
Displacement measured by the sensor
Um ¼ 17:4 mm; at 50% power

Um ¼ 35:4 mm; at 100% power

Subscript m indicates measured value, and subscript c for
calculated value in the following. Strain in specimen center
(calculated value)

2pf
ec ¼ pffiffiffiffiffiffiffiffiffi Um ¼ 25:19Um ðA1:3Þ
E=r

with f ¼ 20 kHz, E ¼ 110 GPa, and r ¼ 4420 kg=m3. Substitut-

ing measured values, we have
ec ðUm ¼ 17:4 mmÞ ¼ 438:3 m; at 50% power

ec ðUm ¼ 35:4 mmÞ ¼ 891:2 m; at 100% power

Strain in specimen center (measured by the gauge)

2 1000
em ¼ Vm ðmÞ ðA1:4Þ
K V1000
266 Appendix 1

where K ¼ 2.055 is the gauge factor, and V1000 ¼ 2.6 volts is

the voltage of calibration at 1000 m of strain. Similarly, substi-
tuting measured values, we have
em ðVm ¼ 1:2 VÞ ¼ 449:2 m; at 50% power

em ðVm ¼ 2:4 VÞ ¼ 898:4 m; at 100% power

Horn 2. The amplification factor Fampl is 9.16. Displace-
ment measured by the sensor
Um ¼ 43:4 mm at 50% power

Um ¼ 84:4 mm at 100% power

Strain in specimen center (calculated value)
ec ðUm ¼ 43:4 mmÞ ¼ 1093 m; at 50% power

ec ðUm ¼ 84:4 mmÞ ¼ 2126 m at 100% power

Strain in specimen center (measured by the gauge and with
V1000 ¼ 1.56 volts)
em ðVm ¼ 1:8 VÞ ¼ 1123 m; at 50% power

em ðVm ¼ 3:56 VÞ ¼ 2221 m; at 100% power

The results are summarized in Table A1.1
This verification is satisfactory. Note that the differ-
ence of strains from the measurement of the gauge and
the calculation, starting from the values measured by the
optical sensor, is about 2%. This demonstrates that the opti-
cal displacement sensor and the strain gauges used are reli-
able in the vibration environment of 20 kHz, and the stress
distribution in the antinodes is good. We can also see that
the measurements at 100% power are close to twice those
at 50% power. The strain gauge measurements give the
best results. The measurements show that the ratio of
vibration between horns 1 and 2 is 2.45 on the average,
while the calculated ratio is 2.6. The difference between
them is about 6%.
 Stress Calibration

Table A1.1 Results of Second Calibration

Fampl Um (50%) Um (100%) ec (50%) ec (100%) em (50%) em (100%)

Horn 1 3.52 17.4 mm 35.4 mm 438.3 m 891.2 m 449.2 m 898.4 m

Horn 2 9.16 43.4 mm 84.4 mm 1093 m 2126 m 1123 m 2221 m

No. 2=No. 1 2.602 2.494 2.384 2.494 2.386 2.500 2.472

267
268 Appendix 1

From these measurements, the stress can be calibrated

in accordance with the voltage of plug 9 in connector J2 that
is 5 volts at 50% power and 10 volts at 100% power. This vol-
tage is proportional to the displacement amplitude of the horn
(at point C of Figure 3.11) and does not depend on load. Tak-
ing account of the linearity, we only consider the state at
100% power in the following discussion.
The converter is always at ambient temperature; that is
to say, the amplitude is constant at a given power. This dis-
placement amplitude is denoted by UC. With the measure-
ment of the gauge at room temperature, the displacement
can be calculated at the head of the specimen (at point A of
Figure 3.11). The values are 35.6 mm and 88.6 mm for horns
1 and 2, respectively, both at 100% power. Because the displa-
cement of point B in Figure 3.11 is the same as that of point A,
horn 1 also gives an amplitude value of 35.6 mm, and horn 2
gives 88.6 mm at 100% power. These values are independent
of the specimens.
At low temperatures, the amplifying horns are more
rigid because their elastic moduli increases. Consequently
the amplification factor decreases slightly. Since UC is con-
stant, we have
   
UB UB
UC ¼ ¼ ðA1:5Þ
Fampl ambient Fampl cold
Using a superscript c for cold, we obtain
 
c c UB
UB ¼ Fampl ðA1:6Þ
Fampl ambient

Formula A1.6 can be used to include the influence of tempera-

ture of the horns.
Having obtained UBc (identical to UAc ), we are able to
determine the stress in the specimen. The stress depends on
the geometry of the specimen and the elastic modulus of the
material. The related formulas are available in Chapter 2,
both for a cylindrical bar and an endurance (fatigue life) spe-
cimen with a longitudinal profile of hyperbolic cosine in the
central portion.
Stress Calibration 269

A1.4. THIRD CALIBRATION

In the above, the verification of the calculations has been

made for ambient temperature. To know the behavior of the
machine at low temperatures, several gauges are mounted
on the cylindrical specimen of TA6VPQ. Theoretically, the
strain is of a sinusoidal function with the maximum in the
center and the minimum at two ends. For this reason, we
mount five small gauges (gauge length, 1.57 mm) in the cen-
tral region, and three long gauges (3.18 mm) in the zone of
small strain. Another specimen of Ti-6A4VPQ is also tested
where a gauge is used to compensate temperature effect.
The resistance of the gauge is 350 O. During the measure-
ment, the two specimens are soaked in liquid nitrogen. The
compensation gauge is linked to the demi-bridge. The results
of the calculations are compared with the measurements at
low temperature to verify the vibration mode analysis.
Material Ti-6A4VPQ was supplied by an industrial com-
pany. The static Young’s modulus at 77
K is 128 GPa. This
yields a resonance length of 134.5 mm at 20 kHz. But the test
machine did not vibrate in resonance under these conditions.
Only when the specimen was shortened by 3 mm did the sys-
tem vibrate in liquid nitrogen. This shows the real modulus is
lower than 122.4 GPa.
The measurements of strain start after the cooling of the
two specimens. For horn 2, the signal from gauges is quite
sinusoidal and the measurements give good results as shown
in Figure A1.7. The differences between the results of calcula-
tions and the measurements by the four gauges are about 5%.
The measurements verify that the antinodes of stress are
situated within 0 to 5 mm from the specimen center. Although
these tests are performed at 50% power from the ultrasonic
generator, the gauges wires broke one after another because
of acceleration. This is the reason that only a small number
of results have been obtained.
Figure A1.8 shows the strain values measured by gauges
for two horns at various levels of power. (Note: This figure
was shown in a simplified form as Figure 3.12.) The measure-
ment points lie very close to the straight line, em ¼ ec. For
270 Appendix 1

Figure A1.7 Results measured by gauges.

horn 1, the largest error is 7.7% at high power. For horn 2, the
error is 5.1% at 50 power, the same value as that of the pre-
vious test (Figure A1.7). Other points in Figure A1.8 are well
aligned and the error is lower than 2%.

