Anda di halaman 1dari 477

2002 ASM International. All Rights Reserved.

Handbook of Residual Stress and Deformation of Steel (#06700G)

Handbook of Residual Stress


and
Deformation of Steel

Edited by
G. Totten
M. Howes
T. Inoue

Materials Park, Ohio 44073-0002


www.asminternational.org

www.asminternational.org

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

www.asminternational.org

Copyright 2002
by
ASM International
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any
means, electronic, mechanical, photocopying, recording, or otherwise, without the written permission of the
copyright owner.
First printing, March 2002

Great care is taken in the compilation and production of this book, but it should be made clear that NO
WARRANTIES, EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE GIVEN IN CONNECTION
WITH THIS PUBLICATION. Although this information is believed to be accurate by ASM, ASM cannot
guarantee that favorable results will be obtained from the use of this publication alone. This publication is
intended for use by persons having technical skill, at their sole discretion and risk. Since the conditions of
product or material use are outside of ASMs control, ASM assumes no liability or obligation in connection
with any use of this information. No claim of any kind, whether as to products or information in this
publication, and whether or not based on negligence, shall be greater in amount than the purchase price of this
product or publication in respect of which damages are claimed. THE REMEDY HEREBY PROVIDED
SHALL BE THE EXCLUSIVE AND SOLE REMEDY OF BUYER, AND IN NO EVENT SHALL EITHER
PARTY BE LIABLE FOR SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES WHETHER OR
NOT CAUSED BY OR RESULTING FROM THE NEGLIGENCE OF SUCH PARTY. As with any material,
evaluation of the material under end-use conditions prior to specification is essential. Therefore, specific
testing under actual conditions is recommended.
Nothing contained in this book shall be construed as a grant of any right of manufacture, sale, use, or
reproduction, in connection with any method, process, apparatus, product, composition, or system, whether or
not covered by letters patent, copyright, or trademark, and nothing contained in this book shall be construed as
a defense against any alleged infringement of letters patent, copyright, or trademark, or as a defense against
liability for such infringement.
Comments, criticisms, and suggestions are invited, and should be forwarded to ASM International.
Prepared under the direction of the ASM International Technical Book Committee (20002001), Sunniva R.
Collins, Chair
ASM International staff who worked on this project included Veronica Flint, Manager of Book Acquisitions;
Bonnie Sanders, Manager of Production; Carol Terman, Production Project Manager; and Scott Henry,
Assistant Director of Reference Publications.
Library of Congress Cataloging-in-Publication Data
Handbook of residual stress and deformation of steel/[edited by] G. Totten, M. Howes, T. Inoue.
p. cm.
Includes bibliographical references and index.
1. SteelFatigueHandbooks, manuals, etc. 2. Residual stressesHandbooks, manuals, etc. I. Totten,
George E. II. Howes, Maurice A.H. III. Inoue, Tatsuo, 1939TA473 .H3215 2001
620.176dc21

2001040042
ISBN: 0-87170-729-2
SAN: 204-7586
ASM International
Materials Park, OH 44073-0002
www.asminternational.org
Printed in the United States of America

Multiple copy reprints of individual articles are available from Technical Department, ASM International.

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

www.asminternational.org

Contributors
I. Alexandru
Faculty of Materials
Technical University of Iasi
Romania

T. Inoue
Department of Energy Conversion Science
Graduate School of Energy Science
Kyoto University, Japan

H. Bhadeshia
Department of Materials Science and
Metallurgy
University of Cambridge, UK

D.-Y. Ju
Saitama Institute of Technology
Saitama, Japan

V. Bulancea
Faculty of Materials
Technical University of Iasi
Romania
V.V. Dobrivecher
Ukraine National Academy of Science
Institute of Engineering Thermophysics
Kiev, Ukraine
T. Ericsson
Linkopings Tekniska Hogskola IKP
Linkoping, Sweden
F.D. Fischer
Vorstand des Institutes fur Mechanik
Montanuniversitat Loeben
Loeben, Germany
A.V. Fomin
Leading Research Scientist
Institute Russian Academy of Sciences
Moscow, Russia
K. Funatani
Nihon Parkerizing Co. Ltd.
Nagoya, Japan
Bo Gong
Department of Metallurgy & Materials Science
University of Toronto, Canada
J. Grum
Faculty of Mechanical Engineering
University of Ljubljana, Slovenia
A.Y. Hassan
Director/Dean, School of Mechanical
Engineering
Universiti Sains Malaysia, Malaysia
K. Heess
Karl Heess GmbH
Lampertheim, Germany
F.T. Hoffmann
ITW Bremen
Bremen, Germany
R. Hoffmann
ITW Bremen
Bremen, Germany

N.I. Kobasko
Ukraine National Academy of Science
Institute of Engineering Thermophysics
Kiev, Ukraine
Z. Kolozsvary
S.C. Plasmaterm S.A.
Tg-Mures, Romania
A.I. Kovalev
Surface Phenomena Research Group
CNIICHERMET
Moscow, Russia
J. Kritzler
Metal Improvement Company, Inc.
Unna, Germany
K.-H. Lang
Institut fuf Werkstoffkunde 1
Universitat Karlsruhe (TH), Germany
R.W. Lewis
Department of Mechanical Engineering
University College of Swansea, UK
D. Lohe
Institut fuf Werkstoffkunde 1
Universitat Karlsruhe (TH), Germany
J. Lu
Universite de Technologie de Troyes
Trayes Cedex, France
T. Lubben
ITW Bremen
Bremen, Germany
M.V. Medvedev
Research Scientist
Institute Russian Academy of Sciences
Moscow, Russia
V.P. Mishina
Surface Phenomena Research Group
CNIICHERMET
Moscow, Russia
V.S. Morganyuk
Ukraine National Academy of Science
Institute of Engineering Thermophysics
Kiev, Ukraine
iii

M. Narazaki
Utsunomiya University
Tochigi, Japan
J. Pan
School of Materials Science and Engineering
Shanghai Jiao Tong University
Shanghai, P.R. China
P. Ramakrishnan
Department of Metallurgical Engineering and
Materials Science
Indian Institute of Technology
Bombay, India
I.A. Razumovsky
Leading Research Scientist
Institute Russian Academy of Sciences
Moscow, Russia
T. Reti
Banki Donat Polytechnic
Budapest, Hungary
C. Ruud
Pennsylvania State University
University Park, PA USA
G. Schleinzer
Vorstand des Institutes fur Mechanik
Montanuniversitat Leoben
Leoben, Germany
B. Scholtes
Institut fur Werkstofftechnik
Universitat Kassel, Germany
K.N. Seetharamu
School of Mechanical Engineering
Universiti Sains Malaysia, Malayasia
G.E. Totten
G.E. Totten & Associates Inc.
Stony Point, NY USA
O.Vohringer
Institut fuf Werkstoffkunde 1
Universitat Karlsruhe (TH), Germany
D.L. Wainstein
Surface Phenomena Research Group
CNIICHERMET
Moscow, Russia
H.W. Walton
Consultant
Forest City, NC USA
Z. Wang
Department of Metallurgy & Materials Science
University of Toronto, Canada

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

G.M. Webster
Union Carbide Corporation
Tarrytown, NY USA
W. Wubbenhorst
Metal Improvement Company, Inc.
Unna, Germany

www.asminternational.org

V.V. Zabilsky
Physical Technical Institute, Ural Branch of
RAS
Ijevsk, Russia

iv

W. Zinn
Institut fur Werkstofftechnik
Universitat Kassel, Germany

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

www.asminternational.org

Contents
Preface ........................................................................... vi

Residual Stress During Hardening Processes

Effect of Materials and Processing

Residual Stresses in Carburized, Carbonitrided,


and Case-Hardened Components ........................................ 189
T. Reti

Material Factors .................................................................. 3


H.K.D.H. Bhadeshia

Residual Stresses in Nitriding............................................... 209


Z. Kolozsvary

Prestress Engineering of Structural Material:


A Global Design Approach to the Residual Stress Problem ..........11
J. Lu

Induction Hardening.......................................................... 220


J. Grum

Residual Stresses and Fatigue Behavior.....................................27


D. Lohe, K.-H. Lang, and O. Vohringer

Hardening by Reheating and Quenching .................................. 248


M. Narazaki, G.E. Totten, and G.M. Webster

Stability of Residual Stresses .................................................54


D. Lohe and O. Vohringer

Metallo-Thermo-MechanicsApplication to Quenching ................ 296


T. Inoue
Control of Residual Stress Formation and Steel Deformation
during Rapid Heating and Cooling...................................... 312
N.I. Kobasko, V.S. Morganyuk and V.V. Dobrivecher

Effect of Residual Stress on Hydrogen Embrittlement


and Stress Corrosion Cracking.............................................70
A.I. Kovalev, V.P. Mishina, D.L. Wainstein, and V.V. Zabilsky

Effect of Cryogenic Cooling on Residual Stresses, Structure,


and Substructure ........................................................... 331
Ioan Alexandru and Vasile Bulancea

Measurement and Prediction of Residual Stress and Distortion


Deflection Methods to Estimate Residual Stress ...........................89
H. Walton

Inducing Compressive Stresses through Controlled Shot Peening .... 345


J. Kritzler and W. Wubbenhorst

Measurement of Residual Stresses ...........................................99


C. Ruud

Residual Stress Formation During Manufacturing Processes


Residual Stress Formation during Casting ................................ 361
R.W. Lewis, K.N. Seetharamu and A.Y. Hassan

Stress Determination in Coatings........................................... 118


J.Albert Sue and Gary S. Schajer

Residual Stress Formation during Casting:


Continuous and Centrifugal Casting Processes........................ 372
D.-Y. Ju

Methods for Determination of Inhomogeneous Residual


Stress Fields ................................................................ 125
I.A. Razumovsky, M.V. Medvedev, and A.V. Fomin
Residual Stress Formation in the Shaping of Materials

Residual Stress Formation Processes during Welding and Joining.... 391


W. Zinn and B. Scholtes

Residual Stress in the Forming of Materials.............................. 141


Z. Wang and B. Gong

Residual Stresses in Powder-Metal Processing........................... 397


P. Ramakrishnan
Residual Stress Formation and Distortion of Rail Steel................. 424
F.D. Fischer and G. Schleinzer

The Effect of Final Shaping Prior to Heat Treatment ................... 150


T. Ericsson

Residual Stresses during Gear Manufacture .............................. 437


K. Funatani

Factors Affecting Final Part Shaping ...................................... 159


P. Jiansheng

Metric Conversion Guide .................................................. 459


Effects of Process Equipment Design...................................... 183
F.T. Hoffmann, T. Lubben, R. Hoffmann, and K. Hee

Index ........................................................................... 465


v

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

www.asminternational.org

Preface
Control of steel deformation is one of the most
common concerns within the metals processing
industry. Numerous surveys have been conducted by various organizations in recent years
to assess the critical needs of the industry. In
nearly every survey that has been conducted, distortion is either the greatest or second greatest
concern among the steel heat treating community. Steel distortion control will exhibit tremendous effects on the profitability of the commercial enterprise. Therefore, it is not surprising that
the ability to understand the overall distortion
process and to be able to design solutions to this
problem typically rank very high on these same
surveys.
In view of the enormous visibility and importance of steel deformation problems, the editors
decided to put together an engineering handbook
on steel deformation. To address this subject
properly, contributing factors to overall steel deformation problems, including material effects,
machining, heating and cooling, must be examined.
This handbook contains 27 articles, divided
into five sections: Effect of Materials and Processing, Measurement and Prediction of Residual Stress and Distortion, Residual Stress Formation in the Shaping of Materials, Residual
Stress During Hardening Processes, and Residual Stress Formation During Manufacturing Processes.
There are five articles in the section Effect of
Materials and Processing. Material Factors
discusses the effects of various material properties such as thermal properties and the interactions of residual stresses on the transformation
products formed and steel deformation during
fabrication. Transformation plasticity is discussed in some detail along with the use of modeling to better understand these processes.
Prestress Engineering of Structural Material
provides a global design approach to understanding the effects of residual stress generated during
surface engineering manufacturing processes
such as PVD and CVD on the material properties
obtained. Some of the topics discussed in this
chapter include developments in the measurements of residual stresses, advanced mechanical
surface treatments, and modeling of fatigue behavior taking residual stresses into consideration.
The effect of residual stresses on fatigue behavior is discussed in detail in the next article.
Examples of topics discussed include stability of
residual stresses, some aspects of fatigue in

steels, influence of residual stresses on cyclic deformation behavior, influence of residual stresses
on crack initiation and propagation, and effect of
residual stresses on S-N curves; an overview of
modeling of the effect of residual stresses on fatigue behavior is provided.
The next article provides an overview of the
stability and relaxation behavior of macro and
micro residual stresses in steel due to thermal
and mechanical treatments. This discussion includes relaxation of residual stresses by annealing, residual stress relaxation by uniaxial deformation, and relaxation by cyclic deformation.
Hydrogen embrittlement of metals, as well as
other types of brittle fracture, result from nucleation and development of micro-cracks caused
by internal stresses. The last article in this section provides an overview of the effect of residual stress on hydrogen embrittlement and stress
corrosion cracking (SCC) of steel. This discussion includes the effect of hydrogen on structure
and transformation of steel, types of hydrogen
embrittlement, delayed fracture in steel, crack
initiation and growth, SCC of low alloy steels,
crack initiation and growth mechanism of SCC
processes, methods of estimating sensitivity to
SCC, effect of alloying elements on resistance
to SCC, and the role of structure and thermal
processing in SCC.
In the section Measurement and Prediction of
Residual Stress and Distortion, the first article
describes a number of simple, inexpensive deflection (dissection) methods used to estimate residual stress of various types of components. The
methods include Almen strip; Navy C-Ring;
plate or bar slitting and deflection; tube slitting
and opening; and bending of bars, H-beams, and
channels.
The next article provides an overview of residual stress measurement methods. Topics include residual stresses arising from various
manufacturing processes, measurement methods
including strain measuring technique, post-stress
relaxation measurement, sectioning and material
removal methods. In addition, strain measurement methods such as x-ray and neutron diffraction, ultrasonic, birefringent and laser, optical
gages, brittle coatings, Barkhausen noise, and
chemical coatings are discussed. Semidestructive methods such as blind hole drilling and ring
coring are discussed.
Measurement of residual stresses in coatings
and thin films is important because their influence on mechanical and physical properties affect component service performance. Stress Devi

termination in Coatings provides a guide for


measuring residual macrostress in coatings, Specific topics include origin of residual stresses in
coatings and residual stress measurement methods including the deflection method, x-ray diffraction, and hole-drilling. A comparison of
these methods is provided.
The last article in this section provides a detailed review of methods used to measure and
subsequent data analysis of inhomogeneous residual stress fields. This discussion includes residual stress as an inverse problem of experimental mechanics, indicator crack method of
measuring residual stress, arbitrary cut-out indicator method, and experimental methods and
equipment including photoelastic coating
method, and optical interferometry. Although
this is a relatively rigorous numerical discussion,
practical examples also are provided.
Residual Stress Formation in the Shaping of
Materials contains four articles. The first article
covers residual stress in the steel forming processes. The steel forming processes included are
cold forming such as wire drawing, and hot
forming such as extrusion, rolling, and forging.
The effects of residual stresses involved in these
processes are reviewed, and specific topics include residual stress in cold metal forming such
as bending of sheet, drawing of wire, rod, and
tube, and residual stresses in deep drawn cup,
sunk tubes, and radial forging products.
The effect of final shaping prior to heat treatment on residual stress formation is discussed in
the next article. The effects of shaping processes
including grinding, milling, turning, shot peening, and straightening on residual stress are discussed. Also discussed is distortion after final
part shaping and experimental and computational studies of these processes.
The next article provides a practical overview
of the factors affecting residual stress and distortion during final part shaping. Included are
discussions of influence of component shape on
heat treatment distortion, the effect of crosssection size and asymmetry, effect of heat treating procedure and machining process on final
component shape, effect of sequence of heat
treating and machining, influence of machining
allowance and stress relieving procedure, influence of residual stresses caused by cutting, methods of manufacturing blanks and effect of original structure, hot-rolled steels or forgings and
effect of banded segregation and carbide segregation, influence of heat treating methods, the
effect of heating including the rules of heating,

2002 ASM International. All Rights Reserved.


Handbook of Residual Stress and Deformation of Steel (#06700G)

quenching and system design, tempering, and


equipment and racking.
A more focused, but practical, discussion on
the effect of process equipment design on distortion follows. Subjects that are covered include
distortion generating process equipment, methods that may be used to minimize equipmentrelated distortion, quench system design, and
press quenching.
Residual Stress During Hardening Processes
contains eight articles. The first article provides
a detailed discussion on the residual stresses in
carburized, carbonitrided, and case-hardened
components. Topics include process considerations for carburized and carbonitrided components, transformations and stress evolution in
carburized and case-hardened components, effect of heat treating operations on residual stress
distribution, relationship between residual
stresses and properties of carburized parts and
modeling and prediction of residual stress field.
The article on residual stresses in steel nitriding includes a discussion of nitrided layer structure as a function of nitriding process, residual
stresses in nitrided layers, influence of residual
stresses on fatigue behavior of nitrided steel
components, and modeling and prediction of residual stresses in nitrided steel components.
The article on residual stress formation in induction hardening processes include an overview of the induction hardening process and
steels used for this process, magnetic flux concentrators, conditions in induction heating and
quenching of machine parts, residual stress surface profiles after induction surface hardening,
stress profiles in the machine part in the loaded
state, workpiece distortion in induction surface
hardening, induction surface hardening of gear
wheels, fatigue strength of materials, and residual stresses after induction surface hardening and
finish grinding.
The next article provides an overview of residual stresses and distortion resulting from reheating and quenching. Topics include phase
transformation during heat treating including
steel transformations, TTT and CCT diagrams,
metallurgical crystal structure, estimation of volumetric change due to steel transformation upon
quenching, cooling of steel with and without
metallurgical transformation, tempering, basic
distortion mechanism, relief of residual stresses,
material movement due to thermal gradients during heating and cooling, material, component
and process effects, retained austenite, quench
severity and uniformity and process design effects on distortion, quench distortion and cracking, quenchant selection, measurement and evaluation of quenching power, estimation of heat
transfer coefficient, wetting behavior and nonuniform quenching, surface conditions, and
quench process modeling and simulation of residual stress and distortion after quenching.
A detailed approach to modeling and simulation of residual stress and distortion applied to
quench processing follows. This discussion is
based on a metallo-thermo-mechanics approach,
and topics discussed include an overview of

www.asminternational.org

metallo-thermo-mechanics and numerical simulation methodology with practical examples.