Figure A1.8 Measurements and calculations of strain for horns 1

and 2.
Stress Calibration 271

From the calibration results, we can conclude that the

concept and the design of the test system is adequate. The
machine can be satisfactorily used to determine stress levels
with about 95% accuracy. The test results will be presented
in the other chapters.
(Wu, 1992) presents calibration procedures of an ultraso-
nic fatigue machine working at high temperatures.
Appendix 2
Remarks on the Statistical Prediction

As stated earlier, it is of interest to point out that many

structural components are subjected to beyond 107 cycles.
Normally the materials characterization and the fatigue pre-
diction are carried out from data limited to between 106 and
107 cycles. The fatigue life requirement of a car engine is
around 108 cycles. The big diesel engines for ships or high-
speed trains work up to 109 cycles. An extreme example for
the technical limit of fatigue life is about 1010 cycles for tur-
bine engines.
In principle, the fatigue limit is given for a number of
cycles to failure. Using the staircase method, for example,
the fatigue limit is given by the average alternating stress,
sD, and the probability of fracture is given by the standard
deviation of the scatter, s. A classical way to determine the
infinite fatigue life is to use a Gaussian function. Roughly
speaking, it is said that sD – 3 s gives a probability of fracture
close to zero. Assuming standard deviation is equal to 10 MPa,
the true infinite fatigue limit should be sD – 30 MPa. How-
ever, our experiments show that between sD for 106 and sD
for 109, the difference is much greater than 30 MPa for many
273
274 Appendix 2

Figure A2.1 S-N curves that typify fatigue test results for testing of
medium-strength steels. (From ASM Atlas of Fatigue Curves, p. 28.).

alloys. The so-called SD approach to the average fatigue limit

is certainly not the best way to reduce the risk of rupture in
fatigue. Whereas one is conscious that it is a convenient
approximation, only experience can remove the ambiguity
by appealing to some tests of accelerated fatigue. The concept
of a fatigue limit is bound to the hypothesis of the existence of
a horizontal asymptote of the S-N curve between 106 and 107
cycles (Figure A2.1). A sample that reaches 107 cycles and is
not broken is often considered to have an infinite life. This
is not a rigorous approach. It is important to understand that
if the staircase method is popular today to determine the fati-
gue limit at 107 cycles, it is because of convenience of the
approximation. A main goal of this work is to introduce a
new practical definition of the fatigue strength which extends
into the gigacycle regime, i.e., 109 cycles or more.

A2.1. REMARKS ON STATISTICAL ANALYSIS IN

THE MEGACYCLE REGIME

It has long been assumed that a log-normal relation takes into

account the scatter of fatigue results up to the megacycle
regime. The staircase method is based on this assumption.
Thus, the prediction of a fatigue strength for the so-called
infinite life depends on this assumption. In order to check this
Remarks on the Statistical Prediction 275

point, an investigation of the distribution of the fatigue life-

time of a high strength aluminium alloy 2024T3 has been
done in Bathias’s laboratory (Bathias, 2003). The specimen
used in the study was a dog-bone type with thickness of
4 mm and length of 130 mm. The test machine was an
INSTRON 8501. It was controlled by computer, for which load
fluctuation is less than 0.1 KN in constant amplitude tests.
More than 100 specimens were tested under constant
amplitude loading at 13 stress levels (the stress levels were
from 175 M to 400 MPa). For tests with maximum stress,
smax, of 240 MPa, 260 MPa, 280 MPa, and 300 MPa, the dis-
persion of lifetimes was very large. About 20 specimens were
run at each stress level to obtain enough samples to statisti-
cally analyze the distribution of logN for a given smax. The
experimental results of fatigue total lifetimes are shown in
Figure A2.2. The test data concentrated into two zones; the
boundary between the zones is about at 106 to 2  106 cycles.
It is clear that the data obtained in these experiments
(see in Figure A2.2) forms a narrow scatter band when the
stress level is high enough (smax > 320 MPa). When the

Figure A2.2 Results of fatigue lifetime of 2024-T3.

276 Appendix 2

Figure A2.3 Cumulative frequency of lifetime.

stress level is less than 300 MPa, the test points are not very
concentrated. A statistical analysis has been done and the
cumulative frequency of logarithms of lifetime on three stress
levels is presented in Figure A2.3. The statistical results
show that for smax ¼ 260 MPa, the dispersion of total fatigue
lifetime is very large. The test points do not conform to a
log-normal distribution (a steep straight line).
It is noted that, in the stress range from 240 MPa to
300 MPa, it is impossible to draft an acceptable S-N curve
with these test data.
All failed specimens have been analyzed by microfracto-
graphy methods. Some different types of inclusions were
found in this alloy and one of them was very fragile. When
the fatigue cracks initiated from these broken inclusions, they
denoted mode B. For those failed specimens in the second
zone, the slip bands were observed very clearly. Thus it could
be supposed that the persistent slip bands initiate the cracks
by mode A.
An S-N curve is drawn separately for the test data for
each different mode. These S-N curves with a new dual form,
composed of two curves, are shown in Figure A2.4.
This new dual S-N curve was also statistically analyzed.
Based on the two mode hypothesis, the cumulative frequency
of total fatigue lifetime at three stress levels is presented in
Figure A2.5. The results showed the experimental data for
each mode as a distribution in accordance with the log-normal
Remarks on the Statistical Prediction 277

Figure A2.4 The new S-N curve of aluminium alloy 2024-T3

noting two modes of initiation.

Figure A2.5 Cumulative frequency of lifetime in two modes.

behavior. A 13.5% and 21% level of significance is noted

separately for mode B and mode A. As compared with the
statistical results in Figures A2.3 and A2.5, the reduction of
dispersion of this new S-N curve shows remarkable improve-
ment.
 References

(Bajons, 1978) Bajons, P., Kromp, W. Determination of magnifica-

tion and resonance length of samples used in ultrasonic
fatigue tests. Ultrasonics Sept, pp. 213–217.
(Bathias, 1993) Bathias, C., Ni, J. Determination of fatigue limit
between 105 and 109 cycles using an ultrasonic fatigue device.
In: Michell, M. R., Landgraf, R. W., eds., Adv Fatigue Lifetime
Predictive Techniques. ASTM STP 1211, Vol. 2, Philadelphia:
ASTM, pp.141–152.
(Bathias, 1994) Bathias, C., Bechet, J., Wu, T., Jago, G. Une
machine de fatigue vibratoire fonctionnant à température
cryogenique et à 20 KHz. Application à l’étude du TA6VPQ à
20 KHZ et 77 K, Rapport – SEP, ITMA=CNAM.
(Bathias, 1996) Bathias, C. A review of fatigue of aluminium matrix
reinforced by particles or short fibres. Materials Science
Forum 217–222:1407–1412.
(Bathias, 1997) Bathias, C., El Alami, K., Wu, T. Y. Influence of
mean stress on Ti6A14V fatigue crack growth at very high
frequency. Engineering Fracture Mechanics 56(2):255–264.

279
280 References

(Bathias, 1998) Bathias, C. Is there an infinite fatigue life in metals

and alloys? EUROMECH 382=Fatigue Life in the Gigacycle
Regime. Paris; June 29–July 1.
(Bathias, 1999) Bathias, C. FFENS July.
(Bathias, 2001) Bathias, C. Designing components against gigacycle
fatigue. Fatigue in very high regime, Vienna, 2–4 July, pp.
97–107.
(Bathias, 2002) Bathias, C. A survey of the progress of piezoelectric
fatigue machines concept. Fatigue 2002, 2963–2971.
(Bathias, 2003) Bathias, C., Bin, C., Marines, I. An understanding
of very high cycle fatigue of metals. IJF 25:1101–1107.
(Bonis, 1997) Bonis, J. Fatigue of nickel base alloys at high
temperature in the gigacyclic domain. Thèse de doctorat,
University of Paris-Sud.
(Buchinger, 1990) Buchinger, L., Stanzl, S., Laird, C. Dislocation
structures produced by low-amplitude fatigue of copper single
crystals at 77 K. Philosophical Magazine A 62(6):633–651.
(Buchinger, 1984) Buchinger, L., Stanzl, S. E., Laird, C. Dislocation
structures in copper single crystals fatigued at low tempera-
ture. Philosophical Magazine 50(2):275–298.
(Dentsoras, 1983) Dentsoras, A. J., Dimarogonas, A. D. Resonance
controlled fatigue propagation in a beam under longitudinal
vibrations. Int J Fracture 23.
(Donald, 1997) Donald, J. K. Introducing the compliance ratio con-
cept for determinating effective stress intensity. Int J Fatigue
19(1).
(Donald, 1997) Donald, J. K., Bray, G. H., Bush, R. W. Introducing
the Kmax sensitivity concept for correlating fatigue crack
growth data. High Cycle Fatigue of Structural Materials,
TMS, (Minerals, Metals, Materials Society).
(Donald, 1999) Donald, J. K., Paris, P. C. An evaluation of DKeff
estimation procedures on 6061-T6 and 2024-T3 aluminum
alloys. Int. J. Fatigue 21.
(Donaldson, 1961) Donaldson, D., Anderson, W. E. Crack propaga-
tion behaviour of some airform materials. Crack propagation
symposium at Cranfield. In: Proceedings Vol. 2, pp. 375–441.
References 281

(Ebara, 1994) Ebara, R. Corrosion fatigue in practical problems. In:

Nisitani, H., ed. Comp. and Exp. Fract. Mech. UK: Comp.
Mech. Publ., pp. 347–376.
(Ebara, 1987) Ebara, R., Yamada, Y. Ultrasonic corrosion fatigue
testing of 13 Cr stainless steel and Ti-6A-4V alloys. In:
Toda, K., ed. Ultrasonic Technology. Tokyo: MYU Research,
pp. 329–342.
(Elber, 1970) Elber, W. Fatigue crack closure. J. Engin. Fracture
Mech 2(1).
(Elber, 1971) Elber, W. The significance of fatigue crack closure.
ASTM STP 486:230–242.
(Fritzemeier, 1982) Fritzemeier, L. G., Paulson, R. R. Discussion on
topics in ultrasonic fatigue. In: Wells, J. M., Buck, O., Roth,
L. D., Tien, J. K., eds. Ultrasonic Fatigue, Proceedings of the
First International Conference on Fatigue and Corrosion
Fatigue up to Ultrasonic Frequencies. The Metallurgical
Society of AIME, pp. 659–664.
(Girard, 1959) Girard, F., Vital, G. Rev Metall 56:25.
(Habertz, 1993) Habertz, K., Pippan, R., Stüwe, H. P. The threshold
of stress intensity range of iron. In: Bailon, J. P., Dickson,
J. I., eds. Fatigue 93EMAS, 525–530.
(Hertzberg, 1996) Hertzberg, R. W. Deformation and Fracture
Mechanics of Engineering Materials. 4th edn. New York: John
Wiley & Sons.
(Hertzberg, 1997) Hertzberg, R. W. Personal reflections of Paul
Paris’ first graduate student. In: High Cycle Fatigue of
Structural Materials TMS.
(Hoffelner, 1982) Hoffelner, W. The influence of frequency on fati-
gue crack propagation of some heat resisting alloys using
ultrasonic fatigue. In: Wells, J. M., Buck, O., Roth, L. D., Tien,
J. K., eds. Ultrasonic Fatigue, Proceedings of the First Interna-
tional Conference on Fatigue and Corrosion Fatigue up to
Ultrasonic Frequencies. The Metallurgical Society of AIME,
pp. 461–472.
(Hopkinson, 1911) Hopkinson, B. Proc. R. Soc. A86:101.
282 References

(Hudak, 1998) Hudak, S. J., et al. Growth of small cracks in aero-

engine disc materials. Final Report USAF Wright Laboratory.
(Jago, 1996) Jago, G. Relation between microstructure and proper-
ties in Ti alloys. Thèse de doctorat, CNAM.
(Jago, 1998) Jago, G., Bechet, J. Influence of microstructure on high
cycle fatigue, tensile tests and fatigue crack growth properties
in (a þ b) Ti6246 alloy. EUROMECH 382, Fatigue Life in the
Gigacycle Regime, Paris, June 29–July 1.
(Jenkin, 1925) Jenkin, C. F. Proc. R. Soc. A109:119.
(Jenkin, 1929) Jenkin, C. F., Lehmann, G. D. Proc. R. Soc. A125:83.
(Jones, 1991) Jones, J. W., et al. Fatigue behaviour of a 2xxx series
aluminium alloy reinforced with 15%SiCp. Met. Trans. A 22:
1007–1019.
(Kikukawa, 1965) Kikukawa, M., Ohji, K., Ogura, K. J. Basic Eng.
(Trans. ASME, D), 87:857.
(Kirchner, 1985) Kirchner, H. O. K., Kromp, W., Prinz, F. B.,
Trimmel, P. Plastic deformation under simultaneous cyclic
and unidirectional loading at low and ultrasonic frequencies.
Mat. Sci. Eng. 68:197–206.
(Kitagawa, 1979) Kitagawa, H., Takahashi, S. Fracture approach to
very small fatigue crackgrowth. Trans. Japan. Mech. Eng.
45:1289–1303.
(Kobayashi, 1991) Kobayashi, M., Fatigue strength prediction of
automobile suspension spring steels. ASME, MD-28, Impact
of Improved Materials Quality on Properties.
(Kong, 1987) Kong, X. Theoretical and numerical study on vibratory
fatigue. Thèse de doctorat, University of technology of
Compiègne.
(Kong, 1991) Kong, X., Saanouni, K., Bathias, C. On the fatigue at
very high frequency. J. Engin. Materials Techn. 113:205–214.
(Kozlov, 1988) Kozlov, A. V., Selitser, S. I. Peculiarities in the plas-
tic deformation of crystals to the acoustoplastic effect. Mat.
Sci. Eng. A(102):143–149.
(Kuzmenko, 1984) Kuzmenko, V. A. Fatigue of structural materials
at high-frequency cyclic loading. In: Valluri, S. R., Taplin, D. M.,
References 283

Rama Rao, P., Knott, J. F., Dubey, R., eds. Advances in Frac-
ture Research. 3rd edn. The Metallurgical Society of AIME,
pp. 1791–1798.
(Laird, 1982) Laird, C. Strain rate sensitivity effects in cyclic
deformation and fatigue fracture. In: Wells, J. M., Buck, O.,
Roth, L. D., Tien, J. K., eds. Ultrasonic Fatigue, Proceedings
of the First International Conference on Fatigue and Corrosion
Fatigue up to Ultrasonic Frequencies. The Metallurgical
Society of AIME. pp. 187–205.
(Laird, 1986) Laird, C., Stanzl, S., de la Veaux, R., Buchinger, L.
The cyclic stress-strain response and dislocation structures
of Cu-16 at % Al alloy. II: Polycrystalline behaviour. Mat.
Sci. Eng. 80:143–154.
(Li, 1992) Li, D. Z. Fretting and protecting. Shanxi Science and
Technique Press, pp. 69, 145, and 178.
(Lindner, 1965) Lindner, B. MS thesis (directed by P. C. Paris),
Lehigh University.
(Lutynski, 1982) Lutynski, C., Simanski, G., McEvily, J. Fretting
fatigue of Ti-6Al-4V alloy. ASTM STP, 780, 150–164.
(Lindley, 1977) Lindley, T. C. Fretting fatigue in engineering alloys.
Int J Fatigue 19(s1):s39–s49.
(Marines, 2003) Marines, I., Baudry, G., Vittori, J. F., Doucet,
J. P., Bathias, C. Ultrasonic fatigue tests on bearing steel
AISI-SAE 52100 at frequency of 20 and 30 kHz, Intrnl.
J Fatigue 25(9–11):1037–1046.
(Manson, 1956) Manson, W. P. Piezoelectric Crystals and their
Application in Ultrasonics. New York: Van Nostrand, p. 161.
(Mason, 1982) Mason, W. P. Use of high amplitude strains in study-
ing wear and ultrasonic fatigue in metals. In: Well, J. M.,
Buck, O., Roth, L. D., Tien, J. K., Dubey, R., eds. Ultrasonic
Fatigue, Proceedings of the First International Conference
on Fatigue and Corrosion Fatigue up to Ultrasonic Frequen-
cies. The Metallurgical Society of AIME, pp. 87–102.
(Masuda, 1994) Masuda, C., et al. Fatigue crack propagation
mechanism of Sic whisker or Sic particle reinforced aluminium
matrix composites. Adv. Composite Mater. 3(4):319–339.
284 References