In the article on the control of residual stress
formation and steel deformation during rapid
heating and cooling, a particular emphasis is on
intensive quenching. This is the first detailed,
article-length discussion of this old, but littleknown technology in the western world. Topics
include mathematical model for calculation of
thermal and stress-strain state, computation of
stress-strain state, possibility of predicting hardening cracks, predicting the deformation of bearing rings during hardening, thermal stresses
formed in carburized steel products due to excessive cooling rates, generalization of computational and experimental results for heating and
cooling of parts with different geometries and
thermal and physical fundamentals of processing
of high-strength materials.
An often contradictory subject is the cryogenic processing of steels, and the detailed overview of the effect of cryogenic cooling on residual stress is presented here. Specific topics
include role of residual stresses within martensitic transformation at cryogenic temperatures,
evaluation of residual stresses after cryogenic
cooling, influence of cryogenic cooling on residual stresses and dimensional stability of steels,
and influence of cryogenic cooling on the structure and substructure of steels.
The practical use of controlled shot peening
to induce compressive residual stresses is described in detail next. This discussion includes a
historic overview of shot peening, elementary
processes of shot peening, workpiece and material process parameters, process monitoring,
process optimization, x-ray diffraction, and industrial examples.
In Residual Stress Formation During Manufacturing Processes, the first article includes an
extensive discussion of residual stress and deformation problems arising from the casting process, and modeling of residual stress formation
during casting. Discussion includes finite element analysis of heat flow during casting, formulation of the elasto-viscoplastic stress model,
and deformation of a solidifying material.
The next article describes residual stress formation during the casting process, and it includes continuous and centrifugal casting. Topics discussed include inelastic behavior and
unified constitutive theory of metallic material
in solidification, analytical method of the
thermal-mechanical problem for the casting process, residual stress formation during semicontinuous casting, residual stress formation during
centrifugal casting, and residual stress formation
during strip casting by the twin-roll method.
The origin and assessment of residual stresses
during welding or brazing is discussed next.
Welding residual stresses are discussed including residual stresses due to shrinkage, quenching, and phase transformations. Characteristic
residual stress distributions in brazed components is also discussed.
The article Residual Stresses in Powder
Metal Processing is divided into two parts. The
vii

first part describes manufacturing of ferrous P/


M parts including powder characteristics, compaction in rigid dies, isostatic compaction, sintering, heat treatment of P/M parts, hot pressing,
roll compaction, powder forging, metal injection
molding, spray forming, warm compaction, and
rapid prototyping. The second part discusses residual stresses in P/M processing including
powder production, compaction of metal powders, sintering of metal powders, pressure sintering and hot isostatic pressing, heat treatment
of P/M parts, and microstructural development
and properties.
Residual Stress Formation and Distortion of
Rail Steel covers the cooling process including
the cooling boundary conditions and heat transfer, residual stress state analysis, weight and frictionthe rail end problem, experimental results;
roller straightening including residual stresses in
unused roller-straightened rails, behavior of rail
steel under plastic deformation, simulation of
roller straightening; and rails in service including residual stresses due to welding and residual
stress formation in rolling contact.
The last article provides a detailed description
of residual stress formation during hypoid gear
manufacture. It includes an overview of residual
stress formation in carburized and hardened
work, profiles and peak magnitudes of residual
stresses, measurement methods including the
Sachs hole-drilling method, x-ray and neutron
diffraction, influence of steel properties on residual stresses, influence of carburizing process parameters on residual stress formation, benefits of
residual stresses on fatigue strength, and the effects of hardness, case depth, intergranular oxidation, influence of shot peening, change of residual stresses during fatigue, and distortion of
carburized and hardened steels.
The preparation of a text of this scope was a
tremendous task. The editors are deeply indebted
to many colleagues for their patience, support,
and assistance; without them this text would not
have been possible. Special thanks go to the
ASM staff who often labor in the background
but who are vital members of the team. Particularly, thanks go to Veronica Flint and Carol Terman of ASM International for their help and encouragement.
Very special thanks go to our families for their
seemingly unending support. Without their understanding and encouragement, this project
would never have been completed.
George E. Totten, Ph.D., FASM
Editor
G.E. Totten & Associates Inc.
Stony Point, NY USA
Prof. Maurice A.H. Howes, Ph.D. (Retired)
Editor
Worcestershire, England
Prof. Tatsuo Inoue, Ph.D., FASM
Editor
Department of Energy Conversion Science
Faculty of Energy Science
Kyoto University
Kyoto, Japan

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p3-10
DOI: 10.1361/hrsd2002p003

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Material Factors
H.K.D.H. Bhadeshia, University of Cambridge

RESIDUAL STRESSES are a consequence of


interactions among time, temperature, deformation, and microstructure (Fig. 1). Material or
material-related characteristics that influence the
development of residual stress include thermal
conductivity, heat capacity, thermal expansivity,
elastic modulus and Poissons ratio, plasticity,
thermodynamics and kinetics of transformations,
mechanisms of transformations, and transformation plasticity.
Many general statements can be made about
the role of material factors in the evolution of
residual stress. Spatial variations in temperature
give rise to nonuniform thermal strains, the effect of which becomes exaggerated when the
material is elastically stiff and has a high yield
strength. A large thermal conductivity helps reduce residual stress by reducing temperature gradients (Ref 2). The dissipation or absorption of
heat depends not only on the external environment of the component but also on internally
generated heatfor example, during adiabatic
deformation or due to the latent heat of transformation. Similarly, the plastic strain distribution
in the component depends both on the constitutive properties and on how the shape deformations due to phase transformations compensate
for the development of stress.
The fundamental material properties are, of
course, temperature dependent. Table 1 illustrates how several key properties might vary
with temperature (Ref 3). Some of these properties, which can to some extent be estimated
quantitatively, are discussed in detail in the sec-

tions that follow; others such as elastic modulus


and thermal conductivity still have to be measured for individual alloys.

Heat Capacity
The dominant contribution to specific heat capacity comes from lattice vibrations (phonons),
since the majority of free electrons are prevented
from participation in heat absorption by the Pauli
exclusion principle. However, for iron and its
alloys, a further important contribution comes
from magnetic changes. The net specific heat capacity can therefore be factorized into three components:

T C

CP{T } C LV

TD

CeT C Pl{T }

where C LV{TD /T} is the Debye specific heat function and TD is the Debye temperature. The funcTable 1

Physical properties that affect the development of residual stress in steels


Temperature, C (F)

Property

Phase(a)

Elastic modulus, GPa

c
P
b

Poisson ratio

c
P
b

c
P
b

Thermal expansivity, K1

Time,
temperature

1. Thermal stress

6. Latent
heat

Stress,
strain

4. Heat of
deformation
5. Stressinduced
transformation

2. Temperaturedependent phase
transformations

3. Transformation
strain
Microstructure

Fig. 1

The coupling of temperature, stress, and microstructure. Source: Ref 1

(Eq 1)

tion C1 corrects C LV{TD /T} to a specific heat at


constant pressure. Ce is the electronic specific
heat coefficient, and C lP is the component of the
specific heat capacity due to magnetism. Figure
2 illustrates the data for ferrite and austenite in
pure iron. Whereas it is well known that ferrite
undergoes a paramagnetic to ferromagnetic transition on cooling below 1042.15 K, the magnetic
properties of austenite are seen from Fig. 2 to be
of some consequence in determining the heat capacity. There are two coexisting electron states
of austenite, one of which is ferromagnetic with
a Curie temperature of 1800 K and the other of
which is antiferromagnetic with a Neel temperature of 55 to 80 K (Ref 4). The balance between
these states changes with temperature, giving
rise to corresponding changes in heat capacity.
The data in Fig. 2 are for pure iron, but there
is now sufficient understanding of the components of heat capacity to enable similar estimates
for iron alloys, using internationally available
computer programs and thermodynamic data-

Thermal conductivity, W/m K

c
P
b

Specific heat capacity, 106 J/m3 K

c
P
b

Yield strength, MPa

c
P
b

0 (32)

200
210
210
200
0.291
0.280
0.280
0.280

15.0
49.0
49.0
43.1
4.15
3.78
3.78
3.76
190
360
440
1600

300 (570)

175
193
193
185

600 (1110)

150
165
165
168

0.309
0.296
0.296
0.296
2.1 105
1.4 105
1.4 105
1.3 105
18.0
41.7
41.7
36.7
4.40
4.46
4.46
4.45
110
230
330
1480

(a) , P, b, and represent allotriomorphic ferrite, pearlite, bainite, and martensite, respectively. Source: Ref 3

0.327
0.310
0.310
0.310

21.7
34.3
34.3
30.1
4.67
5.09
5.09
5.07
30
140
140
1260

800 (1470)

124
120
120
...
0.345
0.325
0.325
...

25.1
27.0
27.0
...
4.90
5.74
5.74
...
20
30
30
...

4 / Effect of Materials and Processing


bases (Ref 6). After all, changes in fundamental
thermodynamic quantities such as enthalpy and
entropy are derived from heat capacity data. It is
surprising that this capability has not yet been
exploited in any calculation of residual stress,
even though the methodology is widely available.

Expansion Coefficient and Density


Table 1 shows that the expansion coefficient
of austenite is larger than that of ferrite; this
might be considered surprising given the lower
density of ferrite. However, the behavior is again

c0
c1
V cm{T } (1 y)V m
{T } y V m
{T }

30
Heat capacity, CP, J mol 1K1

V cm0 6.695(1 2.043 105T 1.52 108T 2)


V cm1 7.216(1 2.043 105T 1.52 108T 2)

40

20

V am{T }
7.061(1 2.043 105T 1.52 108T 2)

Austenite
10
0
75
Ferrite

50

25
0

a reflection of the two coexisting electronic


states of austenite (c0 and c1), each with a thermal expansion coefficient that is identical to that
of ferrite. The c0 component has the lower molar
volume and is the antiferromagnetic form,
whereas the denser c1 form is ferromagnetic. The
relative proportion of atoms in the c0 and c1
states changes with temperature, so that the apparent expansion coefficient of austenite as a
whole, as detected experimentally, is much
larger than that of ferrite (Fig. 3).
The molar volumes (in cm3 /mol) of c0, c1, c,
and over the temperature range of 300 to 1775
K are:

300

600

900

1200

1500 1800

Temperature, K

where y is the fraction of atoms of austenite in


the c1 state, the details of which can be found
elsewhere (Ref 4, 5).
These data are for pure iron, but thermodynamic data can be used to assess how the expansion coefficients would change with alloying,
since there are quite sophisticated treatments of
the effect of solute elements on the magnetic and
other components of the free energies of iron.
Note that the two electronic states picture of
austenite is a simplification of the real scenario,
but first-principles calculations (Ref 7), which

Specific heat capacities of ferrite and austenite in


pure iron, as a function of temperature. The thin
lines represent the combined contributions of the phonons
and electrons, whereas the thicker lines also include the
magnetic terms. The dashed vertical lines represent the Curie, c, and c d transitions. d-ferrite is simply an
alternative historical name for high temperature . Source:
Ref 5

deal with higher levels of complexity, are not yet


applicable to practical alloys.

Plastic Deformation
The familiar mechanisms of plastic deformation are slip, mechanical twinning, and creep.
Phase transformations also cause permanent deformation (Ref 811). In steels, austenite can decompose into a large variety of microstructures
that are distinguished by the atomic mechanism
of transformation (Fig. 4). In a displacive transformation, the change in crystal structure is
achieved by a deformation of the parent structure. A reconstructive transformation is one in
which the change in structure is achieved by a
flow of matter, which occurs in such a way that
strains are minimized.
All the transformations cause changes in
shape (Fig. 5a), which for reconstructive transformations simply reflects the change in density.
For displacive transformations, the shape change
is an invariant-plane strain (IPS), that is, a combination of a shear on the invariant plane and a
dilatation normal to that plane. The strain energy
associated with a constrained IPS is minimized
when the product phase has a thin-plate shape.
This is why Widmanstatten ferrite, bainite, acicular ferrite, and martensite in steels grow in the
form of plates. The distinguishing features of a
variety of deformation modes are compared in
Table 2, and Table 3 describes the shape deformations.
The permanent strain caused by any transformation is called transformation plasticity. A

Fig. 2

Displacive
Reconstructive

Invariant-plane strain shape


deformation with large shear
component. No iron or
substitutional solute diffusion.
Thin plate shape

Diffusion of all atoms during


nucleation and growth.
Sluggish below about 850 K

8.0

7.8

Allotriomorphic
ferrite

Volume, cm3 mol1

7.6

V1

7.4

Widmansttten ferrite
Carbon diffusion during
paraequilibrium nucleation
and growth

Idiomorphic
ferrite

Bainite and acicular


ferrite

7.2

Massive ferrite

7.0

No change in
bulk composition

Carbon diffusion during


paraequilibrium nucleation.
No diffusion during growth

6.8

V0
6.6

300

600

900

1200

1500 1800

Temperature, K

Fig. 3

Molar volumes of the various forms of iron.


Source: Ref 5

Fig. 4

Pearlite

Martensite

Cooperative growth
of ferrite and
cementite

Diffusionless
nucleation and
growth

Transformation products of austenite. Source: Ref 12

Material Factors / 5
phase change in a stress-free material is usually
triggered by heat treatment, when the parent
phase passes through an equilibrium transformation temperature. Alternatively, the application of a stress in isothermal conditions can trigger transformation in circumstances where it
would not otherwise occur. Unusual effects can
occur when stress and temperature work together. The transformation may occur at remarkably low stresses or at very small deviations
from the equilibrium temperature. This is why
even minute stresses can greatly influence the
development of microstructure, and vice versa.
It is not surprising that transformation plasticity
can be obtained at stresses that are much smaller
than the conventional yield stress of the parent
phase.

Transformations, Residual Stresses,


and Related Phenomena
The strains due to phase transformations can
alter the state of residual stress or strain. It is well
known that the martensitic transformation of the
carburized surface of a steel component puts the
surface under compression. It is argued that this
is because of the expansion at the surface due to
formation of the lower-density martensite from
austenite.

Phase transformation can also compensate for


stress. Greenwood and Johnson (Ref 13, 14)
showed that when a phase change is accompanied by a change in volume, the tensile strain
expected when transformation occurs under the
influence of a tensile stress r is given by:
e

5 DV r
6 V rY

(Eq 2)

where rY is the yield stress of the weaker phase


and DV/V is the transformation volume strain.
The role of shear strains associated with transformation has been emphasized in later work by
Magee and Paxton (Ref 15, 16), and subsequently by Fischer (Ref 17), Leblond et al. (Ref
1822), Olson (Ref 23), and Bhadeshia et al.
(Ref 24). Not only does transformation affect
stress, but the latter modifies the development of
microstructure. The microstructure tends to be
more organized when transformation occurs in a
stresss parent phase, because the stress favors
the formation of certain orientations relative to
others. This is illustrated schematically in Fig.
5(b) to (d). These aspects will now be discussed
in more detail, because transformation plasticity
can radically alter the state of residual stress.

Table 2

Reconstructive

Displacive

Single
crystal

Causes permanent change in shape


Invariant-plane strain shape changewith a large shear component
Changes crystallographic orientation
Changes lattice type
Can lead to a density change

Table 3

Allotriomorphic ferrite
Idiomorphic ferrite
Pearlite
Widmanstatten ferrite
Bainite
Acicular ferrite
Martensite
Cementite plates
Mechanical twins ()
Annealing twins (c)

Slip
deformation

Mechanical
twinning

Displacive
transformation

Reconstructive
transformation

Yes
Yes
No
No
No

Yes
Yes
Yes
No
No

Yes
Yes
Yes
Yes
Yes

Yes
No
Yes
Yes
Yes

Shape change due to transformation

Transformation

(a)

Displacive transformations can be regarded as


modes of plastic deformation. Just as a combination of a plane and a direction constitutes a
deformation system for slip or twinning, the
habit plane and displacement vector of the
invariant-plane strain accompanying displacive
transformation completely describe the deformation system responsible for transformation
plasticity. The displacement vector describes the
sense of the macroscopic displacements accompanying transformation and, along with the habit
plane indices, also contains information about
the magnitude of the shear component and dilatational component of the displacements. Typical data for the deformation systems associated
with transformations are listed in Table 4. Note
that reconstructive transformations involve only
a volume change together with diffusional mass
flow, so it is not appropriate to regard them as
deformation systems in the present context.
Given the cubic crystal structure, and the fact
that habit planes tend to be irrational, there will
in general be 24 of these systems per austenite
grain, and they may operate simultaneously to
varying extents. Of course, unlike ordinary slip,

Characteristics of different modes of deformation

Characteristic

Deformation System

Shape change (a)

s(b)

d(b)

Morphology

Volume change
Volume change
Volume change
Invariant-plane strain
Invariant-plane strain
Invariant-plane strain
Invariant-plane strain
Invariant-plane strain?
Invariant-plane strain

0.00
0.00
0.00
0.36
0.22
0.22
0.24
0.21?
1/2
0.00

0.02
0.02
0.03
0.03
0.03
0.03
0.03
0.16?
0.00
0.00

Irregular
Equiaxed, faceted
Spherical colonies
Thin plates
Thin plates
Thin plates
Thin plates
Thin plates
Thin plates
Faceted

(a) An invariant-plane strain here implies a large shear component as well as a dilatational strain normal to the habit plane. (b) s and d refer to the
shear and dilatational strains, respectively. The values stated are approximate and will vary slightly as a function of lattice parameters and the details
of crystallography.

(b)

(c)

(d)

Shape changes accompanying unconstrained


transformations. Note that the horizontal scale
bars are all the same length. (a) The two kinds of shape
changes that occur when a single crystal of austenite transforms to a single crystal of ferrite, as a function of the mechanism of transformation. (b) Polycrystalline sample of austenite. (c) Polycrystalline sample of austenite that has
partially transformed by a displacive transformation mechanism into a random set of ferrite plates. (d) Polycrystalline
sample of austenite that has partially transformed by a displacive transformation mechanism into an organized set of
ferrite plates.

Fig. 5

Table 4

Deformation systems associated with transformations

Phase

Habit plane indices

Displacement vector

(0.363 0.854 0.373)

[0.195 0.607 0.771]

0.185

Bainite

(0.325 0.778 0.537)

[0.159 0.510 0.845]

0.27

Widmanstatten ferrite

(0.506 0.452 0.735)

[0.867 0.414 0.277]

0.36

Martensite

Note: Typical habit plane and displacement directions for low-alloy steels. The indices all refer to the austenite phase. Note that the indices stated are
approximate, since the habit plane and displacement direction are usually irrational. The displacement vector does not quite lie in the habit plane
because the dilatational strain is directed normal to the habit plane. The magnitude of the displacement is given by m, which is the total displacement
including the shear and the dilatational components.

6 / Effect of Materials and Processing


the different deformation systems within an austenite grain cannot intersect, except in special
circumstances where intervariant transformations are possible, as is the case with some
shape-memory alloys. It follows that the ordiN
A (applied tensile stress)

max

Resolution of the applied stress, rA. The normal


stress, rN, and the shear stress, s, both act on the
habit plane. The vector d is the direction along which lie
the shear displacements of the shape deformation. smax is
the maximum shear stress on the habit plane, but s is given
by resolving smax along d. Note that d differs slightly from
the displacement vector of the IPS, which includes a dilatational component in addition to the shear.

Fig. 6

Table 5 Typical values of the mechanical


driving force coefficients
DG/r, J/(mol MPa)

Nature of stress

Uniaxial tension
Uniaxial compression
Elastic crack tip (a)

0.86
0.58
1.42

Tensile
+
Compression
0

Mechanical Driving Force


The interaction of an applied elastic stress
with a phase change can occur in two ways:
1. The stress can alter the driving force for the
transformation.
2. The stress can change the appearance of the
microstructure by favoring the formation of
those variants which best comply with the applied stress.
For reconstructive transformations, only the
hydrostatic component of stress can interact with


Hydrostatic
compression
Stress (below austenite yield strength)

Indication of how the transformation-start temperature (for Widmanstatten ferrite, bainite, acicular ferrite, or martensite) should vary as a function of the
nature and magnitude of an applied stress whose magnitude is less than that of the yield stress.

Fig. 7

Table 6 Sensitivity of transformation-start


temperatures in steels to applied stress
Phase

Nature of stress

Martensite
Bainite
Eutectoid
Martensite

Pressure
Pressure
Pressure
Tensile

Source: Ref 32

Sensitivity, K/MPa

0.06
0.09
0.011
0.06

1000
Free energy, J mol1

Change in bainite-start temperature

(a) The stress state for the crack tip is multiaxial, but the coefficient is
calculated by expressing the stress in terms of the von Mises equivalent
tensile stress. Source: Ref 32

nary notion of work hardening does not apply.


Work hardening nevertheless manifests itself via
a different mechanism, in which the stability of
the austenite increases as it becomes ever more
finely divided.
The Taylor/von Mises criterion (Ref 25, 26)
states that in any given crystal, a minimum of
five independent slip systems is necessary to
produce an arbitrary shape change. A crystal in
a polycrystalline aggregate has to accommodate
the arbitrary deformations of neighboring grains.
Therefore, a polycrystalline material is brittle
unless each grain contains at least five independent slip systems. Similar logic can be applied
to the crystallographic variants of a phase generated by displacive transformation. The habit
plane is predicted theoretically (Ref 27, 28) and
found experimentally (Ref 29) to have irrational
indices. This means that there exist, in principle,
24 possible variants of the habit plane per grain
of austenite (that is, 24 independent deformation
systems). Given this large number of transformation variants available per grain, the Taylor
criterion leads to the conclusion that transformation plasticity can cause, or accommodate,
any externally imposed, arbitrary shape
changeassuming that a sufficient quantity of
parent phase is available. It follows that polycrystalline samples can remain intact at grain
boundaries when transformation plasticity is the
sole mode of deformation.

the volume change. The corresponding interaction with displacive transformations is much
larger because of the shear component of the
IPS.
For displacive transformations, the influence
of stress on the transformation can be expressed
as a mechanical driving force (DGmech), which is
the work done by the external stress in producing
the macroscopic shape deformation (Ref 30, 31):
DGmech rNd ss

(Eq 3)

where rN is the normal stress on the habit plane


and s is the component of the shear stress on the
habit plane that is parallel to the direction along
which the shear displacements of the shape deformation occur (Fig. 6). The strains d and s are
the dilatational and shear components, respectively, of the shape deformation. Some typical
values of the mechanical driving force terms are
given in Table 5. Given a free choice of some
12 to 24 crystallographic variants of the transformation product in each grain of austenite, the
work done by the shear stress is always expected
to be positive, whereas that due to the dilatational component depends on the sign of rN. For
steels, this latter component is relatively small.
Any observed consequences of stress must therefore reflect the dominant role of the shear component unless the stress is purely hydrostatic.
Since the shear stress remains positive irrespective of whether the sample is pulled in tension or uniaxially compressed, and since the
shear component of the shape change is large, a
uniaxial stress will always cause a temperature
increase for displacive transformations in steels.
Hydrostatic stress, on the other hand, has no deviatoric components and consequently interacts
only with the dilatational component of the
shape change. Thus, hydrostatic compression is
expected and found to lead to a decrease in the
transformation temperature (Fig. 7); some data
(Ref 32) on the sensitivity of the transformation
temperature to applied stress are presented in Table 6.