(Mayer, 1994) Mayer, H. R., Laird, C. Influence of cyclic frequency

on strain localisation and cyclic deformation in fatigue. Mat.
Sci. Eng. A187: 23–25.
(Mayer, 1991) Mayer, H. R., Stanzl, S. E., Tschegg, E. K. Fatigue
crack propagation in the threshold regime after rapid load
reduction. Eng. Fracture Mech. 40(6):1035–1043.
(Mayer, 1992) Mayer, H. R., Stanzl, S. E., Tschegg, E. K., Schijve, J.
Fatigue crack growth of Al 2024-T3 under low amplitude two-
step loading. Fatigue Fract. Eng. Mater. Struct. 15(3):265–275.
(Mayer, 1993) Mayer, H. R., Stanzl, S. E., Tschegg, E. K., Tan, D. M.
FEM modeling of stress intensity factors for fatigue crack
growth at ultrasonic frequencies. Engi. Fracture Mech.
45(4):487–495.
(Mayer, 1994) Mayer, H. R., Tschegg, E. K., Stanzl-Tschegg, S. E.
High-cycle fatigue measurements on ceramic materials in
combined loading (cyclic torsion and static compression). In:
Pineau, A. ed. Proceedings of Fourth International Conference
on Biaxial and Multiaxial Fatigue. ESIS, Paris, Vol. 2, pp.
357–368.
(Mayer, 1999) Mayer, H. R. Fatigue crack growth and threshold
measurements at very high frequency. Int. Mat. Rev. 44:1;
34:1.
(McEvily, 1958) McEvily, A. J. Jr., Illg, W. The rate of fatigue crack
propagation in two aluminum alloys, NACA TN 4394.
(McClintock, 1997) McClintock, F. A. The Paris law for fatigue
crack growth in terms of crack tip opening displacement. In
High Cycle Fatigue of Structural Materials, TMS.
(Mitsche, 1973) Mitsche, R., Stanzl, S., Burkert, D. G. Hochfre-
quenzkine-matographie in der metallforschung. Wissenschafli-
cher film 14:3–10.
(Murakami, 2002) Murakami, Y. Metal Fatigue: Effects of Small
Defects and Non-metallic Inclusions. Amsterdam: Elsevier.
(Murakami, 1994) Murakami, Y. Effects of defects, inclusions and
inhomogeneities on fatigue strength. Fatigue 16:163–182.
(Mutoh, 1995) Mutoh, Y. Mechanisms of fretting. Fatigue, JSME,
Int. J Series A38:405.
References 285

(Nadot, 1999) Nadot, Y., Mendez, J., Ranganathan. Fatigue life

assessment of nodular cast iron containing cast defects. FFEM
22:289–300.
(Nakazawa, 1992) Nakazawa, K., Sumita, M., Maruyama, N. Effect
of contact pressure on fretting fatigue of high strength steel
and titanium alloy. In: Attia, H. A., Waterhouse, R. B. (eds.)
Standardisation of Fretting Fatigue Test Methods and
Equipment. ASTM STP 1159, p. 115.
(Neppiras, 1959) Neppiras, E. A. Techniques and equipment for
fatigue testing at very high frequencies. In: Proc. ASTM 59.
Philadelphia: ASTM pp. 691–710.
(Ni, 1991) Ni, J. Mechanical behaviour of alloys in ultrasonic
fatigue. Thèse de doctorat, CNAM.
(Ni, 1994) Ni, J., Bathias, C. Development of an ultrasonic fatigue
device and its application in fatigue behaviour studies. In:
Silva Gomes, et al. (eds).: Recent Advances in Experimental
Mechanics. Rotterdam, Balkerma: pp. 1121–1126.
(Ni, 1995) Ni, J. Determination of stress intensity factor in ultraso-
nic fatigue loading by means of dynamic modal analysis and
three-dimensional finite element calculation. Eng. Fracture
Mech. 52(6):1079–1086.
(Ni, 1996) Ni, J. On the determination of stress intensity factors in
ultrasonic fatigue loading. Int. J. Fatigue 18(7):457–461.
(Ni, 1996) Ni, J., Zhang, X., Nie, J. 3-D Finite Element Computation
and Dynamic Modal Analysis on Ultrasonic Vibration
Systems. Science in China (Series E), Vol. 39, No. 2, pp.
136–140.
(Nishijima, 1999) Nishijima, S. Stepwise S-N curve and fish-eye
failure in gigacycle fatigue. Fatigue Fracture Engineering
Materials Structure 22, Blackwell Science Ltd., pp. 601–607.
(Papakyriacou, 1995) Papakyriacou, M., Mayer, H. R., Tschegg-
Stanzl, S. E., Groschl, M. Near-threshold fatigue crack growth
in Al2O3 particle reinforced 6061 aluminium alloy. Fatigue
Fract. Eng. Mater. Struct. 18(4):477–487.
(Paris, 1962) Paris, P. C. The growth of cracks due to variation in
load. Ph. D. dissertation, Lehigh University.
286 References

(Paris, 1963) Paris, P. C. The fracture mechanics approach to

fatigue. In: Sagamore Conference Proceedings, pp. 107–127.
(Paris, 1965) Paris, P. C., Sih, G. Stress analysis of cracks. ASTM,
pp. 315–381.
(Paris, 1970) Paris, P. C. Testing for very slow growth of fatigue
cracks. Closed Loop (MTS Systems Corp.) 2(5).
(Paris, 1999) Paris, P. C., Tada, H., Donald, J. K. Service load
fatigue damage—a historical perspective. Int. J. Fatigue 21.
(Puskar, 1986) Puskar, A., Varkoly, L. Influence of temperature on
fatigue crack growth behaviour of steels at ultrasonic
frequency. Fatigue Fract. Eng. Mater. Struct. 9(2):143–150.
(Rayaprolu, 1992) Rayaprolu, D. B., Cook, R. A critical review of
fretting fatigue investigations at the Royal Aerospace Estab-
lishment. In: Attia, H. A., Waterhouse, R. B (eds.): Standardi-
sation of Fretting Fatigue Test Methods and Equipment.
ASTM STP 1159, pp. 129–152.
(Realigns, 1972) Realigns, C. T., Sandorff, P. E. Slip Front Mechan-
ism in Mechanical Joints.
(Roth, 1985) Roth, L. D. Ultrasonic fatigue testing. Metals
Handbook. Vol. 8. 9th edn. Mechanical Testing, American
Society for Metals, pp. 240–258.
(Roth, 1982) Roth, L. D., Willertz, L. E., Leax, T. R. On the fatigue of
copper up to ultrasonic frequencies. In: Wells, J. M., Buck, O.,
Roth, L. D., Tien, J. K., eds. Ultrasonic Fatigue, Proceedings of
the First International Conference on Fatigue and Corrosion
Fatigue up to Ultrasonic Frequencies. The Metallurgical
Society of AIME. pp. 265–282.
(Saka€
, 1999) Saka€
, T. et al. Experimental evidence of duplex S-N
characteristics in wide life region for high strength steels.
Fatigue.
(Satoh, 1993) Satoh, T. et al. Fretting fatigue damage and fatigue
crack initiation process. In: Proceedings of 38th Annual
Meeting on Materials Research.
(Schmidt, 1973) Schmidt, R. A., Paris, P. C. Threshold for Fatigue
Crack Propagation and the Effect of Load Ratio and
Frequency. ASTM STP 536, pp. 79–94.
References 287