Limits to Stress-Assisted
Transformation

Chemical
0 Mechanical

1000
Total
2000
300

400

500

600

700

800

900

Temperature, C
Typical magnitudes of the chemical and mechanical driving forces for stress-affected transformation. The mechanical driving force is estimated for an
applied stress that is equal to the yield stress of austenite.
Since this yield stress becomes small at high temperatures,
the contribution of the mechanical driving force also decreases. Therefore, transformation becomes impossible as
the temperature exceeds about 700 C (1290 F).

Fig. 8

At temperatures close to that at which the


equilibrium transformation occurs, an applied
stress can assist reaction when the chemical driving force is insufficient to achieve the change on
its own. There must exist a point, however, when
the applied stress simply cannot provide enough
mechanical driving force to complement the
chemical term to give a driving force large
enough to induce transformation. After all, the
magnitude of the stress that can be applied is
limited by the yield point of the parent phase.
Thus, there are limits to what can be achieved
by the application of stress as a stimulus to transformation (Fig. 8).

Material Factors / 7

Residual stresses are often introduced unintentionally during fabricationfor example,


during welding or heat treatment. A few elegant
experiments illustrate how phase transformations interact with the buildup of residual stress.
Using bainitic, martensitic, and stable austenitic steels, Jones and Alberry (Ref 33, 34) demostrated that transformation plasticity during
the cooling of a uniaxially constrained sample
from the austenite phase field acts to relieve the
buildup of thermal stress as the sample cools. By
contrast, the nontransforming austenitic steel exhibited a continuous increase in residual stress
with decreasing temperature, as might be expected from the thermal contraction of a constrained sample.
When the steels were transformed to bainite
or martensite, the transformation strain compen-

500
AISI 316
9CrMo
2CrMo
Austenite YS

400
Stress, MPa

300
200
100
0
100

200

400

600

800

1000 1200 1400

Temperature, C
(a)

Stress

Plastic
strain in
austenite
Transformation
finished

Stress due to
thermal
contraction
of austenite

Stress due to
thermal contraction
of ferrite

Transformation
begins

sated for any thermal contraction strains that


arose during cooling. Significant residual
stresses were therefore found to build up only
after transformation was completed and the specimens approached ambient temperature (Fig. 9).
The experiments contain other revealing features. The thermal expansion coefficient of austenite (1.8 106 /K) is much larger than that
of ferrite (1.18 106 /K), and yet the slope of
the line prior to transformation is smaller when
compared with that after transformation is complete (Fig. 9). This is because the austenite yields
to accommodate the thermal contraction, which
is possible because the yield strength of the austenite is reduced at elevated temperatures. Ferrite
is strong at low temperatures, so the slope of the
stress/temperature curve (after transformation is
complete) is steeper and consistent with the magnitude of thermal contraction strains.
Interpretation of experimental data of the kind
illustrated in Fig. 9 is difficult in the region of
the stress/temperature curve where transformation occurs. The popular view that the volume
change due to transformation is the major component of transformation plasticity is probably
incorrect for displacive transformations such as
bainite or martensite. The shape change due to
transformation has a shear component that is
much larger than the dilatational term (Table 3).
Admittedly, this shear component should, on average, cancel out in a fine-grained polycrystalline sample containing plates in many orientations (Fig. 5c). However, the very nature of the
stress effect is to favor the formation of selected
variants, in which case the shear component rapidly begins to dominate the transformation plasticity (Fig. 5d).
Bulk transformation strain, %

Transformation under Constraint:


Residual Stress

3
75 MPa compressive stress
2

1
0

2

100

200

300

400

500

600

Transformation time, s
(b)

Temperature

(a) Plot of residual stress versus temperature for


a martensitic (9CrMo), bainitic (2CrMo), and austenitic steel (AISI 316). Adapted from Ref 33, 34. (b) Interpretation of the Jones and Alberry experiments. The thermal
expansion coefficient of austenite is much larger than that
of ferrite.

Fig. 9

Anisotropic Strain and


Transformation Plasticity
When an unstressed polycrystalline sample of
austenite is transformed to plates of ferrite, the
shear caused as each randomly oriented plate
forms is canceled on a macroscopic scale; only
the volume expansion is observed experimentally. However, if the plates do not form at randomfor example, when certain variants are favored because they comply better with the
external stressthe shear strains are no longer
canceled out. Transformation will then lead to
highly anisotropic strains, as illustrated in Fig.
11. Naturally, any anisotropy will be greatest for
displacive rather than reconstructive transformations, given that the former involve large
shear strains.

Modeling Anisotropic
Transformation Strains

1

The residual stress at ambient temperature is


larger when the austenite finishes transformation
at a high temperature. This is because thermal
contraction strains can no longer be compensated by transformation plasticity once the austenite has decomposed. Low transformation temperatures help minimize residual stresses.
High-strength welding alloys used for making
submarine hulls therefore have transformation
temperatures of less than about 250 C (480 F).
Figure 10 illustrates one kind of distortion
found in welds, measured in terms of the angle
h through which the unconstrained plates rotate
as they cool. Table 7 shows how the distortion
depends on the temperature at which the majority of the transformation is completed, for two
manual metal arc welds deposited with a 60 Vjoint preparation in a multipass fabrication involving about 11 layers, with two beads per layer
to complete the joint. The distortion is clearly
larger for the case where the transformation is
exhausted at the higher temperature.

Development of anisotropic transformation


strain when bainite forms under the influence
of a constant, elastic applied compressive stress. Note that
the shear strain associated with the formation of one plate
is about 26%, with a volume change of about 3%. The
potential for anisotropy is therefore much greater than illustrated here.

Fig. 11

Consider a distribution of bainite variants


along all radial directions in a circle with the
compression axis as its diameter (Ref 35, 36).
The circle is divided into 18 equal segments (i
1 18), each segment representing a particular orientation of bainite habit plane. The
choice of 18 segments is convenient and arbitrary. The compression axis of the sample is
taken to be the z direction, the x and y directions

Table 7 Chemical composition, calculated transformation temperature range (DT ), and


measured distortion (h) for two manual metal arc, multipass weld deposits
Composition, wt%
C

Fig. 10

Distortion caused by welding two plates that


were originally flat

0.06
0.06

Si

Mn

Ni

Mo

Cr

DT, C(F)

0.5
0.3

0.9
1.6

...
1.7

...
0.4

...
0.35

802400 (1476750)
422350 (792660)

14.5
8

Source: H.K.D.H. Bhadeshia and L.-E. Svensson, unpublished data, 1994

8 / Effect of Materials and Processing


being radially orientated; the unit vectors x, y,
and z define the orthonormal basis X of the sample, giving a corresponding reciprocal basis X *.
The shear and dilatational components of the IPS
accompanying the growth of bainite are approximately s 0.22 and d 0.03. Thus, the 3
3 deformation matrix describing the shape deformation is given by:
P

1 fse1 p1 f dp1 p1
fse2 p1 f dp2 p1
fse3 p1 f dp3 p1
fse1 p2 f dp1 p2
1 fse2 p2 f dp2 p2
fse3 p2 f dp3 p2

e1
p1
P I fs e2 ( p1 p2 p3) f d p2 ( p1 p2 p3)
e3
p3

where I is a 3 3 identity matrix. A further


reduction of notation is achieved using the
MacKenzie and Bowles notation (Ref 27):
(XPiX ) I fis[X;ei]( pi ;X *)
fid[X;pi]( pi; X *)

(Eq 4)

where the subscript i identifies a particular segment of interest and X and X *, respectively, represent the real and reciprocal bases of the coordinate system in which the deformation is
described. The notation due to MacKenzie and
Bowles (Ref 27) is discussed in detail in Ref 35.
The components of the shear direction and the
dilatation direction are given by:
[X;ei] fi[cos(hi) 0 sin(hi)]
[X; pi] fi[sin(hi) 0 cos(hi)]

120

180

210

(a)

Ui

330

where is the angle between the shear direction


and the direction of the shear component of the
applied stress as resolved onto the habit plane.
To facilitate a two-dimensional analysis, the
value of i is taken to be zero. A positive value
of Ui adds to the chemical driving force (DG ac
G c G a) for transformation; a negative
value thus opposes transformation. Using these
values of interaction energies, the model can be
modified so that the segments transform in an
order of decreasing Ui. There is, however, a further complication. The effect of stress should be
largest when the interaction energy is large compared with the chemical driving force. To allow
for this, the volume fraction fi of each segment

90

60

150

30

180

210

300
270

r
[s sin 2hI cos i d(1 cos 2hi)]
2

120
30

(X Pi X ) [1 0 0]
i1

where (x 1) and (z 1) give the strains


along the x and z directions. These are assumed
to be equal to radial and longitudinal strains eR
and eL, respectively.
It is expected that those segments that comply
best with the applied stress transform most rapidly, whereas the others do so at a lower rate, or
not at all. This can to some extent be incorporated into the model by calculating the energy
change Ui as the stress interacts with the shape
deformation of a particular variant (i). Patel and
Cohens method (Ref 30) gives:

60

150

240

(X Pi X ) [0 0 1]
i1
18

[X; x]

where p is the unit normal to the habit plane and


e is the unit direction along which the shear occurs. This can be written more succinctly as:

90

18

[X; z]

whereas a unit vector along x changes to

fse1 p3 f dp1 p3
fse2 p3 f dp2 p3
1 fse3 p3 f dp3 p3

where hi represents the orientation of the habit


plane of variant i and fi is the volume fraction of
bainite located in segment i.
A unit vector along the z direction changes to
a new vector z given by:

330
240

(b)

300
270

Transformation behavior inherent in the model (Ref 36). The dots illustrate the area fraction of each segment,
which in the calculations is scaled according to the value of Ui DG c. (a) Zero stress. All segments have
equal area fraction, and the order in which they transform is irrelevant. (b) Applied stress of 40 MPa and DG c 400 J/
mol. The area fractions of the segments are no longer equal. The segments in which the distance of the dot from the origin
is largest transform first.

Fig. 12

can be scaled according to the value of Ui


DG ac.
Note that for the model calculations, the transformation occurs with the most favored variants
growing first (Fig. 12). The model thus exaggerates the effect of stress, since in reality, for the
sort of stress levels considered experimentally,
no variant is likely to be entirely suppressed. In
addition, the grains in a polycrystalline sample
are randomly oriented, so that perfect compliance with the applied stress is impossible. Nevertheless, the trends revealed by the model are
expected to be correct.
The experimental data that need explaining,
and their interpretation in terms of the model are
summarized in Fig. 13 and may be stated as follows:
Without any stress, in a random polycrystal-

line sample, the transformation strains are


isotropic. This is easily understood since the
shear components of randomly oriented
plates tend to cancel out (Fig. 13a).
The application of the small stress at a high
transformation temperature (that is, a small
chemical driving force) causes the development of anisotropic strains, the transverse
strain first being negative and then positive
(Fig. 13b). The same effect is observed for a
large stress and low temperature (that is, a
large driving force). The model explains this
effect when it is assumed that the favored variants form first, but that the stress is not large
enough to suppress the eventual formation of
other variants. The signs of eL and eR are always opposite for the favored variants, but
are identical for the rest of the variants.
Therefore, the transverse strain is initially
negative but then becomes positive as transformation progresses. The low-stress/hightemperature situation is equivalent to the
high-stress/low-temperature case because in
both of these circumstances, variants that are
not favored cannot be suppressed. In the former case the stress is too small for suppression, whereas in the latter case the chemical
driving force is too large to permit suppression.
When a large stress is applied at a high temperature, the favored variants dominate.
Therefore, the strains are always of opposite
sign (Fig. 13c).
The model is thus capable of qualitatively explaining all the essential features of the formation of bainite under the influence of a small tensile stress. A uniaxial compressive stress (as
used in the experiments described below) simply
causes a reversal of the signs of the longitudinal
and transverse stresses; there is also a minor effect from the unfavorable interaction between
the compressive stress and the dilatational component of the IPS shape deformation.
The most interesting conclusion to emerge
from comparison of the model with experimental
data is that transformation under the influence of
a mild stress occurs sequentially. Variants that
comply with the applied stress grow first, fol-

Material Factors / 9
lowed by those that do not. This also carries the
implication that the interaction of the stress is
with the growth process (that is, the IPS shape
deformation) rather than the strain field of the
nucleus, which is likely to be different. It is
worth noting that there are similar results for
martensite: most favored variants grow first in
the sequence of transformation under stress (Ref
15, 24).

Summary
Many of the thermal properties of steelsfor
example, heat capacity, thermal expansion coefficients, and latent heats of transformation
are remarkably well understood. Indeed, commercially available thermodynamic databases
and programs can be used to estimate these
quantities as a function of temperature and
chemical composition. This capability has not
been exploited in the analysis of residual
stresses, even though phase diagram calculations
using the same software are now routine in industry and academia.
Other properties, such as elastic modulus, are

not yet calculable in the same manner. It may be


the case that they are insensitive to alloying, but
that remains to be demonstrated in the context
of residual stress analysis.
There is little doubt that transformations in
steel play a major role in the development of
residual stresses. For reconstructive transformations (for example, pearlite), it is the difference
in density between the parent and product phases
that contributes to transformation plasticity. The
plasticity can be much larger for displacive
transformations (Widmanstatten ferrite, bainite,
martensite) because of the large shear component of the shape deformation when these transformation products form. These are quite sophisticated effects which, with few exceptions,
are not incorporated in most residual stress analyses.

3.
4.
5.
6.
7.
8.

REFERENCES
9.
1. T. Inoue and Z. Wang, Mater. Sci. Technol.,
Vol 1, 1985, p 845850
2. P.W. Fuerschbach, in The Metal Science of
Joining, M.J. Cieslak, J.H. Perepezko, S.

10.
11.
12.

L
T

+
L, T
Strain

Strain

13.

14.

Transformation time

15.
16.

Equivalent time

(a)

17.
L

18.

Strain

Strain

L
0

19.

20.
21.
Equivalent time

Transformation time
(b)

22.
L

23.

+
Strain

Strain

0


Transformation time

24.

0
T

25.
Equivalent time

26.

(c)

27.
Schematic of the reported variations (Ref 24) in longitudinal and radial strains during the isothermal formation
of bainite under the influence of a tensile load, presented alongside predictions (Ref 36) from the crystallographic/thermodynamic model. The stresses are all intended to be below the austenite yield strength, and the data in this
case refer to uniaxial tension. (a) Zero stress, any temperature. (b) Small stress, low temperature. (c) Small stress, high
temperature; or large stress, low temperature.

Fig. 13

28.

Kang, and M.E. Glicksman, Ed., Minerals,


Metals and Materials Society, 1992, p 21
30
R. Schroder, Mater. Sci. Technol., Vol 1,
1985, p 754764
L. Kaufman, E.V. Clougherty, and R.J.
Weiss, Acta Metall., Vol 11, 1963, p 323
335
L. Kaufman, in Energetics in Metallurgical
Phenomenon, Vol III, W.M. Mueller, Ed.,
Gordon and Breach, 1967, p 5584
K. Hack, Ed., The SGTE Casebook: Thermodynamics at Work, Institute of Materials,
1996
D.G. Pettifor and A.H. Cottrell, Ed., Electron Theory in Alloy Design, Institute of
Materials, 1992
J.W. Christian, in Decomposition of Austenite by Diffusional Processes, V.F. Zackay
and H.I. Aaronson, Ed., Interscience, 1962,
p 371386
H.M. Clark and C.M. Wayman, Phase
Transformations, American Society for
Metals, 1970, p 59114
J.D. Watson and P.G. McDougall, Acta Metall., Vol 21, 1973, p 961973
J.W. Christian, Physical Properties of Martensite and Bainite, Special Report 93, Iron
and Steel Institute, 1965, p 119
H.K.D.H. Bhadeshia, Bainite in Steels, 2nd
ed., Institute of Materials, 2001, p 1453
G.W. Greenwood and R.H. Johnson, Proc.
Roy. Soc., Vol 238, 1965, p 403422
R.H. Johnson and G.W. Greenwood, Nature, Vol 195, 1962, p 138139
C.L. Magee, Ph.D. thesis, Carnegie Mellon
University, 1966
C.L. Magee and H.W. Paxton, Trans. Met.
Soc. AIME, Vol 242, 1968, p 17411749
F.D. Fischer, Acta Metall. Mater., Vol 38,
1990, p 15351546
J.B. Leblond, G. Mottet, and J.C. Devaux,
J. Mech. Phys. Solids, Vol 34, 1986, p 395
409, 411432
J.B. Leblond, J. Devaux, and J.C. Devaux,
Int. J. Plasticity, Vol 5, 1989, p 551572
J.B. Leblond, Int. J. Plasticity, Vol 5, 1989,
p 573591
J.B. Leblond and J. Devaux, Residual
Stresses, Elseviers, 1989, p 16
J.B. Leblond, Internal Report CSS/L/NT/
90/4022, FRAMASOFT, 1990, p 112
G.B. Olson, Deformation, Processing and
Structure, American Society for Metals,
1982, p 391424
H.K.D.H. Bhadeshia, S.A. David, J.M. Vitek, and R.W. Reed, Mater. Sci. Technol.,
Vol 7, 1991, p 686698
R. Von Mises, Z. Angew. Math. Mech., Vol
8, 1928, p 161
G.I. Taylor, J. Inst. Met., Vol 62, 1928, p
307
J.K. MacKenzie and J.S. Bowles, Acta Metall., Vol 2, 1954, p 138147
M.S. Wechsler, D.S. Lieberman, and T.A.
Read, Trans. Amer. Inst. Min. Metall. Eng.,
Vol 197, 1953, p 15031515

10 / Effect of Materials and Processing


29. A.B. Greninger and A.R. Troiano, Trans.
Amer. Inst. Min. Metall. Eng., Vol 140,
1940, p 307336
30. J.R. Patel and M. Cohen, Acta Metall., Vol
1, 1953, p 531538
31. L. Delaey and H. Warlimont, in Shape
Memory Effects in Alloys, J. Perkins, Ed.,
TMS-AIME, Plenum Press, 1975, p 89114

32. S. Denis, E. Gautier, A. Simon, and G.


Beck, Mater. Sci. Technol., Vol 1, 1985, p
805814
33. W.K.C. Jones and P.J. Alberry, Ferritic
Steels for Fast Reactor Steam Generators,
British Nuclear Engineering Society, 1977,
p 14
34. W.K.C. Jones and P.J. Alberry, Residual

Stresses in Welded Constructions, Welding


Institute, 1977, Paper 2
35. H.K.D.H. Bhadeshia, Worked Examples in
the Geometry of Crystals, Institute of Metals, 1987
36. A. Matsuzaki, H.K.D.H. Bhadeshia, and H.
Harada, Acta Metall. Mater., Vol 42, 1994,
p 10811090

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p11-26
DOI: 10.1361/hrsd2002p011

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Prestress Engineering of Structural


Material: A Global Design Approach
to the Residual Stress Problem
J. Lu, LASMIS, Universite de Technologie de Troyes, France

ALL MANUFACTURING PROCESSES introduce residual stress into mechanical parts,


which influences its fatigue behavior, fracture
strength, and even its corrosion resistance. Few
metalworking methods exist that do not produce
new stresses. The role of residual stress is, therefore, very important when designing mechanical
parts. Over the last few years, an increasing number of studies have been carried out to understand the effects of residual stress on mechanical
performance. This article attempts to present a
global approach to including residual stress in
expected fatigue life calculations, and the possibility of introducing it into mechanical engineering design offices. The definitions and origins of residual stress according to production
methods are first presented. Then, shown are the
problems involved in correctly adapting these
modeling techniques for use in design offices
and the industrial consequences of taking residual stress into account on quality assurance control procedures. This article deals mainly with
residual stress measurement techniques and the
overall necessity to combine destructive (incremental hole-drilling method) and nondestructive
(x-ray and neutron diffraction) methods in order
to precisely evaluate the residual stress distribution. Some new results concerning optical
methods are also discussed. Shown, too, are the
beneficial and harmful effects of residual stress
on the resistance of structures or industrial components, depending on whether they are tensile
or compressive. The different models used to
predict residual stress induced by different types
of processing are then presented. The last section
shows the effect of residual stress on fatigue behavior. A model based on the finite element
method (FEM) for predicting the relaxation of
residual stresses is presented. Finally, prediction
of fatigue life, taking residual stress into account,
using FEM is discussed. A new approach to concurrent engineering applied to the design of mechanical components with residual and applied
stress consideration is presented.

Residual stress is usually defined as the stress


that remains in mechanical parts that are not subjected to any outside stresses. Residual stress exists in practically all rigid parts, whether metallic
or not (wood, polymer, glass, ceramic, etc). It is
the result of the metallurgical and mechanical
history of each point in the part and the part as
a whole during its manufacture.
In the case of structural materials, surface engineering can lead indirectly to the innovation of
genuinely new materials based on conventional
materials. The frustrating slowness of some of
these developments reflects not only difficulties
in scaling up laboratory techniques, but also general conservation in the engineering industries
and a reluctance to change established habits.
One of the factors that contributes to slowing the
pace is that the current approach to surface engineering directly correlates performance to processing parameters; designers are left without the
tools needed to optimize and take into account
modified or prestressed surfaces for a particular
application, and the processors are left without
appropriate goals for surface properties.
The recent European Network of Surface and
Prestress Engineering and Design (ENSPED)
project led by the University of Technology of
Troyes and funded by the European Union is one
of the contributions currently being made to the
development of such tools. The aim is to develop
a project with a hard core of about 18 partners
representing a balanced selection of Europes
major industries, such as SNECMA; European
Aeronautic, Defense and Space (EADS); ABB;
Siemens; Volvo; Fiat; British Aerospace; Robert
Bosch; Hydro Aluminum; Wartisla; and so on.
It is based on building a bridge between the
fields of surface modification and prestress processing in materials engineering and computeraided design in mechanical engineering in order
to find an appropriate way of approaching this
interdisciplinary area. The members of the project will focus on the field of prestress processing. The main goal of the prestress engineering
approach is the optimization of residual stress for

an objective mechanical behavior of the material


and structure. Under this approach, the residual
stress must not be considered as a parameter that
only depends on the material processing conditions, but must also be considered as a parameter
that can be optimized.
There is an increasing interest in how the state
of residual stress affects the mechanical properties of a material and its structure. The failure of
a structure or a mechanical component is not
only due to external loads. Residual stress is an
important parameter in this respect. All manufacturing processes, for example, introduce a
new state of residual stress. This can have a positive effect, such as increasing the fatigue limit in
the case of surface compressive stress, or it can
have a negative effect, such as decreasing the
stress corrosion behavior of a material in the case
of tensile residual stress.
Basic and applied research in the field of residual stress has been stepped up in the last few
years. Residual stress is taken into account in
advanced design in the aerospace, automotive,
and nuclear industries. Even the microelectronics industry is starting to take residual stress into
account for the dimensional stability of electronic packaging.
The introduction of advanced materials has
also contributed to the development of knowledge in the field of residual stress. In fact, many
new materials are multimaterials, for example,
metal-matrix composites, plasma-sprayed coating, physical vapor deposition (PVD), and chemical vapor deposition (CVD) coatings, which
contain residual stress as a result of the thermal
and mechanical incompatibilities of the different
phases of the material or structure.
Figure 1 shows the different fields of research
in which residual stress is taken into account and
its relevance for industrial applications. Three
main fields must be developed for a global approach of prestress engineering: measurement
techniques for the quality control and processing
analysis, processing parameter optimization and
processing modeling, and modern design tool for

12 / Effect of Materials and Processing


the life cycle simulation with residual stress consideration.

Origins of Residual Stress


In general, macroscopic residual stress can be
induced due to:
Nonhomogeneous plastic flow under the ac-

tion of external treatment (shot peening, autofretting, roller burnishing, hammer peening,
shock laser treatment)
Nonhomogeneous plastic deformation during
nonuniform heating or cooling (ordinary
quenching, molding of plastics)
Structural deformation from metalworking
(heat treatment)
Heterogeneity of a chemical or crystallographic order (nitriding or case hardening)
Various surface treatments (enameling,
nickel plating, chrome plating, PVD and
CVD coating)
Differences in expansion coefficients and mechanical incompatibility of the different components of composites (composites with a
metallic and organic matrix, ceramic coatings)

Table 1 shows the different origins of residual


stress for metalworking operations usually carried out in the industry. To produce an industrial
part, one or several of the techniques listed in the
table can be used. To calculate the residual stress
existing in a part, the source of the stress must
be identified first.
Residual Stress Measurement Techniques
in Global Approach and Quality Assurance.
Over the last few decades, various quantitative
and qualitative techniques have been developed
(Ref 1). These techniques are used for the processing optimization and quality control of material. In general, a distinction must be made between destructive and nondestructive methods.

The first series of methods is based on destroying the state of equilibrium in the mechanical component. The residual stress is then evaluated from its relaxation. However, it is only
possible to measure the consequences of stress
relaxation and not the relaxation itself (displacement, fracture, strain). In most cases, the change
in strain is selected as the parameter to be studied. The following procedure is used:
1. Creation of a new stress state by machining
or layer removal
2. Detection of the local change in stress by
measuring the strain or displacement
3. Calculation of the residual stress as a function
of the strain measured using the elastic theory
(analytical approach or numerical calculations such as FEM)
During the recent years, the incremental holedrilling method is extensively used. It is sensitive to the first kind of residual stress, that is, the
macroscopic residual stress. The principle of this
technique is simple. It involves monitoring the
change in strain when a hole is drilled in a component with residual stress. These strain measurements can be related to the original residual
stress distribution in the analyzed sample at the
hole location. The relationship between the
strain and the residual stress can be calculated
with the calibration coefficients Ain and Bin. The
general approach used to determine the Ain and
Bin FEM is detailed in Ref 2. Recently, the high
sensitivity moire interferometer and incremental
hole-drilling method for residual stress measurement has been developed. The theoretical development of a combined method is introduced in
Ref 3 and 4. The relationship between the threedimensional surface displacements produced by
introducing a blind hole and the corresponding
residual stress is established by employing the
existing theoretical solution containing a set of
undetermined coefficients. The coefficients are
calibrated by the three-dimensional finite ele-

Quality control

Manufacturing

Measurement
technique

Prestress process
development and
modeling

Take residual stress into


account in integrated design

Relaxation of
residual stress

Dimensional
stability

Fig. 1

Effect of residual
stress on behaviors

New design

Mechanical
design

Main research fields and industrial application fields in which residual stress is taken into account

ment method. The in-plane surface displacements Ux and Uy and out-of-plane surface displacement Uz produced by the relaxation of
residual stresses are obtained using moire interferometry and Twyman-Green interferometry,
respectively. Figure 2 shows three-dimensional
displacement as a function of the drilling depth.
The main advantage of this technique is the possibility of studying the residual stress with an inplane stress gradient (Ref 5). It is also possible
to study composite materials (Ref 6). In this
case, the through thickness residual stress distribution was evaluated ply by ply in a carbon fiber/epoxy composite. Figure 3 shows an example of residual stress determined by using this
technique.
Recently, this method was used (Ref 7) to
study the Plastic Ball Grid Array (PBGA) package, which is a cost-effective surface-mounting
package with a high-density interconnection,
low profiles, and light weight. It is currently used
in many electronic products, including portable
telecommunication and computing products. A
typical structure in a PBGA package consists of
four layers: a plastic molding compound, a silicon chip, a chip-attach adhesive layer, and an
organic chip carrier. Due to the coefficient of
thermal expansion mismatch between the silicon
chip, the plastic compound, and the organic chip
carrier, considerable residual stresses are developed in the package during the assembly process. The process-induced residual stress can
play a significant role in the reliability of electronic components and packages. Since a PBGA
package is small and the surface layer is made
of a plastic material, it has proved very difficult
to use other existing methods of residual stress
measurement. In this research work, a practical
method has been developed to determine residual stress for electronic packaging. In this
method, blind holes are drilled into the specimens, and relationships are established between
the released surface displacement and the corresponding residual stress by introducing a set
of calibration coefficients. A multilayer three-dimensional FEM is established to determine the
relevant calibration coefficients. The surface displacements are measured accurately in a small
region around the hole. For a practical PBGA
package, the tensile residual stress is determined
in both the plastic molding compound and the
glass/epoxy laminate chip carrier. The method
is accurate, simple, convenient, and practical.
More applications in the field of electronic products are anticipated.
The x-ray diffraction and neutron diffraction
methods are based on the measurement of lattice
strains by studying variations in the lattice spacing of the polycrystalline material. The first
method measures the residual strain on the surface of the material, and the second measures the
residual strain within a volume of the sample.
Diffraction techniques can be used to study all
three types. The peak shift method is sensitive
to the first two, whereas line broadening is sensitive to the second and third types. These techniques have been used to measure the first type

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 13
of residual stress in different phases of advanced
materials, such as metal-matrix composites (in
matrix and in reinforcement) (Ref 8, 9).
Residual stress can be incorporated into the
design of mechanical components. Although this
leads to a better knowledge of the fatigue life of
parts and reduces the safety coefficient at the design stage, it also poses a host of new problems
on a quality assurance level. All statistical controls are only applied today to a few critical components in the aeronautical and nuclear industries; this practice could easily become

Fig. 2

widespread. Rapid ways of checking the residual


stress must therefore be developed. The methods
used industrially (x-ray diffraction and the incremental hole method) will not be sufficient in the
future. Other nondestructive testing (NDT) techniques (ultrasound, magnetic methods, acoustic
emission) are presently being developed. But as
they currently stand, these techniques use physical parameters that depend not only on the residual stress present in the parts, but also on microstructural changes. In the near future, NDT
techniques will be applied at the same time as

Three-dimensional displacements measured for different hole depth by a combination of optical methods and
the incremental hole-drilling method in a shot peened sample. Champ, field

Table 1

the reference techniques. Figure 4 proposes a residual stress inspection plan. Other new techniques, such as neutron diffraction and optical
method, will be introduced in industry. A detailed analysis of different techniques of measurement is out of the scope of this article.

Modeling of Process
The experimental techniques developed previously can contribute to the development of a
residual stress prediction model. The results calculated by the model make it easier to take the
residual stress into account during the mechanical design. The residual stress induced by the
thermal processing is extensively treated in the
other articles of this Handbook. Only some examples concerning the mechanical surface treatments and the comparison between the numerical simulation and experimental validation are
presented briefly.
Two kinds of models can be mentioned: analytical models and numerical models. In this
case, mechanical surface treatment models were
developed for shot peening (Ref 10) and cold
rolling (Ref 11), while several finite element
codes were used or developed for welding,
grinding, heat treatment (quenching), and thermal cutting (Ref 12, 13). The results show good
correlation between the prediction of the model
and the experimental results. But it can also be
seen that three-dimensional calculations are necessary to obtain good results in all directions. If
a two-dimensional calculation is used, the residual stress evaluation correlates well in one direction only. So, in the future, three-dimensional
calculations will be of greater significance for
real case modeling (Ref 13).
In the case of shot peening, a three-dimensional finite element dynamic model was developed to obtain a better description of the shotpeening process and to introduce this approach
into the component design (Ref 14). With the

Main origins of residual stress resulting from different manufacturing processes

Process

Casting
Shot peening, hammer peening, roller burnishing, laser
shock treatment, bending, rolling, chasing, forging,
straightening, extrusion
Grinding, turning, milling, drilling, boring

Mechanical

Thermal

Structural

No
Heterogeneous plastic deformation
between the core and surface of the
part
Plastic deformation due to the removal
of chips
No
No

Temperature gradient during cooling


No

Phase transformation
Depends on the material

Temperature gradient due to heating


during machining
Temperature gradient
Temperature gradient

Phase transformation during machining if


the temperature is sufficiently high
No
Change of volume due to a phase change

No

Thermal incompatibility

Welding
Brazing
Electroplating

Shrinkage
Mechanical incompatibility
Mechanical incompatibility

Temperature gradient
Thermal incompatibility
Mechanical incompatibility

Thermal spraying (plasma, laser, HVOF)

Mechanical incompatibility,
microcracking
Mechanical incompatibility
Mechanical incompatibility

Thermal incompatibility, temperature


gradient
Mechanical incompatibility
Mechanical incompatibility

New chemical component with volume


modification
Microstructural change (HAZ)
New phase at interface
Composition of plating depending on bath
used
Change of phase in plating

Quenching without a phase transformation


Surface quenching with a phase change (induction, EB,
laser, plasma, classical methods)
Case hardening, nitriding

PVD, CVD
Composite

EB, electron beam; HVOF, high-velocity oxygen fuel; HAZ, heat-affected zone

Change of phase
No

14 / Effect of Materials and Processing


help of three-dimensional modeling, the enormous influence of shot interaction is verified by
simulating simultaneous impacts. This simulation is very similar to the industrial process. Results of the residual stress obtained by the simulation are closer to the experimental results, if a
three-dimensional FEM is used (Fig. 5). The
higher the coverage rate, the lower the intensity
of the stress. This study shows how FEM can be
used to model shot peening and determine the
associated residual stress field.
The results obtained for both a single impact
and several impacts follow the general distribution of measured residual stress fields. This type

of analysis seems promising for studying the repercussions of a large number of parameters.
However, these methods are only efficient when
a maximum number of experimentally measured
parameters can be introduced. They also need to
be improved if they are to take most phenomena
into account. The proposed models could be
completed by analyzing the influence of the type
of contact, the direction of the shot during the
impact, behavior of the material, and so on. It
could then be possible to introduce the residual
stress field into metal parts in order to study their
behavior, using numerical simulation, for several
types of mechanical and fatigue life tests.

70
60

Average per lamina fibers


Average in two laminas of fibers

Average per lamina of epoxy


Average in two laminas of epoxy

50
40
Residual stress, MPa

30
20
10
0

112.5

225

337.5

450

562.5

675

787.5

900

10
20

Effect of Residual Stress on the


Mechanical Strength of Materials
Generality. When a part is subjected to a field
of elastic residual stresses characterized by a tensor rR, on which is superposed a field of service
stresses defined by the tensor rS, the real stress
to which the part is subjected is characterized by
the tensor rR rS (Fig. 6). If the residual
stresses are added to the service stresses (residual tensile stress, for example), the part is locally
overloaded due to residual stress. If, on the contrary, an appropriate finishing operation (shot
peening or roller burnishing, for example) is
used to introduce residual compressive stress,
the part is relieved of some of the load locally
and the mechanical performance of the materials
is increased as a result.
Figure 7 shows the properties of materials that
are influenced by residual stress. In the subsequent section, several quantitative examples of
the effect of residual stress are given.
Influence on the Fatigue Strength (Initiation Crack Phase). Residual stress plays an extremely important role with respect to the fatigue
strength of materials. It can be considered to be
a mean or static stress superimposed on the cyclic stress. As the mean stress rm increases, the
fatigue strength decreases. This is demonstrated
in the Haigh and Goodman diagrams.

30

y
40
50
Hole depth, m

Fig. 3

Residual stress distribution in the half-thickness of a carbon fiber/epoxy composite

S
R + S

I M

Mechanical components
manufacturing lines

Continuous control using


NDE techniques (US,
magnetic, AE)

Fig. 6

Superposing of residual stress and service stress

100

Good
quality

50

Defect
Assembly
of parts

Statistical analysis
of results

Residual stress, MPa

0
Measurement of the residual
stress distribution through
the thickness of the part
using conventional techniques
(x-ray, hole drilling)

0.10

0.20

0.30

Fatigue

0.40

50
100

Corrosion
under tensile
stress

150
200

Residual
stress

250
300

Breaking

srr L = 0.3 mm
szz L = 0.3 mm

Friction

Tensile
strength

350
400
Improvement of
process
Residual stress inspection plan for the purposes
of quality assurance. NDE, nondestructive evaluation; US, ultrasonic; AE, acoustic emission

Fig. 4

450
Depth, mm
Modeling of the residual stress distribution induced by shot peening using the 3-D finite element method

Fig. 5

Dimensional
stability

Fig. 7

Interfacial
bond strength
of coating

Effect of residual stress on the performance of


materials

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 15
leads to a tangential residual stress equal to or
slightly greater than the longitudinal stress. The
thickness of the material subjected to residual
compressive stress is in the same order of magnitude as the layer transformed during treatment
(Ref 15). Fatigue tests were carried out by the
French Technical Center for Mechanical Industry (CETIM) on 36 mm diameter XC42 steel cy-

Quenching treatment, after induction heating,


introduces very high residual compressive stress
into the hardened layer, which results from the
increase in volume of the martensitic structure
with respect to the ferrito-pearlitic structure (this
applied to the treatment of annealed steel, for
example). In induction-quenched cylindrical
bars, the residual stress on the surface usually
Table 2

Effect of quenching conditions and residual stress on fatigue strength

Type and depth of treatment


at 45 HRC, mm (in.)

Surface
hardness, HRC

Residual stress stabilized


at the fatigue limit, MPa (ksi)

Fatigue limit
after 5 106
cycles, MPa (ksi)
rm

ra

A, induction 2.7 (0.11)

5556

596 (87)

584 (85)

B, induction 4.2 (0.17)

5556

623 (90)

610 (88)

C, induction 4.7 (0.19)


D, water quenched after through
heating without stress-relieving
annealing 3.5 (0.14)

5459
6061

670 (97)
780 (113)

660 (96)
750 (109)

Longitudinal
stress

Transverse
stress

128 (19)
243 (35)
273 (40)
341 (49)
655 (95)
863 (125)
777 (113)

468 (68)
571 (83)
583 (85)
676 (98)
603 (87)
1132 (164)
1156 (168)

Source: Ref 16

Sample type

2.10, MPa

State

Crack
initiation site

BX, MPa

As welded

207

40

Post weld heat treated

207

37

Shot peened

392

519

Weld toe

1
Multipass

Fig. 8

Effect of residual stress on the fatigue strength of E690 welded joints. rRX, residual stress perpendicular to the
fillet (rRX); rRI, principal maximum residual stress evaluated at a depth of 0.1 mm below the surface. Source:

Ref 17

r rR rf*
106
ZAT D.T.
(40-75 mm)
ZAT B.S.
(40 mm)

Nontreated metal

da/dN, m/s

107

108

Rs = 0.1
E36Z

109

10

20

30

40

50

60

70

80

K, MPa m

Fig. 9
Ref 18

lindrical bars, quenched after induction heating,


and subjected to repeated bending stress (Ref
16). The results obtained are presented in Table
2. It can be seen that the higher the residual compressive stress, the greater the fatigue strength.
The resulting gain in fatigue strength produced
by the residual stress can be as much as 50% of
the fatigue strength of the base material treated.
Figure 8 shows the effect of residual stress on
the fatigue strength of welded HLE (E690) steel
joints (Ref 17). Three cases are show: as-welded
(residual tensile stress), stress relieved (no residual stress), and shot peened (residual compressive stress). A marked increase in the fatigue
strength was observed in the case of shot peening.
Influence on Fatigue Failure (Propagation
Phase) and Brittle Fracture (Ref 13, 18, 19).
In the case of welded assemblies, the presence
of welding defects at the weld toe and the geometric profile of the latter generally lead to a
limited period of crack initiation. The cracking
phase must be considered by taking into account
the residual stress field induced by the welding
operation.
The decisive influence of the residual stress
field on the crack propagation speed has been
demonstrated (Ref 18). Figure 9 shows the results of cracking as a function of the residual
stress. Relieving residual stress by heat treatment
changes the crack propagation speed considerably when the stress is high.
In the case of a brittle fracture, cleavage starts
in a grain when the local stress reaches a critical
value of rf* and it generally propagates without
difficulty in the adjacent grains by producing a
brittle fracture. The tensile residual stress rR, in
addition to the applied stress r, initiates this type
of failure for low loads, such that:

Effect of a residual stress-relieving treatment on the cracking speed in the HAZ (butt-welded assembly of an
E36Z steel). ZAT D.T., HAZ of the heat treated sample; ZAT D.T., HAZ of the as-welded sample. Source:

Once cracking has been initiated, the applied


stress alone can be enough to allow propagation
to continue at a high speed. Failure is therefore
very sudden. Residual stresses that facilitate the
initiation of brittle fracture by cleavage are therefore very dangerous for steels under load at low
temperature. This is why the stress relieving of
welded joints is also recommended.
Grain slips come up against inclusions and
create concentrated stresses at their interface that
lead to fracture of either the interface or the inclusion. Cavities then appear for a critical initiation stress and grow by plastic deformation of
the matrix until their coalescence leads to ductile
fracture at least on a microscopic level. The
speed at which the cavity grows is not only proportional to the plastic deformation speed but
also to the degree of triaxial state of the stresses
and to the ratio of the mean stress to the ultimate
stress. Coalescence is a plastic instability phenomenon that no doubt occurs for a critical cavity size. Tensile residual stress not only facilitates the initiation of cavities but, by increasing
the mean stress, also accelerates growth. These
two effects combine to decrease the critical elon-

16 / Effect of Materials and Processing


gation of ductile fracture. However, this is only
important if the ductility is already very low in
the absence of residual stresses, since plastic deformation can eliminate them.
Effect on Stress Corrosion (Ref 20, 21).
Stress corrosion is a mechanical and chemical
cracking phenomenon that can lead to failure under the combined effect of tensile stress and a
corrosive environment. Cracking is generally
transcrystalline and can appear on all types of
materials, such as aluminum alloys, steels, copper, titanium and magnesium. The introduction
of residual compressive stress can considerably
increase the fatigue life of parts subjected to
stress corrosion. Tests carried out on magnesium
test specimens placed under stress in a salt solution gave the following results:

used, can remain in the coatings and in the substrates. They are of several types: microstresses
in the grain, produced during cooling, and
macrostresses affecting the entire coating.
Macrostresses are created not only by cooling
but also by the difference in temperature between the substrate, the sprayed layer, and the
outside surface. The differential contraction thus
produced between the various materials, due to
the difference in physical and mechanical properties, determines the stresses in the coating and
the coating-substrate interface. These stresses
therefore influence the mechanical and thermomechanical behavior of the coated parts.
In order to appreciate the quality of a coating,
three types of damage to parts in service can be
considered:

Ground test specimen: failure after two min-

The coating deteriorates rapidly.