(Shoeck, 1982) Shoeck, G. Calculation of the stress intensity in

ultrasonic resonance. Z. Metallkde 9:576–578.
(Sirian, 1982) Sirian, C. R., Conn, A. F., Mignogna, R. B., Green, R.
E. Method of measuring elastic strain distribution in speci-
mens used for high frequency fatigue testing. In: Wells, J.
M., Buck, O., Roth, L. D., Tien, J. K., eds. Ultrasonic Fatigue,
Proceedings of the First International Conference on Fatigue
and Corrosion Fatigue up to Ultrasonic Frequencies. The
Metallurgical Society of AIME, pp. 87–102.
(Soderberg, 1986) Soderberg, S., Colvin, T., Salama, K., Vingsbo, O.
Ultrasonic fretting wear of a plain carbon steel. J. Engin.
Materials Technol. Transactions of the ASME 108:153–158.
(Speidel, 1980) Speidel, M., Stanzl, S., Tschegg, E. Ermüdung von
stahl X20Cr13, risswachstum bei lastwechselfrequenzen von
10
3 bis zu 21 kHz, Bestimmung des grenzwertes DK0 mit
ultraschall, Zeitschrift für werkstofftechnic. 11:305–308.
(Stanzl, 1981) Stanzl, S. A new experimental method for measuring
life and crack growth of materials under multi-stage and
random loadings. Ultrasonics 11:269–272.
(Stanzl, 1981) Stanzl, S., Tschegg, E. Fatigue crack growth and
threshold measured at very high frequencies. Metal Sci.
(Stanzl, 1984) Stanzl, S., Hollanek, W., Tschegg, E. K. Fatigue and
fracture under variable-amplitude loading at ultrasonic fre-
quency. In: Valluri, S. R., Taplin, D. M. R., Rao, T. R., Knott,
J. F., Dubey, R., eds. Advances in Fracture Research. Vol. 5,
pp. 3645–3651.
(Stanzl, 1986) Stanzl, S. Fatigue testing at ultrasonic frequencies.
J. Soc. Environm. Eng. 25(1):11–16.
(Stanzl, 1991) Stanzl, S., Mayer, H. R., Tschegg, E. K. The influence
of air humidity on near-threshold fatigue crack growth of
2024-T3 aluminium alloy. Materials Science and Engineering
A147:45–54.
(Stanzl, 1993) Stanzl-Tschegg, S. E., Mayer, H. R., Tschegg,
E. K. High-frequency method for torsion fatigue testing.
Ultrasonics 31(4):275–280.
288 References

(Stanzl, 1993) Stanzl-Tschegg, S. E., Mayer, H. R., Tschegg, E. K.,

Beste, A. In-service loading of AlSill aluminium cast alloy in
the very high cycle regime. Int. J. Fatigue 15(4):311–316.
(Stanzl, 1996) Stanzl, S. Ultrasonic fatigue. Fatigue 96:1887–1898.
(Stanzl, 2001) Stanzl, S. International Conference of Very High
Cycle Fatigue II. Vienna, 2–4 July.
(Stickler, 1982) Stickler, R., Weiss, B. Review of the application of
ultrasonic fatigue test methods for the determination of crack
growth and threshold behaviour of metallic materials. In:
Wells, J. M., Buck, O., Roth, L. D., Tien, J. K., eds. Ultrasonic
Fatigue, Proceedings of the First International Conference on
Fatigue and Corrosion Fatigue up to Ultrasonic Frequencies.
The Metallurgical Society of AIME, 135–171.
(Sun, 2001) Sun, Z. Ph. D. thesis: Fatigue Crack Growth Threshold
at High Frequency, CNAM,-Paris.
(Sun, 1999a) Sun, Z., Wang, Q., Bathias, C. Experimental investiga-
tion on fretting fatigue of a high strength steel at ultrasonic
frequency. Fatigue 99, Beijing, China, pp. 169–174.
(Sun, 1999b) Sun, Z., Bathias, C., Baudry, G. Fretting fatigue
experiment at ultrasonic frequency. Report of ITMA.
(Tada, 2000) Tada, H., Paris, P. C., Irwin, G. R. The Stress Analysis
of Cracks Handbook, ASME Press.
(Tao, 1996) Tao, H. Ultrasonic fatigue of Ti and alloys at cryogenic
temperature. Thèse de doctorat, CNAM.
(Tao, 1996) Tao, H., Bathias, C. Experimental study on fretting-fati-
gue at very high frequency. Revue de Métallurgie–
CIT=Science et Génie des Matériaux, pp. 687–695.
(Tein, 1975) Tein, V. V., Seibert, P. E. Fretting of structures for
modern V. G. fighters. AGARD-CP, p. 161.
(Thanigaiyarasu, 1988) Thanigaiyarasu, G. Contribution to the
study of fatigue life and crack growth in vibration fatigue.
Thèse de doctorat, University of technology of Compiègne.
(Tschegg, 1981) Tschegg, E. K., Stanzl, S. Fatigue crack propaga-
tion and threshold in b.c.c. and f.c.c. metals at 77 and 293 K.
Acta Metal. 29:33–40.
References 289

(Vasudevan, 2003) Vasudevan, A. K. The effect of microstructure

and environment on fatigue crack growth in 7049 aluminium
alloy at negative stress ratios. Intl. J. Fatigue 25:1209–1216.
(Wang, 1996) Wang, Q., Bathias, C. Réalisation d’un système de
fatigue vibratoire sur tôles minces. Partie I : Conception et
mise au point, Contrat Renault-Rapport 96-2, ITMA=CNAM.
(Wang, 1997) Wang, Q., Bathias, C. L’évolution de la température
en fatigue vibratoire, Rapport 97–2, ITMA=CNAM.
(Wang, 1998a) Wang, Q., Berard, J. Y., Rathery, S., Bathias, C.
Gigacycle fatigue of a spheroidal graphite cast iron,
EUROMECH 382=Fatigue Life in the Gigacycle Regime.
Paris, June 29–July 1.
(Wang, 1998b) Wang, Q., Dubarre, A., Baudry, G., Bathias, C. High
cycle fatigue and life prediction of high strength steels,
EUROMECH 382=Fatigue Life in the Gigacycle Regime.
Paris-June 29–July 1.
(Wang, 1998c) Wang, Q., Sun, Z., Berard, J. Y., Rathery, S.,
Bathias, C. High cycle fatigue crack initiation and propagation
behaviour of high-strength spring steel wire, EUROMECH
382=Fatigue Life in the Gigacycle Regime, Paris, June 29–
July 1.
(Wang, 1998d) Wang, Q., Sun, Z., Berard, J. I., Rathery, S., Bathias,
C. High cycle fatigue crack initiation and propagation
behaviour of high-strength spring steel wire. EUROMECH
382=Fatigue Life in the Gigacycle Regime. Paris.
(Waterhouse, 1971) Waterhouse, R. B., Tayler, D. E. The initiation
of fatigue cracks in a 0.7% carbon steel by fretting. Wear
17:139.
(Wells, 1982) Wells, J. M. Ultrasonic fatigue, AIME-ISBN member
0-89520-397-9.
(Weiss, 1982a) Weiss, B., Stickler, R., Fembock, J., Pfaffinger, K.
Evaluation of fatigue limits and crack growth properties of
PM-Mo alloys tested at cyclic frequency of 200 Hz and
20 kHz. In: Wells, J. M., Buck, O., Roth, L. D., Tien, J. K.,
eds. Ultrasonic Fatigue, Proceedings of the First International
Conference on Fatigue and Corrosion Fatigue up to Ultrasonic
Frequencies. The Metallurgical Society of AIME, pp. 505–537.
290 References

(Weiss, 1982b) Weiss, B., Stickler, R., Schider, S., Schmidt, H. Cor-
rosion fatigue testing of implant materials (Nb, Ta, stainless
steel) at ultrasonic frequencies. In: Wells, J. M., Buck, O.,
Roth, L. D., Tien, J. K., eds. Ultrasonic Fatigue, Proceedings
of the First International Conference on Fatigue and Corrosion
Fatigue up to Ultrasonic Frequencies. The Metallurgical
Society of AIME, pp. 387–411.
(Whitlow, 1982) Whitlow, G. A., Willertz, L. E., Tien, J. K. Effects of
frequency on the corrosion fatigue of Udimet 720 in a molten
sulfate salt. In: Wells, J. M., Buck, O., Roth, L. D., Tien, J. K.,
eds. Ultrasonic Fatigue, Proceedings of the First International
Conference on Fatigue and Corrosion Fatigue up to Ultrasonic
Frequencies. The Metallurgical Society of AIME, pp. 321–331.
(Willertz, 1980) Willertz, L. E. Ultrasonic fatigue, International
Materials Reviews, N. 2, pp. 65–77.
(Willertz, 1982) Willertz, Rust, T. M., Swaminathan, V. P. High and
low frequency corrosion fatigue of some steam turbine blade
alloys. In: Wells, J. M., Buck, O., Roth, L. D., Tien, J. K.,
eds. Ultrasonic Fatigue, Proceedings of the First International
Conference on Fatigue and Corrosion Fatigue up to Ultrasonic
Frequencies. The Metallurgical Society of AIME, pp. 333–348.
(Wright, 1971) Wright, G. P., O’Conner, J. J. Finite-element analy-
sis of alternating axial loading of an elastic plate between two
elastic rectangular blocks with finite friction. Intl. J. Engin.
Sci. 9:555.
(Wu, 1991) Wu, T., Bathias, C. Contrat DRET N 8707101,
ITMA=CNAM.
(Wu, 1992) Wu, T. Modelisation de la fissure en fatigue vibratoire à
haute température. Application aux alliages à base de nickel.
Thèse doctorale, Ecole Centrale de Paris.
(Wu, 1994) Wu, T., Bathias, C. Application of fracture mechanics
concepts in ultrasonic fatigue. Eng. Fracture Mech. 47(5):
683–690.
(Wu, 1994) Wu, T., Ni, J., Bathias, C. An automatic ultrasonic fati-
gue testing system for studying low crack growth at room and
high temperatures. Amzallag, C. ed., Automation in Fatigue
and Fracture: Testing and Analysis, ASTM STP 1231, ASTM,
Philadelphia, pp. 598–607.
References 291

(Wu, 1996) Wu, T., Jago, G., Bechet, J., Bathias, C. Accelerated
vibratory fatigue test by ultrasonic frequency at cryogenic
temperature. Eng. Fracture Mech. 54(6):891–895.
(Yeske, 1982) Yeske, R. A., Roth, L. D. Environmental effects on
fatigue of stainless steel at very high frequencies. In:
Wells, J. M., Buck, O., Roth, L. D., Tien, J. K., eds. Ultrasonic
Fatigue, Proceedings of the First International Conference on
Fatigue and Corrosion Fatigue up to Ultrasonic Frequencies.
The Metallurgical Society of AIME, pp. 365–385.
 Listing of Materials Tested in
Gigacycle Fatigue

Iron and steels [Aluminium alloys]

Cast iron A291
AISI 1006 AZ91
4240 Titanium alloys
D 38MSV5 TI6A4V4
54SiCr6 Ti6246
55SiCr7 TiAl
100C6 Nickel alloys
SUJ2 Astrology
304 INCO 706
403 INCO 718
17-6PH U500
13-8Mo N18
12Cr Cu metal
Aluminium alloys Metal matrix composites
AS5U3G Al203=6061
AS10U3G Al203=AS7G06
AS10G SIC=2080
2024 SIC=2124
2020 SIC=AC8
2219 SIC=AC4
8090
293
 Index

100C6 steel 2080=15%SiCp composite, S-N

fatigue curve for, 101, 237 curves, 107
chemical composition of, 100 304 stainless steel, S-N curve, 103,
mechanical properties of, 100 233
12% Cr steel 4240 Rep-B, fretting fatigue test, 204
chemical composition of, 99 4240 steel
mechanical properties of, 99 S-N curves of, 93
S-N data for, 234 subsurface initiation, 93
17%SiC surface-subsurface transition
mechanical properties of, 108 in, 253
composition of, 108 6061 alloys
17-4 steel mechanical properties of, 171
and 12%Cr steels, partial closure predicted
experimental results, 99 threshold for, 144
fatigue crack growth with 6061=Al2O3=15p, 172
striation formation, 212 6061=Al2O3=21p, 172
internal initiation in, 254 6246 Ti alloy, microstructure, 113
20 kHz-resonance method, 31 7055 aluminum
2024-T3 applied stress intensity for, 140
FCG curves of, 161 load method for, 140
high maximum stress partial closure 2=Pi0
intensity data, 141 method for, 141
partial closure model for, 144 7075 alloy, S-N curves, 107

295
296 Index

7075 T6, fatigue crack growth [Aluminium alloys,]

data, 134, 136 at low and high frequency, 210
9310 Gear steel, threshold gigacycle fatigue of, 239
behavior, 137 in liquid hydrogen
and magnesium alloys, 119
S-N curves for, 223
Aluminium matrix composite, 105
A533 steel, threshold data for, 138 Aluminium-lithium alloy, 121
Acoustic wave train, 83 fretting fatigue of, 194
AISI403 stainless steel, cathodic Anisotropy of metals, 243
polarization effect of, 226 ANSYS, calculation, 41–43
Al2024 alloy Applied stress intensity for 7055
chemical composition of, 160 aluminium, 140
low amplitude two-step Armco iron, frequencies, 213
loading, 164 Astroloy, 216, 217
mechanical properties of, 160 composition of, 152
Al6061 alloys, chemical crack growth rate of, 153
composition of, 170 heat treatment of, 152
Al6061-T6 alloy, 170 mechanical properties of, 152
Al-Li8090 nickel-based alloys, 151
chemical composition of, 121 Asymptotic approximation, 31
mechanical properties of, 122 Austenitic stainless steel, 102, 181
mechanical properties of, 194 fatigue crack growth
S-N curve of, 122 rates, 181–183
Alloys stress ratio, 183
Al6061-T6, 170 Autoclave description, 75
aluminium, 158 Automobile engines, fatigue life of, 1
aluminium-lithium, 121
at cryogenic temperatures, 118
gigacycle fatigue failures in, 1
nickel-based, 151 b.c.c. and f.c.c. crystalline structure,
non-ferrous, 109 173
tested to gigacycle fatigue Blaha effect, 215
chemical composition of, 230 Boundary collocation, 31
mechanical properties of, 230
Ti-Al intermetallic, 127–132
titanium, 109, 119, 146
without inclusions, Calibration tests, 64–66
gigacycle fatigue of, 249 Carbon steel, frequency effects, 213
AlSi cast alloy, 167 Cast aluminium alloy, fatigue crack
chemical composition of, 167 under the surface in, 256
material 1, FCG curve of, 169 Cast iron
material 2, FCG curve of, 169 fatigue crack propagation, 187
surface initiation, 255 gigacycle fatigue of, 231
Aluminium alloys, 117, 158 S-N curves, at low and high
AS5U3G, S-N curve, 242 frequency, 210
Index 297