The properties of the substrate are modified

utes
Shot peened test specimen: no cracking after
12 days under the same conditions
The tests conducted by W.H. Friske show that
the fatigue life is 1000 times greater for a shotpeened 304 grade stainless steel part than it is
for a non-shot-peened part (Ref 20). Tests carried out by CETIM on Z6CN18.9 stainless steel
produced similar results (Ref 21).
Effect on Adhesion of Coatings (Ref 22
25). Most coatings are produced for a specific
reason, particularly to improve the corrosion and
wear resistance of the base material, or to provide a thermal barrier for use at high temperature. But this is only achieved if the coating adheres to the substrate correctly. Adhesion
therefore indicates correct preparation of the surfaces to be coated and the quality of the coating
operation. The last few years have seen the appearance of plasma-spraying techniques, both at
atmospheric pressure and at reduced pressure.
These processes offer a high degree of flexibility
for coatings in critical areas. However, high residual stress, inherent to the coating method

R
Kc

7
6

300

200

4
3

100

2
1
0

Residual stress, MPa

Kc, MPa m

1
2

TT

STT

100

Influence of heat treatment on the residual


stress and the toughness of the interface: case
of plasma-sprayed coatings at atmospheric pressure. TT,
heat treated; STT, as-sprayed

Fig. 10

by the coating.

The damage is common to both materials. It

is located at the interface and jeopardizes both


the adhesion and the fatigue life.
C. Richard et al. have shown that decreasing
the residual stress by thermal treatment of the
coating considerably improves adhesion at the
interface (Ref 23). Figure 10 illustrates the effect
of residual stress. It can be seen that the apparent
toughness of the interface is improved by 100%
when heat treatment is applied. There is a high
level of residual tensile stress in the test specimen without heat treatment. When the level of
residual tensile stress increases, the true toughness of the coating decreases. An increase in the
residual compressive stress produces the opposite effect.
Influence of Residual Stress on the Tensile
Strength, Friction, Wear and Dimensional
Stability. The effect of residual stress on the tensile strength is obvious, particularly in structures
made of composite materials or when the prestressed layer is very thick compared with the
thickness of the parts. In composites, residual
stress is produced as a result of the thermal and
mechanical incompatibility of the reinforcements and matrix. This can influence the macroscopic properties of composites under tensile
or compressive stress (Ref 26).
Little research has been carried out on the effect of residual stress on friction and wear properties. Their role is often masked by other parameters. The increase of hardness during treatment
and changes in the toughness and adhesion of
antiwear coatings due to residual stress can considerably affect the resistance to friction. Up until the present, this effect has been integrated into
the global parameter of adhesion. In the future,
work will be carried out to try to determine the
real effect of residual stress.
The problem of dimensional stability has been
known for a long time. When a part is machined
that contains residual stress produced by heat
treatment or welding, the shape of the part can
change after operation due to the relaxation of
residual stress. This is why stress-relieving treatments are frequently used to avoid this type of

defect. Reference 27 gives a very methodical approach to defining the criteria and processes relating to relieving stress in welded structures.
The same type of reflection can be applied to
other types of parts.

Taking Residual Stress into Account


when Calculating the Expected
Fatigue Life
In the previous section, the different effects of
residual stress on the mechanical strength of
structures and materials were mentioned. Although the ability to quantitatively estimate the
fatigue life taking residual stress into account is
just beginning, it is still too early to extend these
predictions to other types of stress that are far
more complex and involve physical and chemical phenomena. Statistics show that failures of a
purely mechanical origin are mainly due to fatigue. It is for the reasons indicated previously
that this article only addresses problems concerning the prediction of fatigue life. Two articles (H. P. Lieurade and A. Pellissier-Tanon) in
Ref 13 deal with the question of predicting the
effect of residual stress on crack propagation
phase. Although, they concern welded structures, the concepts developed in these two references can be applied to other types of structures. By limiting the approach to prediction of
the fatigue life to the fatigue cracking initiation
stage, the problem of predicting the fatigue life
of mechanical components subjected to a high
cycle fatigue can be analyzed.
Calculating the Effect of Residual Stress on
the Fatigue Strength. Based on the experimental results mentioned previously, it would seem
that a linear relationship of the Goodman type
can be used to take residual stress into account:
ra rD

rD
(rm rR)
Rm

(Eq 1)

where ra is the amplitude of admissible stress,


rm is the mean fatigue stress, rD is the purely
reverse tensile fatigue limit, Rm is the true rupture strength, and rR is the residual stress measured in the direction of the applied service
stress. The numerous studies mentioned in Ref
28 show that the effect of residual stress is
greater when the properties of the materials are
high.
If we try to represent the development of ra
according to the residual stress rR by an equation
of the following type:
ra rD rR

(Eq 2)

the experimental results generally show that


increases with the strength of the material; for
example, in the case of machining stresses in an
XC38 grade steel, Syren et al. found:
0 in the annealed state
0.27 when quenched and tempered
0.4 when quenched

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 17
Unfortunately, these results are in contradiction with an equation of the Goodman type. In
Eq 2, the coefficient is none other than what
is usually called the endurance ratio:

rD
Rm

This parameter decreases as the rupture strength


of steels increases.
This apparent contradiction is probably explained by the fact that the residual stress relaxation phenomenon has not been taken into account. The value of the residual stress rR to be
introduced into equation of type 1 or 2 mentioned previously, must correspond to the stabilized fatigue stress, or the coefficient of influence
will include the relaxation process. The article
by D. Lohe and O. Voehinger in this Handbook
presents a large number of results concerning the
relaxation of residual stress under mechanical
and thermal loading. References 29 and 30 provide the information on the relaxation mechanism of residual fatigue stress as a function of

800
A
B Repeated
C bending tests
D
Repeated tension test

700

a, MPa

600
500
400
300
200
100
0
0

1000
m, MPa

2000

Use of Haigh diagrams to take longitudinal residual stress into account (XC42 steel
quenched after induction heating). A through D represent
results of repeated bending tests; X, repeated tension test.
Table 2 presents fatigue test results that correspond to A to
D.

Fig. 11

a, MPa
A
B
Repeated
C
bending tests
D
Repeated tension test

400

200
100

100

200

300

400

Pmax, MPa
Use of the Dang Van criterion to take residual
stress into account (XC42 grade steel quenched
after induction heating). A through D represent repeated
bending test; X, repeated tension test.

Fig. 12

salt f (A, B Pmhp, C Palt )

where Pmhp is the mean hydrostatic pressure, salt


is the amplitude of octahedral shearing or amplitude of the maximum shearing, and Palt is the
amplitude of hydrostatic pressure.
An example can be given as follows:
E
salt A B PD
mhp C Palt

300

200 100

the cyclic behavior of the materials. However,


Syrens results show that relaxation is much
greater when the mechanical properties are
lower.
When these experimental results are used with
the residual stress measured after carrying out a
fatigue test, and therefore stabilized, it is sometimes possible to use an equation such as Eq 2.
In the case of the fatigue bending test on cylindrical XC42 steel bars quenched after induction
heating (Table 2), the fatigue test results for the
different treatments correspond perfectly to the
Haigh diagram, provided any possible influence
of transverse residual stress on the fatigue stress
is ignored (Fig. 11).
It is not possible, however, to extend these
results to all materials and to the different manufacturing processes that introduce residual stress.
Also, preliminary tests are needed to validate the
methodology. The use of residual stress in calculations based on endurance diagrams of the
Haigh or Goodman type usually only allows for
an estimation of the increase in fatigue strength
as a function of the residual stress. Secondly, this
approach only allows for the combination of uniaxial stresses. Yet the residual stresses produced
by the various manufacturing methods used to
make the part are always multiaxial. The stresses
on the surface are biaxial while those inside the
part are triaxial. Depending on the area in which
the fatigue crack is initiated (on or below the
surface), the biaxial or triaxial stresses need to
be included when calculating the fatigue life.
This raises the problem of choosing a multiaxial
fatigue stress criterion. A simplified approach
based on an endurance diagram can therefore
only be an approximation.
The test carried out by the CETIM (Ref 16)
shows that the traditional Mises and Tresca criteria can only be used in the presence of higher
mean or residual stress. In this case, it is preferable to use criteria (Ref 31) that include the
amplitude of octahedral shearing (socta) or the
maximum shearing (sa) and maximum hydrostatic pressure (Pmax), as indicated here:

(Eq 3)

where A, B, C, D, and E are material constants.


If Dsalt is taken on the maximum shearing
plane, D E l and B C, the result is the
Dang Van criterion (Ref 32).
If Dsalt is taken on the octahedral shearing
plane, when D E 1 and B C, the Crossland criterion results (Ref 33); when D E
1 and C 0, the Sines criterion (Ref 34); and
when D E 1 and B C, the Kakuno criterion (Ref 35).
This type of development can be continued to

invent new criteria, but it leads to complications because of the increasing number of parameters that need to be determined. Even with a
linear relationship of the Dang Van type, two
Wohler curves have to be determined to obtain
at least the two points needed to produce the diagram. If additional constants are added, the test
plane will be even greater, which means that the
criterion cannot be used in industry. As a result,
the criterion to be used must be simplified as
much as possible. This case deals with radial
loading problems (r1 K1 r2 K2 r3), and a
relationship of Eq 3 is sufficient. To simplify
matters further, the Crossland or Dang Van criterion can be used. In the case of combined and
out-of-phase loading, new criteria have been developed to take the out-of-phase effect into account (Ref 3639).
But as yet, these criteria have not been validated in a study in which combined and out-ofphase residual stresses have been taken into account. When the fatigue stress is complex, it is
also very difficult to calculate the expected residual stress relaxation.
When fatigue cracks are initiated on the surface, the stresses to be taken into account are
biaxial; this gives the following for the Crossland or Dang Van criterion:
socta
sa
Pmax

2
3

r 1a r 2a r r
2

1a 2a

r1a
2
1
(r1a r2a r1m r2m r1R r2R)
3

where r1a, r2a represent the amplitude of the


main reversed fatigue stresses (r1a r2a); r1m,
r2m represent the average value of the main fatigue stresses; and r1R, r2R are the residual stress
values measured in the two main directions (stabilized values).
To use the multiaxial fatigue criteria, the reference curve for the material being considered is
needed, just as it is when using the Goodman or
Haigh diagram. Reference 16 shows that the use
of Crossland or Dang Van criteria takes the increase in the bending fatigue strength into account perfectly as a function of the residual
stress introduced by the various treatment conditions (Fig. 12).
When the multiaxial aspect is brought into the
picture, the method that consists in introducing
residual stress into the calculation in the same
way as a mean stress, therefore, seems to give
satisfaction. The whole problem lies in defining
the residual stresses to be included in the calculation.
Taking residual stress into account is essential
for correct prediction of the fatigue limit. Figure
13 shows the important role played by compressive residual stress. If it is not taken into account,
the fatigue strength is underestimated (Fig. 13a).
If the residual stress measured or calculated is
used without taking relaxation of the residual
stress into account, the fatigue strength is over-

18 / Effect of Materials and Processing


estimated (Fig. 13b). The correct method consists in calculating the fatigue strength after taking relaxation into account (Fig. 13c).
In order to correctly evaluate the effect of residual stress, various problems must be solved:
Measuring methods must be available to de-

a, MPa

a, MPa

(a)

a, MPa

Pmax

Fig. 13

termine residual stress in the critical zone. A


large range of measuring methods currently
exists and it is possible to take measurements
(Ref 1) in most of the cases studied, particularly as a result of development of the x-ray
method and the incremental hole method.
When a complete measurement profile is

Relaxation
des R

Pmax

Initial
(b)

Pmax

(c)

Illustration of the different methods used to take residual stress into account

a, MPa (Surface)

Xian
CETIM

300

200

100

100

200

300

400

500

Pmax, MPa (Surface)

Fig. 14

Fatigue results for a 45SCD6 steel treated under different conditions in a Dang Van diagram, taking the residual
stress on the surface into account. Xian, results obtained by Xian Jiaotong University; CETIM, results obtained

by CETIM

a, MPa

available of the residual stress in the surface


layer in which the fatigue crack is initiated,
the profile sometimes has high stress gradients. It is then difficult to know what stress
value to choosethe surface stress or the
stress peak that is often slightly below the surface (in the case of shot peening, for example).
To make a correct calculation, it would be
necessary to use calculation methods that take
the stress gradient into account and make the calculation not only for a single point, but for a
sufficient thickness of the material (thickness of
critical layer) for it to be representative of the
basic volume in which the fatigue damage process occurred (Ref 31). Figures 14 and 15 give
an example of processing of our results (details
can be found in Ref 29) in the case of the fatigue
of a shot-peened spring steel. Figure 14 shows
the fatigue results on a Dang Van diagram, taking both the residual stress and its relaxation into
account. A fairly good correlation can be observed. This indicates that a multiaxial fatigue
criterion taking the hydrostatic pressure into account can be used to predict the fatigue strength
in the presence of residual stress. Since, in this
case, the crack initiation zones are below the surface, calculations were made for different critical
layer depths. Figure 15 shows the results obtained for a critical layer depth of 100 lm. Better
alignment of the experimental points was observed. This example illustrates the possibility
of improving the calculation precision by using
the critical layer thickness approach. It is particularly relevant in the case of notched parts.
It has long been known that residual stresses
are not stable when they are subjected to fatigue
loading. To calculate the expected fatigue life,
precise information is therefore needed in order
to introduce stable residual stress values into the
calculation presented previously, that is, the
stress values that are really likely to be present
in the part during the best part of its lifetime.
The stress must therefore be measured on a
part already under cyclic loading or relaxation of
the residual stress estimated according to expe-

600
500

350
With
Without

400

A Consideration
600 MPa

200

a, MPa

300
Xian
CETIM

300
550 MPa
2 106
5 106

100
250
300

200

100

100

200

300

400

500

Fatigue results for a 45SCD6 steel treated under different conditions in a Dang Van diagram, taking the residual
stress into account using the critical layer approach. Xian, results obtained by Xian Jiaotong University;
CETIM, results obtained by CETIM

50

100

150

200

Pmax, MPa

Pmax, MPa

Fig. 15

107

240

Prediction of the fatigue life using the Dang


Van criterion for a shot-peened 35NCD16
steel, without taking residual stress into account

Fig. 16

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 19
rience or modeling. In Ref 29 and 30, a complete
model was presented using FEM to determine
the stabilized residual stress after fatigue loading. This estimation of the residual stress can
then be used to calculate the fatigue life of a part,
taking residual stress into account. Despite the
initial definition of the Dang Van criterion,
which proposes that it only be used in cases of
fatigue strength with an unlimited number of cycles, an attempt was made to extend this criterion
to include a limited fatigue life with a very large
number of cycles (more than 2 106) (Ref 40).
Figure 16 shows an example of the fatigue life
estimated by calculation. First, a fatigue strength
diagram of the Dang Van type was defined according to the fatigue life obtained from a series
of fatigue life contours. Then the residual stress
relaxation was calculated using FEM (Ref 29).
Finally, the stabilized residual stress was introduced into the diagram. In this example, in the
case of a loading of 550 MPa, the point corresponding to loading is inside the limit of the
fatigue life at 107 cycles. No failure occurs. For
a loading of 600 MPa, the point corresponding
to loading including the residual stress is between the line corresponding to 5 106 cycles
and that of 2 106 cycles. Failure therefore

alt

2 Hardening effect

3 Roughness increase

1
3

Pmax

Fig. 17

Illustration of the effect of the surface finish and


strain hardening on the fatigue strength

Role of hardening
and residual stress

Fatigue life increasing, %

20
Role of
residual stress
10

0
Role of superficial hardening
10
20

30

40

50

HRC
Effect of the resistance of the base metal on the
increase in the fatigue strength after shotpeening treatment, distinguishing between the effect of
strain hardening and that of residual stress

Fig. 18

A new roughness that changes the local stress

concentration

Additional strain hardening of the surface due

to plasticizing

A new metallurgical structure of the surface

layers
Figure 17 shows the effect of the surface finish
and strain hardening on the fatigue strength of
materials. It can be seen that an increase in the
roughness decreases the safety area, and strain
hardening increases the safety area, provided it
does not damage the material.
In the case of thermal or thermochemical surface treatments (induction quenching, case hardening, etc.), for example, it is necessary to take
the new fatigue strength of the treated layer into
account in the calculation.
The problem is more complex in the case of
residual stress introduced by plastic deformation
(pre-straining, machining, shot peening, roller
burnishing), since it is more difficult to distinguish between the influence of residual stresses
and residual microstresses present in the grains
of the deformed material and that of strain hardening of the material.
Evans (Ref 41) made this distinction in the
case of shot peening; to do so, he carried out
three types of fatigue tests on materials with
various mechanical properties:
Fatigue tests on a non-shot-peened material
Fatigue tests on a shot-peened material
Fatigue tests on a shot-peened material, but

with a mean test stress rm, which compensated for the surface residual stress. In this
type of test, the effect of the macroscopic residual stress is cancelled out and the fatigue
strength obtained only depends on the increase in the mechanical properties of the
plastically deformed material and the residual
microstresses distributed throughout the material. The results obtained are presented in
Fig. 18.

40

30

occurs between 2 and 5 million cycles. The tests


give an average fatigue life for the previously
mentioned loading in the order of 3.5 million
cycles. This example shows that, if the cyclic
properties of the material are correctly known, it
is possible to predict the fatigue strength of the
material in the presence of residual stress.
However, it should not be forgotten that other
factors must also be taken into account in calculating the fatigue lifethe introduction of residual stress is often accompanied by other
changes in parameters that have an influence on
the fatigue strength. In particular, these include:

It can be observed that, for materials with low


resistance, the increase in the fatigue strength is
mainly due to surface strain hardening. On the
other hand, for highly resistant materials, it is
mainly the influence of the residual stress that
governs the fatigue strength. When materials
have low elastic limits, the stresses introduced
by shot peening relax much more easily than
they do when the elastic limit is high. This test
only shows a general tendency and does not ex-

actly show what the author is trying to demonstrate, since in the two types of tests for the same
material, the ratio of R(rmin /rmax) is modified.
As has been shown, the residual stress relaxation
changes when the R ratio is varied (Ref 29). The
real contribution of each factor will therefore be
different from that indicated in Fig. 18.
It can thus be seen that taking residual stress
into account in the calculation requires a serious
examination of the different parameters involved. When reliable results are needed, fatigue
tests will no doubt need to be carried out on the
part or the structure concerned. However, modeling enables the variation in the different parameters to be rapidly simulated in order to find
an optimal solution.
Partial Summary. The previous results show
that it is now possible to take residual stress into
account in calculations designed to predict the
fatigue life using a global approach. This must
take the relaxation of residual fatigue stress into
account, as well as the other effects (strain hardening, hardness) introduced by the manufacturing method used. A multiaxial fatigue criterion
that can integrate both the problem of residual
stress and the effect of the stress gradient applied
to a zone in the presence of stress concentration
has been developed, that is, the Crossland or
Dang Van criterion. It is used for a stabilized
state of residual stress, averaged out for a basic
volume of damage (thickness of critical layer),
and applied within a network of contours that
represents the fatigue life. In the future, tests will
be carried out to validate this type of criterion in
the case of combined stresses on notched parts
in the presence of residual stress.

Incorporating the Notion of


Residual Stress into the Design
Office
Incorporating the notion of residual stress into
the design office must be gradual and can be divided up into several phases.
Today, very few industrial sectors consider
the residual stress parameter directly. In technical specifications, requirements are included that
are often closely related to residual stress without actually naming it. An Almen intensity must
be guaranteed in the case of shot peening, for
example, a roller-burnishing load, a machining
procedure or a minimum treated thickness in the
case of thermal or thermochemical treatment,
and a maximum dimensioning tolerance in the
case of a machined or welded part.
In the first phase of incorporation, a semiquantitative notion can be used to evaluate the
increase in performance in terms of fatigue life
or fatigue strength. A few examples can be presented. Table 3 gives an example of the effectiveness of shot peening in increasing the fatigue
life of different types of mechanical parts, and
Fig. 19 shows the beneficial role played by roller
burnishing on the fatigue strength of spherical
graphite cast iron crankshafts. Figure 20 shows

20 / Effect of Materials and Processing

of residual stresses is realized in the first series of fatigue cycles.