[Cast iron] Computer control system, 56

spheroidal graphite, Computer testing system,
S-N curve of, 235 advantages of, 66–67
stress levels, 219 Coolants, 218–219
test frequency on fatigue Copper
crack growth in, 188 chemical composition, 174
Catenoidal approximation heat treatment of, 175
of a circle, 20 mechanical properties of, 175
Cathodic polarization effect of, Corrosion fatigue, 219
AISI403 stainless steel, 226 surgical implant materials, 226
Chemical composition Udimet, 720 225
12% Cr steel, 99 Corrosion pits, 224
Al2024, 160 Crack growth, 6
Al6061 alloys, 170 and initiation, relation
Al-Li8090, 121 between, 244
alloys, tested to gigacycle and threshold, 133–206
fatigue, 230 Crack growth life,
AlSi Alloy, 167 estimation of, 244–249
cast iron SGI52, 104 Crack growth rate
copper, 174 of Astroloy, 153
high strength steel, 91 of N18, 154
Inconel, 706, 155 Crack initiation, 6
low carbon steel, 102 in cylindrical samples, 241
mild steel, 174 in gigacycle fatigue, 232
spheroidal graphite cast mechanisms, 109, 242–243
iron SGI52, 104 in ultrasonic fatigue loading, 117
steel, 304, 174 Crack propagation behavior, 166
SUJ2, 100 Crack tip stress intensity factor,
Ti-6246 alloy, 251 calculation of, 30–49
Ti6246, 110 Crack tip zones of plasticity, 135
Ti-Al, 129 Cr-Si steel, S-N data of, 97
Circle and catenoidal Cr-V steel, S-N data of, 96
approximation, 21 Cryogenic temperature, 220–223
Cold-rolled carbon steel sheet alloys at, 118
composition of, 179 fatigue tests, 69
mechanical properties of, 179 Cryostat, S-N temperature
Command box, 65 gradient, 222
Composition Cumulation fatigue test, 149
of 17%SiC, 108 fracture aspects of, 150
of Astroloy, 152 Curve and fatigue strength,
of cold-rolled low carbon S-N data, 87–132
steel sheet, 179
of superalloy N18, 123
of T6A4V, 115
of U500, 114 D38 Steel, S-N curves
Compressed air cooling systems, 219 at low and high frequency, 210
298 Index

Database of threshold SIF, DKth, 189 Fatigue crack growth specimens,

Defects dimensions of, 30
and crack growth, 6 Fatigue Crack Growth (FCG)
at interior of the alloys, 254–255 study, 31
Deformation mechanism, 215 Fatigue curve
Detuning effect of the resonant 100C6 steel, 237
vibration, 48 safe, 5
Diesel engines, fatigue life of, 1 Fatigue initiation
Dimensions of ultrasonic fatigue, mechanisms of, 242
three-point bending, 86 site, 240
Dislocation structures, 214–216 in Ti-6246 alloys, 250
Displacement node (DN), 27, 29 Fatigue life
Gaussian function, 4
Murakami formula, 246
specimens, dimensions of, 30
Elastic wave theory, 11, 12–15 Fatigue limit, 3, 6
Elastic wave, one-dimensional average, SD approach, 4
formulas of, 28 environmental, 224
Element mesh, finite, 37 modeling and S-N data, 89
Engineering alloys, fatigue strength S-N curve and, 208
at conventional frequencies, 225 staircase method, 4
at ultrasonic frequencies, 225 in water, 225
Environmental effects, 223–227 Fatigue machine
Environmental fatigue limit, 224 high pressure, 73
Estimation of crack growth life, piezo-electric, 73
244–249 Fatigue properties
surgical implant materials,
corrosion, 226
with variable amplitude
Fatigue crack growing mechanism loading, 10
ultrasonic vs. conventional Fatigue results
loading regimes, 209 100C6, 101
Fatigue crack growth 1SUJ2 bearing steel, 101
behavior trends, 133–145 T6A4V, 121
in cast aluminium alloy, 256 Fatigue strength
in cast iron, 187 engineering alloys, 225
general relationship, 145 reduction factor, 203
striation formation T6A4V, 213
17-4PH, 212 Fatigue test machine
T6A4V, 211 full resonance system, 53
threshold (DKth), 6 low temperature and
Fatigue crack growth data high frequency, 70
7075-T6, 134, 136 measuring systems, 52
T6A4V, 199 performance of, 51–86
Ti-6246 alloy, 252 recording system, 52
titanium alloys, 146 three-point bending, 82
Index 299

[Fatigue test machine] Fretting fracture surface, 204

with computer, 54 Fretting scar on T6A4V, 198–199
Fatigue testing Fretting system
cryogenic temperatures, 69 for piezo-electric machine, 80
ultrasonic fretting, 75 schematic, 80
Fatigue-limit differences at Fretting wear, 81, 216
109 cycles, 220 Full resonance system, 53
FCG curve
AlSi cast alloy
material 1, 169
material 2, 169 Gigacycle fatigue
2024-T3, 161 alloys without inclusion, 249
in humid air, 161 aluminium alloys, 239
in a vacuum, 161 cast iron, 231
N18, 154, 159 mechanisms, general
Inconel 706, 156–159 discussion of, 253
at 400 C, 158 Ni base alloys, 234
thin steel sheet, 180 steels, 231
FCG specimen, with titanium alloys, 234
amplifying horns, 44 Gigacycle S-N curve
FEM (finite element method), 16, 31 concept of, 5
Ferrous materials, 90–105 shape of various alloys, 229–230
Finite element method (FEM), 16 Green’s function, 31
Finite element models, three-
dimensional, 44
Fractography, 147
Fracture aspects, of cumulative Heat effect, 216–220
fatigue, 150 Heat treatment
Fracture mechanics concepts, Astroloy, 152
31–43, 133–145 copper, 175
Frequencies, Armco iron, 213 high strength steel, 91
Frequency and resonance length, 15 Inconel 706, 155
Frequency effect mild steel, 175
carbon steel, 213 steel 304, 175
at cryogenic temperature, 209 superalloy N18, 124
on experimental results, 207 T6A4V, 115
Frequency, 214 U500, 115
Fretting fatigue crack High maximum stress intensity
initiation point, 206 data, of 2024T-3, 141
propagation path, 206 High pressure fatigue machine, 73
Fretting fatigue, 76, 194 High speed trains,
4240 Rep-B, 204 fatigue life of, 1
aluminium-lithium alloy and High strength steel, 91–94
titanium, 194 fretting fatigue of, 200
high strength steel, 200 High temperature test
slip amplitude, 76 equipment, 67
300 Index

High-cycle fatigue properties of Low temperature test

Ti-Al, 130 equipment, 69
high frequency, 70
LP (loop patch) structure, 215

Inclusions
N18 alloy, 256
role of, 243 Magnesium alloys, 119
Inconel 706, 155 Martensitic stainless steel, 98
at 400 C, FCG rate of, 158 S-N curves, 239
chemical composition of, 155 Maximum strain values, 54
FCG rate of, 156–159 Mechanical aspects of initiation, 231
heat treatment of, 155 Mechanical properties
mechanical properties of, 155 100C6, 100
Infinite fatigue life, 244 12% Cr Steel, 99
Inhomogeneous microstructure, 17% SiC, 108
254–255 6061 alloys, 171
Initiation zone for low cycle to Al-Li8090, 122, 194
gigacycle failures, 241 alloys tested to gigacycle
Initiation and crack growth, 244 fatigue, 230
mechanical aspects of, 231 Astroloy, 152, 220
Internal crack initiation cold-rolled low carbon
17-4 steel, 254 steel sheet, 179
rail steel, 254 copper, 175
rotating-bending stress high strength steel, 91
correction, 125 Inconel 706, 155
Isoprobability of failure, 4 low carbon steels, 102
mild steel, 175
spheroidal graphite cast iron,
SGI52, 104
K, formula for ultrasonic fatigue, 38 steel, 304 175
Kitagawa diagram, 248 SUJ2, 100
superalloy N18, 124
T6A4V, 115
U500, 115
Load method, for 7055 aluminum, Mechanisms of fatigue
140 initiation, 242
Longitudinal elastic waves and Microfractographic analyses, 90
resonance frequency, 11–15 Microstructural aspects and
Loop patch structure, LP, 215 damage, 229–257
Low carbon ferritic steel thin Mild steel
sheets, 101 chemical composition of, 174
Low carbon steel, 102, 179 heat treatment of, 175
chemical composition of, 102 mechanical properties of, 175
mechanical properties of, 102 temperature effect, on fatigue
S-N curve, 102, 232–233 crack growth rates, 178
Index 301