The cyclic behaviors of material are very im-

Table 3 Increase in the fatigue life of various mechanical components as a result of shot
peening
Type of part

Type of stress

Spindles
Shafts
Gear box
Crankshafts
Aircraft coupling rods
Driving rods
Cam springs
Helical springs
Torque rods
Universal joint shaft
Gear wheel
Tank chain
Weld
Valve
Rocker arm

Increase in the fatigue life (%)

Reverse bending
Torsional
Fatigue life tests in service
Fatigue life tests in service
Tensile compression
Tensile compression
Dynamic stress
Fatigue life in service
Dynamic stress
Reverse bending
Fatigue life tests
Fatigue life tests
Fatigue life tests
Fatigue life tests
Fatigue life tests

4001900
700
80
3000, but highly variable
105
45
100340
3500
140600
350
130
1100
200
700
320

250
200
Alternated
stress

150

0
100
0

50
Repeated
stress

0
70

S-

G
70

S-

G
70

S-

G
40

S-

G
40

70

S-

S-

llin
g

en

ng

in

ro

en

io

ce

rfa

su

ct

di

at

re

tri

pe

fe

ot

du

in

sh

ni

tt

ou

ni

ni

te

io

ith

By using the cyclic behaviors of material, a


simplified method to calculate residual stress relaxation has been proposed in the first section.
Secondly, a method to predict fatigue life by taking the stabilized residual stresses into account
is presented. This design tool is based on the
FEM. It has been applied to shot-peened
35NCD16 grade steel. The different fatigue parameters often used in material research are studied. On the other hand, an experimental investigation about this material had been done by
Bignonnet (Ref 42). The results of the study
show that this design tool on fatigue developed
by the Laboratory of Mechanical Systems Engineering (LASMIS) (Ref 43) is able to take into
account different loading parameters.
The residual stresses, however, can be relaxed
by the deliberate application of thermal or mechanical energy. They will be especially relaxed
when the structure is subject to cyclic loading.
The relaxation phenomenon depends on a complex interaction of a number of factors, such as
the applied stress amplitude, the number of cycles of loading, the state of initial residual
stresses, and the nature, origin, and mechanical
properties of the material. In this article, only
relaxation during cyclic loading and the influence of the stable residual stresses on the fatigue
life are studied.
In only a few cases are the residual stresses
systematically analyzed using measurement of
the residual stress state during and after fatigue
testing. This is usually a difficult, time-consuming task. Several fatigue tests under tensile and
torsion loading with different stress amplitudes
have been done.
A numerical method for the prediction of the
residual stress distribution during and after fatigue has been developed. Finite element software is used for incorporating cyclic plasticity
into the calculation. A simplified method to calculate stabilized residual stresses was proposed.

stresses relax with the increase of the number


of cycles up to a stabilized state of cyclic
properties.

For a cyclic hardening material, the relaxation

Maximum admissible stress, MPa

Example of Integrated Design Tool

For a cyclic softening material, the residual

According to a mechanical approach, we can


make the following general predictions (Ref 30):

Fig. 19

Effectiveness of roller burnishing in increasing the fatigue strength

Thermal treatment by electron beam, R = 0.1


TT by CO2 laser, R = 0.1
TT by CO2 laser, R = 1
TT by laser shock, R = 0.1

Treatment or process

a horizontal comparison of gains to be expected


in terms of fatigue strength from various surface
treatments. The results presented here are not at
all exhaustive and are taken from a limited bibliography. However, this figure should not be
taken as a reference, since the geometry of the
test specimens differs for each type of treatment.
In certain cases, this parameter can have an important effect on the gain achieved. Each industrial sector must carry out this type of comparison for the treatments and materials used in order
to help engineers design their products more effectively.
The second phase consists of predicting the
fatigue life using the notions developed in the
previous paragraph.
The third phase is the development of integrated tools for taking the residual stress into
account. The following paragraph presents an
example of such a design tool on fatigue for
three-dimensional components.

ST by nitriding R = 0.1
TT by induction R = 0.1
TS carburizing R = 0.1
ST by surface rolling R = 0.1
200%
ST by surface rolling R = 0.1
ST by shot peening R = 0.1
0

Fig. 20

25

50
Increase, %

75

Beneficial effect of various surface treatments on the fatigue strength (maximum gain reported in the literature).
TT, thermal treatment; ST, surface treatment

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 21
portant in the prediction of residual stress relaxation. They make it possible to calculate the residual stress state for a corresponded number of
cycles.
Then, a method to predict fatigue life while
taking these stabilized residual stresses into account is presented. The fatigue criterion under
multiaxial stresses is used. After the relaxation,
the residual stresses can be combined directly
with the cyclic loading. They change the value
of the mean stresses and influence the fatigue
life.
As a result, a design tool on fatigue for threedimensional components, FATIGUE3D, has
been developed in the laboratory of LASMIS

Calculation of
residual stresses

Outside solicitation,
cyclic loading

(Ref 43). It can take the residual stresses into


account when calculating the fatigue life of a
structure. It is based on the FEM. With this tool,
an iso-colored image of distribution of fatigue
life or safety factor can be obtained.
These methods mentioned above are applied
to shot-peened 35NCD16 grade steel. The different fatigue parameters often used in material
research are studied. A comparison with the experimental results has been carried out (Ref 43).
Basic Steps for the Calculation. Figure 21
shows the elementary steps of the calculation of
fatigue life while taking into account the residual
stresses.
First, the initial state of the residual stresses

Introduction of initial field


of residual stresses

Measurement of
residual stresses

Prediction of
residual stress

Cyclic behaviors
of material

Fatigue, 3D

Mechanical properties
of material

Distribution of fatigue life,


safety factor, and
admissible residual stress

Fig. 21

Basic steps of calculation while taking into account the residual stresses

N1

Example and Results


N1

Fig. 22

Principle of calculating fatigue life

s = OM/OM

M
N-cycles
M

Fig. 23

can be obtained by the measurement method or


by the simulation of surface treatment. Then, this
initial field of residual stresses can be introduced
by FEM using ABAQUS (Ref 44, 45). Furthermore, a stabilized field of residual stresses can
be predicted after an elastic-plastic calculation,
in which the cyclic behaviors of material are
used. This stabilized field of residual stresses after relaxation must be taken into account when
calculating the fatigue life. At last, with the help
of a design tool on fatigue, FATIGUE3D, the
distribution of fatigue life and that of the safety
factor on fatigue can be obtained.
Principle and Basic Function. According to
the criteria of fatigue under multiaxial loading,
the fatigue life changes along with two parameters. One of them is maximum static pressure;
another is equivalent octahedral stress. Therefore, two stress-number of cycles (S-N) curves
are needed under simple cases (for example,
bending or torsion). With these two curves, a line
can be determined for a certain number of cycles
(life) in the stress plane P s (Fig. 22).
A different fatigue life is represented by a different line. If there is a complex stress state, it
gives a point in the plane P s. Respectively,
the line to which this point is nearest represents
the fatigue life of that stress state. In this way,
the fatigue life for any complex stress state and
for any part of a component can be predicted. A
distribution of the fatigue life is helpful for the
designer.
On the other hand, the safety factor on fatigue
is another interesting parameter for the designer.
FATIGUE3D provides a distribution of the
safety factor. In Fig. 23, N-cycles are the designed life and M is the point given by the stress
state. The safety factor on fatigue is defined as S
OM/OM.

P
Principle of calculating the safety factor on fatigue

Fig. 24

Three-dimensional mesh

Material and Geometry. The shot-peening


process is a widely used technique because it can
produce a field of residual stress on the surface
of a part. These stresses are compressive and will
improve the properties on fatigue. As an example, a shot-peened part made of 35NCD16 grade
steel (0.35% carbon, 3.7% chromium, 0.3% molybdenum and having tensile properties of: a
yield strength of 1000 MPa, an ultimate tensile
strength of 1100 MPa, and an elongation of
17.5%) is studied. Many results of fatigue experiments are available regarding tension compression and torsion loading (Ref 44).
The geometry of the part is defined from the
real geometry so that the testing results can be
compared. It is similar to the fatigue test sample,
only the numerical values (radius, length) vary
for tension compression or torsion loading. The
difficulty is to define a thin mesh near the surface
(0.20.4 mm, or 0.0080.016 in.) where the residual stresses are introduced while maintaining
a normal mesh in other zones. Figure 24 shows
the three-dimensional mesh of the part.
Introduction of the Initial Field of Residual

22 / Effect of Materials and Processing


100

100
Axial stresses
under tensile loading

100

Measurement
Calculated

200

Tangential stresses
under tensile loading

0
Residual stresses, MPa

Residual stresses, MPa

300
400
500

Measurement
Calculated

100
200
300
400
500

600
600
0.05

0.10

0.15

0.20

0.25

0.30

0.05

0.10

0.20

0.25

0.30

0.4

0.5

0.6

100

100
Axial stresses
under torsion loading

100

0
Residual stresses, MPa

0
Residual stresses, MPa

0.15
Depth, mm

Depth, mm

Measurement
Calculated

200
300
400
500
600

Tangential stresses
under torsion loading
Measurement
Calculated

100
200
300
400
500
600

700
0

0.1

0.2

0.3

0.4

0.5

0.6

0.1

Depth, mm

Fig. 25

0.3
Depth, mm

Initial distribution of the residual stresses

Stress. The initial residual stress distribution is


calculated from the residual stresses measured
by the x-ray diffraction method in the upper
layer where initial stresses are introduced by surface treatment. In depth, the stresses are calculated for the structural equilibrium. Figure 25
shows the comparison of initial residual stresses
between the calculation and the experiment.
Prediction of the Residual Stress Relaxation. In order to predict the relaxation of the residual stresses, a simplified method proposed by
Lu et al. has been used (Ref 30). It supposes that
the relaxation of the residual stresses depends on

elastic-plastic properties of material, because the


plastic deformation is the main cause of the relaxation. So, the cyclic behaviors of material are
very important for the prediction of the residual
stresses. However, they can be measured by experiment.
Figure 26 shows the cyclic behaviors of
35NCD16 grade steel, a cyclic softening material.
When the stabilized residual stresses are calculated, the behavior corresponding to NR /2 is
used. After an elastic-plastic calculation with
FEM, the relaxation of the residual stresses un-

Rec(MPa) = 920 55,14 log N


1 cycle
2
3
15
50
150

/2, MPa

1000
900

NR/2

800
700
600

35 NCD 16
(Rm (UTS) = 110 MPa)

0.5

0.75

N,
cycles
1
2
3
15
50
150
NR/2

Re(Y.S.),
MPa
940
920
910
870
830
800
660

t/2, %

Fig. 26

0.2

The cyclic behavior of 35NCD16 grade steel. Re, yield stress; R ec, cyclic yield stress as a function of number
of cycles; Det, axial plastic strain range during cyclic loading

der the cyclic loading can be obtained. Part of


the results and their comparison with the test values have been shown in Fig. 27. In the case of
the traction-compression loading, two loading
levels were calculated. The same level for torsion loading with two different cycle numbers
was also analyzed. Figure 28 shows the stabilized residual stresses for different traction-compression load levels that are used in the prediction of the fatigue life. Unfortunately, it is
impossible to compare with the experiment results.
Prediction of the Fatigue Life. For predicting the fatigue life under a complex stress state,
it is necessary to have two S-N curves. One is
for simple traction or traction compression; another is for alternated torsion. These curves are
basic data of a material, and they can be obtained
by experimental method.
Using the program FATIGUE3D, a design
tool on fatigue developed in the LASMIS laboratory, a distribution of fatigue life and a distribution of safety factor for a structure can be obtained. In this example, the fatigue life under
traction-compression cyclic loading and torsion
loading has been calculated. The stabilized residual stresses play a role as static load. Figure
29 shows the results of fatigue life and their comparison with the experimental results. It is very
clear that the proposed method is available.
Prediction of Admissible Residual Stress.
In this approach, a calculation method has been
developed that can predict the admissible residual stress for a given fatigue life. This tool can

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 23
100

100
103 cycles, traction compression 550 MPa

100

102 cycles, traction compression 700 MPa

Axial measurement
Axial calculated
Tangential measurement
Tangential calculated

Residual stresses, MPa

Residual stresses, MPa

200
300
400

Axial measurement
Axial calculated
Tangential measurement
Tangential calculated

100
200
300
400

500

500
0

0.10

0.05

0.15

0.20

0.25

0.30

0.05

0.10

0.15

Depth, mm

105 cycles, torsion 380 MPa

Residual stresses, MPa

Residual stresses, MPa

100
200
300
400

Axial measurement
Axial calculated
Tangential measurement
Tangential calculated

0
100
200
300
400
500
600

500
0

0.1

0.2

0.3

0.4

0.5

0.1

Fig. 27

life. In other zones, tensile residual stress is admissible.


The fatigue life can then be used to deduce a
stabilized residual stress. An example of defining
the quenching conditions according to the residual stress field obtained before relaxation is analyzed.
An inverse technique designed to obtain the
quenching conditions is then developed (Ref 46,
47). In this technique, it is necessary to simulate
quenching using the FEM. To simulate quench-

100
Tangential residual stress, MPa

100
Traction-compression
550 MPa
600 MPa
620 MPa
650 MPa
700 MPa

200

300

Traction-compression
550 MPa
600 MPa
620 MPa
650 MPa
700 MPa

0
100
200
300
400
500

0.05

0.10

0.15
Depth, mm

Fig. 28

0.3

0.4

0.5

The relaxation results of the residual stresses

be used during the design phase to evaluate


whether or not residual compressive stress needs
to be introduced. The computer code also indicates the level of the residual stress and the zone
in which it needs to be introduced. In this way,
the treatment condition during the mechanical
design phase can be defined. Figure 30 shows a
map of the prestress zone indicated for a notch
sample. It can be seen that in the high stress concentration area, residual compressive stress must
be introduced to obtain a predetermined fatigue

0.2

Depth, mm

Depth, mm

Axial residual stresses, MPa

0.30

102 cycles, torsion 380 MPa


100

Axial measurement
Axial calculated
Tangential measurement
Tangential calculated

100

0.25

200

100

0.20

Depth, mm

0.20

0.25

0.30

0.05

0.10

0.15
Depth, mm

The stabilized residual stresses under different traction-compression loading cases

0.20

0.25

0.30

ing under different conditions, the physical and


mechanical data of the material need to be
known. Generally speaking, these parameters
depend on the temperature, but the surface heattransfer coefficient depends not only on the temperature, but also the sample geometry, the quenchant, and the quenching temperature. In order
to solve the problem, it is considered that there
is no change in any of the quenching parameters
or the sample geometry. Only the quenching
temperature changes. Quenching at different
temperatures was carried out. This gives the variation of the surface heat-transfer coefficient as a
function of the temperature for each quenching
temperature. It should be mentioned that numerical methods alone are not sufficient to obtain
the heat-transfer coefficient. The temperature
must be measured during quenching. Since the
heat-transfer-coefficient curve is known, the parameters of these curves as a function of the
quenching temperature can be defined. Once this
coefficient is known, quenching can be modeled
at different temperatures. The quenching temperature in the case of water quenching varies
between 20 and 80 C (68 and 176 F). The lower
the quenching temperature, the higher the level
of the residual stress field introduced. Initially,
the residual stress field obtained was found to be
lower than at 20 C (68 F) and higher than at

750
Without residual stresses
Experimental
Numerical

700
650
600
550
500

Without residual stresses


Experimental
Numerical

420
400
380
360
340
320

450
105

Fig. 29

440
Torsion load, MPa

Traction-compressionn load, MPa

24 / Effect of Materials and Processing

106
107
Fatigue life N cycles

300
105

106
107
Fatigue life N cycles

Summary

Results of fatigue life

S11

Value
2.24E+02
4.24E+01
+1.39E+02
+3.20E+02
+5.01E+02
+6.82E+02

1
3
2

Fig. 30

Prediction of admissible residual stress (or the prestressed area) for a notched sample under bending loading,
using the finite element method

Mechanical component
development

Introduction of a surface
and prestress treatment

Optimization of geometry
according to the choice
of manufacturing process
No
Stress analysis by
a finite element method
Adjustment of a
finite element model

Is fatigue life
in accordance
with
specifications?

Yes
Stop

Experimental checking of
stress: rapid prototyping

No

Fig. 31

Are results
coherent?

Yes

80 C (176 F). After this condition has been


verified, the quenching temperature can be determined by varying the quenching temperature
during simulation. After each simulation, the
calculated residual stress field is compared to
that obtained from the fatigue life. A limit condition can be defined to improve the precision of
the results. When comparing the residual stress
fields, the hydrostatic pressure was used at each
point.

Fatigue life
estimation

Various connections between residual stress-integrated design and other sectors that use the concurrent engineering approach

Residual stress plays a very important role


with respect to the different properties of materials. The gain obtained from the presence of residual stress can be enormous. This article attempts to show the effects of residual stress
through the example of fatigue strength. Here, it
has been shown that it is now possible to predict
the fatigue strength of materials, taking residual
stress into account. The results of this study
show that it is possible to predict the residual
stress relaxation and fatigue life, with consideration of the influence of residual stress by the
FEM. It has been found that the calculated results of fatigue life at surface agree very well
with experimental results.
Although the author is not in a position to provide the same type of calculation tools for other
properties, such as corrosion resistance and the
adhesion of coatings, it is now reasonable to expect that the notion of residual stress will be
gradually introduced into the design stage of mechanical parts. Numerical modeling of the behavior beforehand saves a considerable amount
of time because of the reduction in the number
of experimental tests required. These tests are
often very long and costly, but they have proved
to be indispensable. The problem of taking residual stress into account at the design stage will
become more and more critical with the development of new materials (multimaterials, etc.)
and new treatments (combined treatments, etc.)
With the development of different experimental and numerical techniques, it is now possible
to introduce residual stress into the design office
for the integrated design of mechanical components, thus offering a new concurrent engineering approach applied to the design of mechanical
components taking residual and applied stress
into consideration. Figure 31 shows the different
connections between residual stress-integrated
design and other sectors that use the concurrent
engineering approach. A mechanical component
designer can simulate dynamic characteristics,
material processing, and product life.
Consideration of residual stress is becoming
increasingly important for two reasons: the introduction of multimaterials that induce residual
stress, and the need for the designer to reduce
the weight of components in order to remain
competitive. Basic research has brought a better
understanding of the phenomena relating to residual stress.
The main aim is to develop an integrated qual-

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 25
ity control tool. For industrial applications, future developments are necessary:
Measurement techniques

grant Brite-EuRam, BRRT-CT985090, ENSPED project), and the National Science Foundation of China (two bases project) is acknowledged.

Improvement of ultrasonic and magnetic

measurement methods
Integration of the portable optical method for
strain measurement in destructive techniques
Processing and materials
Development for the industrialization of new

prestress processes: ultrasonic shot peening,


laser shock, surface nanocrystallization, and
so on.
Optimization of residual stress in advanced
materials: functional gradient coating system,
for example, plasma spray coating, PVD,
CVD, or diamond-like carbon (DLC) coating;
metal-matrix composite; electronic packaging; intermetallic materials
Modeling
Development of advanced tools for the pre-

diction of initial residual stress in various processes: forging, casting, machining, welding,
plasma spray coating, composite processing,
electronic packaging manufacturing, mechanical surface treatment
Integration of the residual stress modeling
method into a global life design (3-D) and the
creation of optimization codes for the design
of mechanical components using the concurrent engineering approach
With the recent progress in the field of research and development, it is now possible to
introduce a new area of research: prestress engineering of structural materials (PESM or
PRESMA). In the near future, the mechanical
engineer will be able to use these tools in a prestressed concrete approach to civil engineering.
The ENSPED program in Europe is an excellent
example of the cooperation of industrial and research partners. Using this new approach, the fatigue behavior and mechanical behavior of new
materials can be considerably improved, increasing two- or threefold in some cases. In fact, many
failures occur in the area close to the surface.
Improving the mechanical behavior is often only
necessary on the surface (fatigue, fretting fatigue, wear, stress corrosion). As a result, this
type of approach can be expected to develop
very quickly in order to introduce adequate prestressing treatment in the area indicated by the
advanced design tool.
ACKNOWLEDGMENTS
The author is grateful to all the people at LASMIS working in the field of residual stress for
their help during the preparation of this paper:
Prof. J.L. Chaboche, Dr. D. Retraint, Dr. E. Rouhaud, Dr. X.L. Gong, Dr. L. Couturier, Dr. R.
Akrache, Dr. Z. Wu, Dr. S. Rasouli Yazdi, Mr.
A. Milley, Mr. D. Deslaf, Mr. B. Guelorget, Mr.
A. Voinier, Mr. F. Belahcene, Mr. K. Gong. Financial support from the European Union (under

12.
REFERENCES
1. J. Lu, Ed., Handbook of Measurement of Residual Stresses, Society for Experimental
Mechanics, USA, 1996
2. J. Lu, A. Niku-Lari, and J.F. Flavenot, Latest Development of Residual Stresses Measurement by the Hole Drilling Method, Materiaux et Techniques, Dec 1985, p 709718
3. Z. Wu, J. Lu, and B. Han, Study of Residual
Stress Distribution by a Combined Method
of Moire Interferometry and Incremental
Hole Drilling. Part I: Theory, J. Appl.
Mech., ASME, Sept 1998
4. Z. Wu, J. Lu, and B. Han, Study of Residual
Stress Distribution by a Combined Method
of Moire Interferometry and Incremental
Hole Drilling. Part II: Implementation of
Applied Mechanics, J. Appl. Mech., ASME,
Sept 1998
5. Z. Wu and J. Lu, Residual Stresses by Moire
Interferometry and Incremental Hole Drilling, Experimental Mechanics, Advances in
Design, Testing and Analysis, Vol 2, I.M.
Allison, Ed., Conf. ICEM 98, 2428 Aug
1998, (Oxford), p 13191324
6. Z. Wu, X.K. Niu, J. Lu, and P.G. Ifju, Study
of Process-Induced Stress in Orthotropic
Composite LaminateCarbon/Epoxy [O2 /
902]2s, Proceedings of the SEM Spring
Conference on Experimental and Applied
Mechanics, 13 June 1998, (Houston,
Texas), p 179181
7. Z. Wu, J. Lu, and Y. Guo, A New Method
of Residual Stress Measurement on Electronic Packaging, Experimental Mechanics,
Advances in Design, Testing and Analysis,
Vol 2, I.M. Allison, Ed., Conf. ICEM 98,
2428 Aug 1998, (Oxford), p 987991
8. J. Lu, J.F. Flavenot, and S. Thery, Study on
the Effect of the Finishing Treatment on the
Residual Stress Gradient in SiC Reinforced
Aluminum Metal Matrix Composite, J.
Compos. Technol. Res., ASTM, Vol 12 (No.
4), 1990, p 232238
9. J. Li, J. Lu, M. Perrin, M. Ceretti, and A.
Lodini, Study on the Residual Stress in
Cold-Rolled 7075 Al-SiC Whisker Reinforced Composites by X-ray and Neutron
Diffraction, J. Compos. Technol. Res., July
1995, p 194198
10. J. Lu, P. Peyre, C. Omam Nonga, A. Benamar, and J.F. Flavenot, Mechanical Surface
Treatments, Current Trends, and Future
Prospects, Surface Modification Technologies VIII, T. S. Sudarshan and M. Jeandin,
Ed., TMS, Sept 1994, p 589601
11. A. Benamar, J. Lu, J.F. Flavenot, P. Barbarin, G. Chalant, and G. Inglebert, Modeling Residual Stresses by Cold Rolling an

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

Austenitic Stainless Steel, Residual Stress,


V. Hauk, H.P. Hougardy, E. Macherauch,
and H.D. Tietz, Ed., DGM, Verlag, 1992, p
891900
J.B. Roelens, F. Maltrud, and J. Lu, Determination of Residual Stresses in Submerged
Arc Multi-Pass Welds by Means of Numerical Simulation and Comparison with Experimental Measurements, Welding in the
World, Vol 33 (No. 3), 1994, p 152159
Proceedings of the National Conference on
the Residual Stress and Design Office, CETIM, (Senlis), 1991
D. Deslaef, E. Rouhaud, and J. Lu, Finite
Element Modeling of Shot-Peening Processes, Proceedings of the SEM Spring Conference on Experimental and Applied Mechanics, 13 June 1998, (Houston, Texas),
p 419422
A. Chabenat, and J.F. Flavenot, Measurement of Residual Stress on the Quenched
Axles after Heating by Induction, Traitement thermique/Revue de metallurgie,
(Heat treatment/Journal of Metallurgy), No.
124, 1978, p 233239 (in French)
N. Skalli, and J.F. Flavenot, Prediction of
Fatigue Life by Simulation with the Residual Stress Considerations, Conference proceedings, Spring Conference, SF2M
(French Society for Metal and Materials),
2223 May 1984, p 98117 (in French)
H.P. Lieurade, P. Castelucci, J.F. Flavenot,
and J. Lu, Efficiency of Improvement Techniques on the Fatigue Strength as a Function
of the Type of Welded Joint, Welding in the
World, Vol 31 (No. 4), 1991, p 268271 (in
French)
H.P. Lieurade, C. Maillard-Salin, and M.
Truchon, Fatigue Cracking of Welded Joints
of High Strength Steel, Proceedings of
IABSE, Colloquium, Lausanne, 1982, p
137144 (in French)
D. Francois, Role of the Residual Stress on
the Fracture, Conference Proceedings, Residual Stress in Welding Structure, CETIM,
SENLIS, Dec 1987, p 8797 (in French)
W.H. Friske, Shot Peening to Prevent the
Corrosion of Austenitic Stainless Steels,
Rockwell International Cooperation, 1975
A. Niku-Lan`, M. Menier, and G. Beranger,
Effect of the Surface State on the Stress
Corrosion Behaviour of a Stainless Steel,
CETIM Informations, No. 78, Dec 1982
C. Richard, J. Lu, J.F. Flavenot, and G. Beranger, Study of Residual Stress in the
Plasma Sprayed NiCrAlY Coating and
Characterization of Interface Toughness,
Memoires et Etudes Scientifiques, Revue de
metallurgie, May 1991, p 295306 (in
French)
C.S. Richard, G. Beranger, J. Lu, and J.F.
Flavenot, The Influence of Heat Treatments
on the Adhesion of Plasma Sprayed NiCrAlY Coatings, Surf. Coat. Technol., Vol
82, 1996, p 99109

26 / Effect of Materials and Processing


24. C.S. Richard, G. Beranger, J. Lu, J.F. Flavenot, and T. Gregoire, Four-Point Bending
Tests of Thermally Produced WC-Co Coating, Surface and Coatings Technology, (No.
78), 1996, p 284294
25. C.S. Richard, J. Lu, G. Beranger, and F. Decomps, Study of Cr2O3 Coating Materials,
Part II: Adhesion to a Cast Iron Substrate,
J. Therm. Spray Technol., ASM International, Dec 1995, Vol 5 (No. 4), p 347352
26. R.J. Arsenault, Tensile and Compressive
Properties of Metal Matrix Composites,
Metal Matrix Composites: Mechanisms and
Properties, R.K. Evertt and R.J. Arsenault,
Ed., p 133167
27. P. Balladon, Stress Relieve After Welding,
Criterion and Process, Conference Proceedings, Residual Stress in Welding Structure,
CETIM, Senlis, Dec 1987, p 4569 (in
French)
28. R. Lematre, J.L. Lebrun, and J. Maeder,
Residual Stress and Fatigue, Materiaux et
Techniques, Sept 1982 (in French)
29. J. Lu, J.F. Flavenot, A. Turbat, and D. Francois, Modeling of the Relaxation of Residual Stress under Fatigue Loading, Conference proceedings, Residual Stress at Design
Office, CETIM, Senlis, Nov 1991, p 7198
(in French)
30. J. Lu, J.F. Flavenot, and A. Turbat, Prediction of the Residual Stress Relaxation During Fatigue, Mechanical Relaxation of Residual Stresses, ASTM STP993, 1998, p
7590
31. J. Lu, J.F. Flavenot, A. Diboine, S. Lasserre,
C. Froustey, M. Bennebach, and T. PalinLuc, Development of a General Multiaxial
Fatigue Criterion for High Cycles of Fatigue
Behavior Prediction, Multiaxial Fatigue
and Design, ESIS 21, A. Pineau, G. Cail-

32.

33.

34.

35.

36.

37.
38.
39.

40.

letaud, and T.C. Lindley, Ed., Mechanical


Engineering Publications, London, 1996, p
477487
K. Dang Van, On the Fatigue of Metals, Science Technique Armement, Memorial de
lartillerie francaise, 3ie`me Fascicule, 1973,
p 647722 (in French)
B. Crossland, Effect of Large Hydrostatic
Pressure on the Torsional Fatigue Strength
of an Alloy Steel, Int. Conf. on Fatigue of
Metals, IME/ASME, (London), 1956, p
138149
G. Sines, Fatigue Criteria under Combined
Stress or Strains, Transactions ASME, J.
Eng. Mater. Technol., Vol 13, April 1981,
p 8290
H. Kakuno and Y. Kawada, A New Criterion of Fatigue Strength of a Round Bar
Subjected to Combined Static and Repeated
Bending and Torsion, Fatigue of Engineering Materials and Structures, Vol 2, 1979,
p 229236
I.V. Papadopoulos, Cyclic Fatigue of
Metal, a New Approach, Ph.D. dissertation, Ecole Nationale des Ponts et Chausse`es, France, Dec 1987 (in French)
A. Galtier and J. Seguret, Multiaxial Fatigue
Criterion, Revue Francaise de Mecanique,
No. 4, 1990, p 291299 (in French)
A. Deperrois, On the Fatigue Limit Calculation, Ph.D. dissertation, Ecole Polytechnique, France, June 1991 (in French)
R. Akrache and J. Lu, 3D Calculation of
High Cycle Fatigue Life under Out-of-Phase
Multiaxial Loading, Fatigue and Fracture
of Engineering Materials and Structures,
Vol 22, July 1999, p 527534
J. Lu and J.F. Flavenot, Prediction of the
Residual Stress Relaxation During Fatigue
Loading and Taking the Residual Stress in

41.

42.

43.
44.
45.

46.

47.

Multiaxial Fatigue Criteria into Account,


Proceedings of 2nd International Conference on Residual Stress, (Nancy, France),
Elsevier Applied Science, Beck et al., Ed.,
Nov 1988, p 784790
W.P. Evans, R.E. Ricklefs, and J.F. Millian,
Local Atomic Arrangements Studied by Xray Diffraction, J.B. Cohen and J.E. Hillard,
Ed., Proc. of a Symposium, (Chicago), Feb
1965, Chap. 11
A. Bignonnet, Fatigue Behavior of a Shot
Peened 35NCD16 Grade Steel, Evolution of
Residual Stress Due to the Shot Peening as
a Function of the Type of the Loading, Proceedings of Ninth Conference on Shot Peening, CETIM, Senlis, France, Nov 1985, p
107116 (in French)
FATIGUE3D, User Manual, UTT-LASMIS, Jan 2000
ABAQUS Theory Manual, Hibitt, Karlsson,
and Sorensen, Inc., 1996
A. Milley, E. Rouhaud, and J. Lu, Prediction
of Residual Stress Relaxation in Three-Dimensional Structures under Tensile and
Torsional Loadings, Fifth International
Conference on Residual Stress, Vol 1, T. Ericsson, Ed., (Linkoping), 1618 June 1997,
p 424429
S. Rasouli Yazdi and J. Lu, Simulation of
Quenching and Fatigue Relaxation of Residual Stresses in Aluminium Parts, Fifth International Conference on Residual Stress,
Vol 1, T. Ericsson, Ed., (Linkoping), 1618
June 1997, p 490495
S. Rasouli Yazdi and J. Lu, Inverse Method
for Determining the Quenching Conditions
Using a Residual Stress Field, Proceedings
of the Third Int. Conf. on Quenching Control and Distorsion, ASM, (Prague), March
1999

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p27-53
DOI: 10.1361/hrsd2002p027

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Residual Stresses and Fatigue Behavior


D. Lohe, K.-H. Lang, O. Vohringer
University of Karlsruhe, Germany

COMPONENTS MAY FAIL DURING SERVICE in very different ways. Important aspects
are:
Onset of plastic deformation or exceeding of

allowable plastic deformation

Crack initiation or exceeding of allowable

crack length

Unstable crack propagation that results in

fracture of the component

Instability such as buckling, which may result

in elastic or plastic collapse and, hence, catastrophic failure


In practice, failure of a component is almost
always determined by the interaction of several
external parameters, such as unexpected high
static and/or cyclic loads, friction, thermal energy, oxidation, corrosion, and so forth. On the
other hand, the component itself may contribute
to premature failure, if there is confusion of the
component material, wrong heat treatment, surface decarburization, and so forth.
The residual stress state in a component may
be one of the most important parameters influencing its service behavior, particularly regarding high-strength material states. Residual
stresses may reduce yield or elastic collapse
loads and may promote corrosion cracking (Ref
1, 2). With respect to fatigue, which is the topic
of this article, residual stresses may alter the cyclic deformation behavior, promote or retard
crack initiation, accelerate or decelerate crack
propagation, and may be beneficial or detrimental to finite fatigue life and the endurance limit.
It is important to realize that in a component locally very high amounts of residual stresses may
exist. For example, in a spring steel up to 2000
MPa compressive residual stresses may be produced by a combined warm and stress peening
process (Ref 3).
In practice, no component is free of micro residual stresses. Almost all components have
macro residual stresses to a certain extent. Only
a small number of components exist where
macro residual stresses are negligible, for example after a suitable stress relaxation heat treatment. The reasons for residual stress formation
in components are treated in other articles in this
Handbook. The most important ones are the various manufacturing processes. The fundamental

mechanisms of residual stress formation are inhomogeneous plastic deformation caused by mechanical, thermomechanical, or thermal attack
and the development of constraint between different constituents of a component such as
phases or coatings during the formation of the
constituents (e.g., during the coating process)
and/or during cooling from a higher process
temperature (e.g., the sintering temperature).
Manufacturing processes not only result in the
formation of definite macro residual stress states,
but may also effect other changes of the component state close to its surface:
Formation of a specific topography
Work-hardening or (in very hard material

states) work-softening processes and hence


alteration of the micro residual stress state
Phase transformation
Crack initiation
All these processes may change the fatigue behavior of a given component. It is hardly possible to completely separate these influences, even
though this is very important for the understanding of the existing relationships.

Residual Stress
Stability of Residual Stresses. For the assessment of the influence of residual stresses on
fatigue behavior, the stability of the residual
stress state is of utmost importance. The reader
is referred to the article Stability of Residual
Stresses in this Handbook, which deals with the
relaxation of residual stresses due to monotonic
or cycling loading or due to thermal energy. It
is shown there that the micro residual stress state
is more stable against mechanically and thermally induced relaxation than the macro residual
stress state. On the other hand, it is also proved
that during fatigue the micro residual stress state
of a given steel may be changed by cyclic hardening and/or cyclic softening processes, which
are closely related to cyclic plastic deformations.
Hence, during fatigue loading of a component
with locally varying macro as well as micro residual stress states, complex interactions of the
macro residual stress state with the cyclic loading stresses and the micro residual stress state

with work-hardening and/or work-softening


processes occur. In both cases, the amount of
cyclic plastic deformation is the most important
parameter. Since during fatigue loadings that result in technically relevant lifetimes or in infinite
life, the cyclic plastic deformations decrease
with increasing hardness, it is generally expected
that the influence of macro residual stresses is
lower in low-strength steels than in high-strength
steels.
Comparison between Loading Stresses and
Residual Stresses. Since the alteration of the
micro residual stress state caused by manufacturing processes changes the local hardness
within a component and hence the local strength,
it is convenient to comprise the corresponding
influence on the fatigue behavior of the component as a change of the local fatigue resistance
of the steel used for the component.
On the other hand, stable macro residual
stresses are static stresses and may be regarded
as locally varying mean stresses. However, it is
important to realize that there are definite differences between loading mean stresses and residual stresses:
Loading mean stresses exist as a consequence

of acting external forces, moments, pressure


and/or internal pressure, and eventually temperature gradients. Residual stresses exist as
a consequence of inhomogeneous plastic deformation and/or as a consequence of constraints between different constituents of a
component.
Distribution and sign of loading mean
stresses depend on the external loading, the
geometry of the component, the elastic-plastic deformation behavior of the material (if
maximum loading stresses exceed the yield
strength or the cyclic yield strength), and on
thermophysical material properties (if temperature gradients exist). Distribution and
sign of residual stresses depend on the
amount and extension of inhomogeneous
plastic deformation in relation to the whole
component volume, on the interaction of mechanically and/or thermally driven deformation during manufacturing processes, on the
strength of the material state, and/or on the
volume fraction, the elastic as well as the

28 / Effect of Materials and Processing

Stress amplitude (a), MPa

thermophysical properties of the constituents


of the component that impose constraint on
each other.
Loading mean stresses are in equilibrium
with external forces and moments. Residual
stresses are in equilibrium with themselves
regarding balance of forces and moments
with relation to any sectional area and axis,
respectively. The latter is also true for thermally induced loading stresses.
Loading mean stresses may be changed and
finally disappear, if external loadings and/or

Nf
Ni
Ns

Number of cycles, N

Fig. 1

Wohler curves for slip band formation (Ns), microcrack initiation (Ni), and failure (Nf). Source:

Plastic strain amplitude (a,p), %

Ref 5

0.4
0.3
0.2
0.1
0
1

10

102

103

104

105

Plastic strain amplitude (a,p), %

Some Aspects of Fatigue of Steels

Plastic strain amplitude versus number of cycles


of normalized AISI 4140 steel. Source: Ref 68

The fatigue behavior of metallic materials is


characterized by different processes. Generally,
it contains four successive stages A, B, C, and
D:

0.15
Inhomogeneous
softening

Homogeneous
hardening

0.10

0.05

01

10
102
Number of cycles, N

Cyclic deformation curve and development of


inhomogeneous strain distribution during cyclic
softening of normalized AISI 4140 steel. Source: Ref 68

Fig. 3

The interaction of the residual stress state with


the cyclic loading stresses determines the fatigue
behavior of components. The following sections
of this article discuss the consequences of this
interaction on cyclic deformation behavior, on
crack initiation and crack propagation, as well as
on the fatigue life and on the endurance limit.
Throughout this article, no special account is
given for multiaxial loading or residual stress
states and equivalent stresses are not used. Instead, this article considers simple loading states
and those components of the residual stress
statewhich is of course always multiaxial
that are relevant for the interaction of the cyclic
loading and the residual stress state under consideration. With regard to notched specimens
in which a multiaxial stress state always exists
upon mechanical loadingthe treatment is
based almost entirely on nominal stresses and
stress amplitudes. Hence, the fatigue strengths
and the cyclic loadings corresponding to finite
fatigue lives given in this article are nominal
stress amplitudes. First of all, the section Some
Aspects of Fatigue of Steels introduces some
basic definitions and relations of fatigue. This is
done here rather briefly. The phenomenon and
the processes of the fatigue of steels are presented in general and in detail in Volume 19 of
the ASM Handbook (Ref 4).

106

Number of cycles, N

Fig. 2

if existing temperature gradients are changed


and finally removed. Residual stresses may be
changed and eventually disappear by thermally and/or mechanically induced relaxation.
Loading mean stresses are not influenced by
cyclic plastic deformation in stress-controlled
loading, unless stress gradients become very
highfor example in the root of sharp
notches, where mean stress redistribution
may occur. Residual stresses always relax, if
the cyclic loading exceeds certain threshold
values (see the article Stability of Residual
Stresses in this Handbook).