Miner’s law for crack propagation Pores in N18 alloy, 256

rate, 166 Powder metallurgy alloys, 255–256
Mixed rupture, transgranular and Power generator, 52
intergranular, 190 PSBs (persistent slip bands), 214–215
MMC, fatigue curve for, 109 Pulsed loading, 219
Murakami model, 90–91
fatigue limit at 109 cycles, 246
N18 nickel alloy and, 248
Rail steel
internal initiation in, 254
S-N curve, 237
N18 nickel alloy, 152 Resonance length
crack growth rate of, 154 analytical solution of, 22
fatigue crack growth rate, 217–218 crack growth specimen, 25–27
FCG rate curves of, 154, 159 endurance specimen, 22
frequency, 214 frequency and, 15
at high temperature, 123 transition section, 27
Murakami model and the, 248 Rotating-bending, internal crack
pores and inclusions, 256 stress correction, 125
S-N curves, 124, 240 Rupture feature ultrasonic vs.
Ni base alloys, gigacycle fatigue, 234 conventional loading
Nickel-based alloys, 151 regimes, 209
Astroloy, 151
Nodes of displacement (DN), 27
Nodes of stress (SN), 27
Non-axial test equipment, 75 SGI52, S-N data, 104
Non-ferrous alloys, 109 SIF amplitude, 43
Nonsymmetrical test equipment, displacement and energy
55–56 approaches to, 45–49
SIF, static, 43
Slip amplitude, fretting fatigue, 76
Slip dissolution model, 227
Partial closure model SN (nodes of stress), 27
2=Pi0 method for 7055 S-N (stress=cycles) curve, 2
aluminium, 141 12% Cr steel, 234
2024T-3, 144 2080=15%SiCp composite, 107
predicted threshold corner for 304 stainless steel, 103
6061, 144 4240R, 93
Persistent slip bands (PSBs), 214–215 4240U, 92
Piezo-electric transducer, 52 7075 alloy, 107
Piezo-electric fatigue machine, 4, 73 Al-Li8090, 122
fretting system, 80 aluminium alloys
Plateau stress, 214–216 in liquid hydrogen, 223
Polycrystalline copper at low and high frequency, 210
temperature and fatigue crack carbon steels, 3
growth rates, 176 cast iron, 210
302 Index

[S-N (stress=cycles) curve,] [Steel 304]

Cr-Si steel, 97 temperature fatigue crack growth
Cr-V steel, 96 rates in, 177
D38 steel, 210 Steel 4240, S-N curves, 236
fatigue limit, 208 Steel rate, and threshold, 212–214
and modeling and, 89 Steel ratio, austenitic stainless
fatigue strength, 87–132 steel, 183
low carbon steel sheet, Steels, gigacycle fatigue of, 231
102, 232–233 Strain values, 65
Martensitic stainless steel, 239 Stress calculation, 64
N18 nickel base alloy, 240 Stress concentration, 243
N18, 124 Stress intensity factor,
rail steel, 237 three-dimensional
SGI52, 104 computational, 43–49
SGI52, 105 Stress magnification factor, 21
spheroidal graphite cast iron, 235 Stress node (SN), 29
spring steel, 238 Stress ratio influence, on T6A4V
steel 4240, 236 fatigue crack growth, 146
SUJ2 bearing steel, 125 Stress=cycles (S-N) curve,
Ti6246, 114, 120 definition of, 2
Ti-6246, 221 Striations, 148
Ti-6A4V, at room temperature Subsurface initiation,
and high pressure in 4240 steel, 93
hydrogen, 227 SUJ2 bearing steel
Ti-6A4VPQ at, 221 chemical composition, 100
titanium alloy, 241 fatigue results of, 101
typical, 3 mechanical properties of, 100
Udimet, 500, 118 S-N curve of, 125
Specimen dimensions, 31 tension-compression, 126
Specimen geometry, 32 Superalloy N18
Spheroidal graphite cast iron composition of, 123
SGI52, 103–104, 186 heat treatments of, 124
chemical composition of, 104 mechanical properties of, 124
S-N curve, 235 Surface initiation, 255
Spring steel, 94 Surface-subsurface transition, in
chemical composition of, 96 4240 steel, 253
mechanical properties of, 96
S-N curves, 238
Stainless steel 304, S-N curve, 233
Staircase method, 6 T-1 steel, threshold data on, 138
Static tensile results of Ti-Al, 129 T6A4V (Ti 6A4V)
Statistical conditions, 243 composition of, 115
Steel 304 experimental results of, 116
chemical composition, 174 fatigue crack growth, 121, 199, 213
heat treatment of, 175 stress ratio influence on, 146
mechanical properties of, 175 with striation formation, 211
Index 303

[T6A4V (Ti 6A4V)] [Ti-6246 alloy]

fretting scar on, 198 chemical composition of, 110
heat treatments of, 115 fatigue crack growth rate of, 252
mechanical properties of, 115 fatigue initiation in, 250
Temperature and fatigue crack quantitative microstructure
growth rates features of, 112
mild steel, 178 S-N curves of, 114, 120, 221
polycrystalline copper, 176 thermomechanical processes of,
steel 304, 177 111, 250–251
Temperature rise, 181 Ti-6A4V
Test control program, life curve, 228
flow diagram, 57 S-N curve, at room
Test equipment temperature=high pressure
high temperature, 67 hydrogen, 227
low temperature, 69 Ti-6A4VPQ, S-N curves, 221
non-axial, 75 Ti-Al alloys, 127–132
nonsymmetrical, 55–56 chemical composition of, 129
thin sheet, 71 high-cycle fatigue properties of, 130
variable amplitude, 55–56 static tensile results of, 129
Test frequency on fatigue crack Titanium alloys, 109, 119, 146, 249
growth in cast iron, 188 fatigue crack growth rate, 146
Testing machines, basic structure, gigacycle fatigue of, 234
52–55 heat sensitivity, 219
Thermomechanical processing, S-N curves, 241
of Ti-6246 alloys, 250–251 Torsion fatigue testing, 81
Thin sheet Transgranular rupture, 190
crack growth specimen, 73 Transverse motion, 28–30
FCG rate of, 180 Transverse waves, 12–15
test equipment, 71
test system, 74
Three-point bending, ultrasonic
fatigue experiments Udimet 500 (U500), 117
fatigue testing, 82 composition of, 115
specimen dimensions, 86 heat treatments of, 115
system for, 83 mechanical properties of, 115
Threshold behavior in 9310 fatigue data, 208
gear steel, 137 S-N data of, 118
Threshold data Udimet 720, corrosion fatigue of, 225
A533 steel, 138 Ultrasonic fatigue
T-1 steel, 138 data, industrial applications of, 11
Threshold level crack fatigue crack tests, data
growth rates, 2 acquisition, 59
Threshold region, 2 loading, crack initiation in, 117
Threshold stress intensity slow fatigue, 149
factors, 190 test, and natural frequencies, 11
Ti-6246 alloy test method, 9
304 Index

[Zones of plasticity] Vibration displacement and stress,

vibration displacement and ultrasonic fatigue and, 201
stress, 201 Vibration modes, 38
Ultrasonic frequency regime,
calculating SIF in, 31
Ultrasonic fretting fatigue testing, 75
schematic experimental system, 78 Water, fatigue limits in, 225
Ultrasonic horn, 52 Wrought aluminium alloys, 240
Ultrasonic specimen geometry
crack growth specimen, 16
endurance specimen, 16
Unreinforced alloy 6061-T6, 172
Young’s modulus. See Mechanical
properties.

Variable amplitude test equipment,

55–56
Variable section specimen, 15–20 Zones of plasticity, crack tip, 135