Stage A: Cyclic deformation with work-hard-

ening or/and work-softening effects in the total volume (homogeneous loading such as
push-pull of smooth specimens) or in the
highest loaded regions (inhomogeneous loading such as bending, torsion, or any loading
of notched components) and development of
persistent slip bands at the surface
Stage B: Microcrack initiation and propagation. Normal evolution of a macrocrack in the
surface region
Stage C: Stable macrocrack propagation connected with changes in the state of the material in the crack tip plastic zone

Stage D: Unstable crack propagation and fail-

ure
An example of the consequences of these fatigue
stages is shown in Fig. 1 in an S-N diagram
(Wohler diagram) for the plain carbon steel SAE
1020 (German grade C 20), which presents experimental results for persistent slip band formation (ra /log Ns curve), microcrack initiation
(ra /log Ni curve), and failure (ra /log Nf curve)
(Ref 5).
Cyclic Deformation Behavior. Investigations on the cyclic deformation behavior in stage
A using computerized servohydraulic test systems enable the determination of the complete
stress/total-strain response at distinct numbers
of cycles that yields to hysteresis loops. In a
stress-controlled test, the plastic strain amplitude
ea,p plotted versus log N results in a cyclic deformation curve that corresponds to the microstructural changes in the material during cyclic
loading. Decreasing (increasing) ea,p values with
an increasing number of cycles N are typical for
a cyclic work-hardening (work-softening) behavior.
Characteristic cyclic deformation curves of
the low-alloy steel AISI 4140 (German grade 42
CrMo 4) in a normalized state are shown in Fig.
2. During stress-controlled push-pull tests with
R rmin/rmax 1 and ra  RP0.2, the plastic
strain amplitudes systematically change with the
stress amplitudes applied. In the initial stage of
cyclic loading, a quasi-linear elastic behavior is
found. This stage is followed by a period of increasing ea,p values due to cyclically induced
work-softening effects that are connected with
inhomogeneous deformation processes caused
by fatigue Luders bands. This can be proved with
photoelastic investigations. A typical result obtained for this steel is presented in Fig. 3. Plastically deformed areas appertaining to the individual stages of inhomogeneous cyclic
deformation were registered photographically
and marked as hatched areas on the specimens
that are allied to the cyclic deformation curve.
Similar macroscopic softening effects occur in
all normalized states of plain carbon steels and
low-alloy steels with carbon contents below 0.6
wt% (Ref 68).
The quasi-linear elastic period shortens and
the extent of cyclic work softening grows simultaneously with increasing stress amplitude
ra. After reaching a maximum of softening, the
materials behavior is mainly characterized by
work hardening, and the ea,p values decrease
with increasing number of cycles. At the end of
the specimens life, the plastic strain amplitude
increases again, however fictitiously as a consequence of changes in the compliance of the
specimens due to crack propagation. The plot of
ra versus ea,t at N Nf /2, the so-called cyclic
stress-strain curve, of this steel is presented in
Fig. 4 together with the monotonic-stress/totalstrain curve at nearly the same strain rate e . In the
elastic-plastic range, the cyclic curve is lower
than the monotonic curve due to the work softening. The cyclic stress-strain curves of plain

Residual Stresses and Fatigue Behavior / 29

Monotonic:
 0.4 101s1

400
300

Cyclic:
 0.3 101s1
N = Nf/2

200
100
0
0

0.1
0.3
0.2
Total strain (t) or total strain amplitude (a,t), %

Monotonic stress-strain curve and cyclic stressstrain curve at Nf /2 of normalized AISI 4140 steel
at nearly the same strain rates. Source: Ref 68

Stress amplitude (a), MPa

Fig. 4

0.8 wt% C

0.45

400

0.22

300

0.01
200

100
0

0.1

0.2

0.3

Plastic strain amplitude (a,p), %

810

750

0.2
0.1

700

600

500

Fig. 6
68

a = 850 MPa

0.3

10

102
103
104
Number of cycles, N

105

 (one) 101s1

800

Cyclic:
 0.6 101s1
N = Nf/2

600
400
200
0

10

Total strain (t) or total strain amplitude (a,t), %


Monotonic stress-strain curve and cyclic stressstrain curve at Nf /2 of quenched-and-tempered
AISI 4140 steel at nearly the same strain rates. Source: Ref
68

Fig. 7

Fig. 5 Cyclic stress-strain curves of plain carbon steels


with the indicated carbon contents (wt%) in normalized conditions. Ref 810

0.4

Monotonic:

1000

0.4

Plastic strain amplitude (a,p), %

0.5

Stress () or stress amplitude (a), MPa

500

ments is characterized by a continuous increase


of the plastic strain amplitude that lasts up to
failure. The cyclic softening sets in after an incubation number of cycles Ni, which depend on
the applied loading amplitude. Figure 6 shows
the cyclic deformation curves ea,p versus log N
at different stress amplitudes for AISI 4140 tempered at 570 C. As a consequence of this cyclic
deformation behavior, the cyclic stress-strain
curves of quenched-and-tempered steels proceed
below the monotonic curves (see Fig. 7). In this
material state, the cyclic softening is caused by
an inhomogeneous deformation behavior. The
plastic deformation develops localized in distinct
zones and remains localized in these zones up to
failure. Figure 8 shows results from photoelastic
investigation carried out on quenched-and-tempered AISI 4140. In the plastically deformed
zones, the cyclic softening is caused by a rearrangement of the dislocations to energetically
more favorable dislocation structures such as cell
structures. As a consequence, the stored strain
energy and the hardness decrease. Within these
plastically deformed zones, microcrack initiation
and propagation occur (Ref 68).
Crack Initiation. The plastic strains generated by a cyclical loading lead to the formation
of microcracks in the material after a certain

0.3

0.2

a = 750 MPa
Inhomogeneous
softening

0.1

0
10

106

Plastic strain amplitude versus number of cycles


of quenched-and-tempered AISI 4140 steel. Ref

number of cycles. These fatigue cracks usually


develop at the surface. Only in special cases
for example, very heterogeneous materials such
as case-hardened steelor at special loading
conditionssuch as Hertzian pressurefatigue
cracks may initiate below the surface. The first
microscopic cracks can appear in a very early
stage of the cyclic loading. Depending on the
loading conditions and the material, microcracks
can already be detected after 0.1 to 10% of the
lifetime and number of cycles to failure, respectively.
In ductile materials, changes of the surface
structure precede fatigue crack initiation. Fatigue
slip bands (persistent slip bands, PSBs), which
often are connected to extrusions and intrusions,
arise from irreversible gliding. These persistent
slip bands are often places in which first fatigue
cracks initiate. The crack initiation at persistent
slip bands often is typically for lower to middle
loading amplitudes. At higher amplitudes, cracks
may also initiate at grain boundaries. Crack initiation and the first phase of crack propagation
in ductile materials are controlled by shear
stresses. Within the first grain layers, the cracks
grow at an angle of 45 to the load axis. This
phase is called stage I crack growth. After reaching a certain length, the cracks become the socalled stage II cracks. Further propagation of theses cracks is controlled by normal stresses. Stage
II cracks grow approximately rectangular to the
load axis. In brittle materials, crack initiation is
often observed at inclusions and other defects.
Crack Propagation. To describe the stress
and strain field in front of a crack tip the loading
details and the geometry of a cracked specimen
or component must be considered. If the plastic
zone in front of a crack tip is small in comparison
with the dimensions of the specimens or components, the linear elastic fracture mechanics are
applicable, and the local loading conditions in
front of the crack tip can be described by the
stress intensity factor, K. Therefore, Paris,
Gomez, and Anderson (Ref 11) considered the
stress intensity factor as a driving force for the
propagation of a crack and suggested a crack
growth equation in the form:
da
C (dK)m
dN

Plastic strain amplitude (a,p), %

Stress () or stress amplitude (a), MPa

carbon steels with 0.01, 0.22, 0.45, and 0.80 wt%


C (German grade Ck 01, Ck 22, Ck 45, and C
80) in normalized conditions, determined from
cyclic deformation curves at N Nf /2, are
drawn up in Fig. 5. With increasing carbon content and pearlite content, respectively, an increasing stress amplitude at ea,p constant occurs due to a phase-hardening mechanism (Ref
810).
The cyclic deformation behavior of quenchedand-tempered steels in stress-controlled experi-

102

103

104

Number of cycles, N
Cyclic deformation curve and development of
inhomogeneous strain distribution during cyclic
softening of quenched-and-tempered AISI 4140 steel.
Source: Ref 68

Fig. 8

(Eq 1)

where da/dN is the crack growth rate, DK the


range of the mode I stress intensity factor, and
C and m are constants (Ref 12). With this equation, which is the well-known Paris law, it is
possible to describe the crack propagation in the
midrange of the crack growth rate. Figure 9
shows the qualitative features of the crack
growth behavior in a double logarithmic plot of
the growth rate versus the range of the stress
intensity factor for experiments with constant
stress range Dr rmax rmin. Distinct curves
are observed for different mean stresses rm or
values of R rmin /rmax, respectively. The
lower parts of the curves exhibit a threshold
value of DK below which crack growth is essentially arrested. Typically, the crack growth

Crack growth rate, log da/dN

30 / Effect of Materials and Processing

R1
R2
R3
Stress intensity factor range, log K

Fig. 9

Crack propagation rate versus stress intensity factor range for stress ratio values R1  R2  R3

rate in this region is below 108 mm/cycle. The


upper parts of the curves represent the crack
growth behavior as the Kmax values approach the
critical stress intensity factor KIc, which marks
the onset of unstable crack propagation. Then the
crack growth rate increases boundlessly.
The Paris law does not represent the complete
crack growth curves shown in Fig. 9. The upper
part of the curves depends on KIc, and the lower
part on the threshold value DKth. Therefore, several complementations were suggested. An equation that incorporates all of the features of crack
growth curves has been used by NASA (Ref 13).
It has the form:

Quantitatively, the following relationships are


valid:
ea,p ef N a
f

and
ea,e rf /E N b
f

(Eq 6)

with ef the fatigue ductility coefficient, rf the


fatigue strength coefficient, the fatigue ductility exponent, b the fatigue strength exponent,
and E the Youngs modulus. Rearrangement of
both equations yields to:
ea,p N fa constant

da
C (1 R)m DKn (DK DKth)p

dN
[(1 R) KIc DK]q

(Eq 5)

(Eq 7)

and
(Eq 2)
ea,e E N b
ra N b
constant
f
f

log a
log f
log Rm

log Rf

log Ns

Strain amplitude (a)

Fig. 10

log Ne log Nf

Stress Wohler curve (S-N curve)

f/ E

E N
N
Number of cycles to failure, Nf

Fig. 11

Total strain Wohler curve

This equation is able to account for the threshold


effect, the R ratio effect, and the final instability
effect. By suitable choice of the exponents, this
equation can accept a number of different variants of crack propagation laws. For m p q
0, for example, the Paris law is obtained.
The concepts of linear elastic fracture mechanics and the use of the stress intensity factor
are limited to cases in which small-scale yielding
develops at the crack tip. Moreover, the applicability of Eq 1 and 2 is restricted to sufficiently
long macroscopic cracks. Short cracks can grow
at smaller ranges of the stress intensity factor as
DKth. At comparable DK values, they can exhibit
clearly higher crack growth rates than long
cracks (Ref 14).
Lifetime Behavior. The lifetime behavior of
steels is commonly described in an S-N diagram
(Wohler diagram) (Ref 15), where the loading
amplitude is plotted versus the number of cycles
to failure in double logarithmic scaling. For
stress-controlled fatigue tests, the S-N curve is
frequently described by three straight lines (see
Fig. 10). In the range of the quasi-static strength
Rm (Nf  Ns) and the fatigue strength Rf(N 
N0f ), the endurable stress amplitude is constant
and given by the ultimate strength Rm and the
fatigue strength Rf, respectively. In the range of
finite life (Ns  Nf  Ne), the Basquin relation
(Ref 16):

Fatigue strength, Rf

ra rf N b
f

often describes the lifetime behavior satisfactorily. rf is the fatigue strength coefficient and b
the fatigue strength exponent.
In the case of total-strain-controlled fatigue
tests the ea,t values in the total strain Wohler
curve (Fig. 11) can be separated into the elastic
and plastic parts:

Re
Goodman
0

Rf

Gerber

ea,t ea,e ea,p


Re
Mean stress m

Fig. 12

Endurance limit versus mean stress

(Eq 3)

Rm

(Eq 4)

From experience it is known that both parts extend exponentially from the number of cycles to
failure Nf, which means that the plots log ea,e
versus log Nf and log ea,p versus log Nf are linear.

(Eq 8)

which are the well-known relationships from


Coffin-Manson (Ref 17, 18) (Eq 7) and Basquin
(Eq 8).
Tensile mean stresses promote crack initiation
and accelerate crack propagation. Therefore, the
fatigue strength Rf decreases with increasing
mean stresses. This correlation is normally represented in a Smith (Ref 19) or in a Haigh (Ref
20) diagram. As an example of such fatigue
strength diagrams, Fig. 12 shows a Haigh diagram. The fatigue strength Rf is plotted versus
the mean stress rm. The Goodman approximation (Ref 21):
Rf R0f (1 rm /Rm)

(Eq 9)

which represents a straight line between the fatigue strength R0f at rm 0 and the tensile
strength Rm, is frequently used to estimate fatigue strength. For practical use, the allowable
stresses are restricted by the demand that macroscopically no plastic deformation should occur. This restriction leads to a straight line connecting the yield strength Re at both axes.
Through this, the allowable stress amplitudes lie
in the hatched area in Fig. 12. The slope of the
Goodman straight line is often also described as
mean stress sensitivity M R0f rm /Rm. The
Goodman approximation normally leads to conservative estimations. Often, the Gerber parabola
(Ref 22):
Rf R0f [1 (rm /Rm)2]

(Eq 10)

is used as a nonconservative estimation of the


fatigue strength. Mostly, experimental results
range between the Goodman and the Gerber approximations.
In the range of infinite lifetime, the influence
of mean stresses can be described by damage
parameters. Two frequently used damage parameters are the one proposed by Smith, Watson,
and Topper (Ref 23):
PSWT (ra rm) ea,t E

(Eq 11)

and the one proposed by Ostergren (Ref 24):

Residual Stresses and Fatigue Behavior / 31


(Eq 12)

ra and rm are the imposed stress amplitude and


mean stress, respectively, ea,t and ea,p are the resulting total strain amplitude and plastic strain
amplitude, respectively, and E is the Youngs
modulus. Normally, for ea,t and ea,p the values at
half of the lifetime are used.

Influence of Residual Stresses on


the Cyclic Deformation Behavior

0.4
a in MPa
Unpeened
Shot peened

0.3

380

0.15

10

102

103

104

105

106

Number of cycles, N

0.4
a in MPa

450

Unpeened
Shot peened

0.3

550

600
0.05

400
375
350
350
400
275

0
1

10

(b)

102
103
104
Number of cycles, N

1.0
900
0.8

105

102
103
104
Number of cycles, N

10

a in MPa
Unpeened
800
Shot peened

750
700

0.2

500
400
550
600

(c)

10

102
103
104
Number of cycles, N

105

a in MPa
Unrolled
0.03
Deep rolled

600
650
550
600

0.02

500
0.01

550
500

0
102
103
104
Number of cycles, N

10

(b)

105

106

Cyclic deformation curves for stress-controlled


push-pull loading of (a) normalized and (b) nitrocarburized specimens of the SAE 1045 steel in unpeened
and deep-rolled conditions. Source: Ref 27

Fig. 14

0.10
Untreated
0.08
Shot peened
0.06
Deep rolled

0.04
0.02
0

106

Cyclic deformation curves for stress controlled


push-pull loading of (a) normalized, (b)
quenched-and-tempered (730 C/2 h), and (c) quenchedand-tempered (570 C/2 h) smooth specimens of the steel
AISI 4140 in unpeened and shot peened conditions.
Source: Ref 25, 26

Fig. 13

106

0.04

1
1

105

106

0.6
0.4

500
450

1
450

550

0.2
0.1

a in MPa
Unrolled
Deep rolled

510

0.10

(a)
Plastic strain amplitude (a,p), %

(a)
Plastic strain amplitude (a,p), %

600

325
300
250

0.1

Plastic strain amplitude (a,p), %

0.20

350

0.2

Plastic strain amplitude (a,p), %

Plastic strain amplitude (a,p), %

Characteristic Examples. The influence of


macro and micro residual stresses on the cyclic
deformation behavior can be studied very well
after mechanical surface treatments, for example

by shot peening or deep rolling. Characteristic


cyclic deformation curves for stress-controlled
push-pull loading of different heat treated
smooth specimens of the steel AISI 4140 (German grade 42 CrMo 4) are compared in Fig. 13
in unpeened and in shot peened conditions with
compressive residual stresses at the surface (Ref
25, 26). In the normalized state (Fig. 13a), the
onset of cyclic deformation is different in both
conditions, since the shot peened specimens with
surface compressive residual stresses rrs

Plastic strain amplitude (a,p), %

POST (ra rm) ea,p E

10

100
103
Number of cycles, N

104

105

Cyclic deformation curves of different mechanically surface treated materials for stress-controlled push-pull loading in the low-cycle fatigue (LCF) regime of the plain carbon SAE 1045 steel. Stress amplitude
ra 350 MPa, Almen intensity I 0.120 mm A, rolling
pressure p 150 bar; surface residual stresses: shot peening rrs  500 MPa, deep rolling, approximately 600/
350 MPa. Source: Ref 28, 29

Fig. 15

400 MPa show cyclic softening from the first


cycle and higher plastic strain amplitudes during
the first cycles for stress amplitudes ra between
250 and 350 MPa. After a certain number of cycles, the opposite tendency can be detected and
the plastic strain amplitudes of the shot peened
conditions are smaller than those of the unpeened
material states. However, for the same ra values
the plastic strain amplitudes of both conditions
approach another at relatively high numbers of
cycles. Corresponding results for a quenchedand-tempered (730 C/2 h) AISI 4140 steel are
presented in Fig. 13(b). In the unpeened condition, the characteristic cyclic deformation behavior of quenched-and-tempered steels occurs with
a quasi-elastic incubation period, which is followed by cyclic softening until crack initiation.
After shot peening that generates surface compressive residual stresses rrs 410 MPa, the
onset of cyclic softening is shifted to smaller
numbers of cycles. Furthermore, it is interesting
to note that for identical stress amplitudes and
comparable numbers of cycles, the higher plastic
strain amplitudes are always measured for the
shot peened specimens. Figure 13(c) shows a
compilation of cyclic deformation curves for another quenched-and-tempered (570 C/2 h) AISI
4140 steel with a higher strength compared with
the steel condition in Fig. 13(b). In this case, the
shot peened condition that has surface compressive residual stresses rrs 530 MPa is characterized for all investigated ra values by small
measurable plastic strain amplitudes during the
first cycle that diminish or disappear first of all
with an increasing number of cycles. After a subsequent regime of quasi-elastic behavior, cyclic
softening, which yields to lower plastic strain
amplitudes and larger numbers of cycles to failure in comparison with the unpeened conditions,
is dominant.
The effect of deep rolling on the behavior of
the cyclic deformation curves is presented in Fig.
14(a) for the normalized plain carbon steel SAE
1045 (German grade Ck 45) (Ref 27). At comparable stress amplitudes essentially smaller
plastic strain amplitudes are measured in the
deep-rolled condition. In this materials state, the
cyclic softening effects are extremely restricted,
but continue during the whole lifetime until
crack initiation. As a consequence of the smaller
plastic strain amplitudes at identical stress amplitudes, higher numbers of cycles to failure are
observed for the deep-rolled condition compared
with the normalized ones. Figure 14(b) shows
the influence of deep rolling in the case of a nitrocarburized SAE 1045 steel. Due to the different microstructures in the near-surface areas of
the specimens, the same stress amplitudes yield
plastic strain amplitudes that are five times
smaller than those of the normalized state. The
deep-rolling process modifies the cyclic deformation curves of the nitrocarburized specimens
in a manner similar to the normalized specimens.
However, the reduction of the plastic strain amplitudes is much less pronounced.
Figure 15 shows plastic strain amplitudes as a
function of the number of cycles plotted for

Cyclic stress-strain curves of deep-rolled AISI


304 (rolling pressure 150 bar, surface residual
stresses r  350 MPa) using compact as well as hollow
specimens with different percentages of strain-hardened
layers compared with an untreated state and a shot peened
condition (rrs  450 MPa). Source: Ref 28, 29

damage is clearly correlated with plastic strain


amplitude, the benefit of mechanical surface
treatments becomes obvious. The influence of
mechanical surface treatments on cyclic plasticity can be summarized in cyclic stress-strain
curves, which correlate stress amplitudes and
plastic strain amplitudes for, for example, half
the number of cycles to failure (see the section
Cyclic Deformation Behavior). Figure 16, as
an example, shows data for an austenitic AISI
304 steel in shot peened as well as deep-rolled
conditions compared with untreated specimens
(Ref 28, 29). Effects of mechanical surface treatments are more distinct with a higher ratio between the area of affected surface layer and the
cross section of the specimen. In the case presented here, hollow thin-walled specimens were
prepared from compact specimens for analysis.
As shown in Fig. 16, cyclic yield strength considerably increases compared with compact
specimens, if only the fatigue behavior of the
surface layers of mechanically surface treated
components is investigated.
If residual stresses are present in the root of
notched specimen, marked changes always occur
in the initial parts of the notch root cyclic deformation curves (ea,p /log N curves) and in the
notch root cyclic mean strain curves (em /log N
curves) that depend on the sign and the magnitude of the residual stresses. Characteristic examples are presented in Fig. 17 for normalized

True plastic strain amplitude (a,p), %

True plastic strain amplitude (a,p), %

32 / Effect of Materials and Processing


push-pull loading with the stress amplitudes indicated for a normalized SAE 1045 steel (Ref
28, 29). Results for annealed as well as for shot
peened or deep-rolled conditions are shown. One
can clearly see that both mechanical surface
treatments considerably diminish the plastic
strain amplitudes. Due to the thicker affected
surface layer in the case mentioned first, the effect for deep-rolled states is more pronounced
than for shot peened states. Because fatigue

Stress amplitude (a), MPa

700
600
500

Deep rolled (1 mm wall thickness)


Deep rolled (0.5 mm wall thickness)
Deep rolled (0.3 mm wall thickness)

400
300
200

Untreated
Deep rolled
Shot peened

100
0
0

0.1
0.2
0.3
0.4
Plastic strain amplitude (a,p), %

0.5

Fig. 16

rs
l

0.4

n,a in MPa
rs = 15 MPa
rs = 490 MPa

0.3
220
0.2

240
200

220
175

0.1

150

0
1

10

102

170 150
190

103

104

105

106

0.06
n,a = 200 MPa
0.04
rs = 710 MPa
460

0.02

320
0
1

10

102

103

104

105

106

Number of cycles, N

(a)

Number of cycles, N
0.06

0.30

n,a in MPa

220

190

170

150

0.15

0.00

Plastic mean strain (m), %

Plastic mean strain (m), %

(a)

n,a = 200 MPa

0.04

0.02
460

rs = 710 MPa
0.2
1

0.15
1
(b)

10

102

103

104

105

106

(b)

10

102

103

104

105

106

Number of cycles, N

Number of cycles, N

True plastic strain amplitude (a) and plastic


mean strain versus number of cycles (b) for
stress-controlled push-pull loading of normalized notched
specimens (notch factor Kt 3.0) of the SAE 1045 steel in
unpeened and shot peened conditions. Source: Ref 30, 31

Fig. 17

320

True plastic strain amplitude (a) and plastic


mean strain versus number of cycles (b) for
stress-controlled push-pull loading of quenched-and-tempered (400 C/2 h) notched specimens (notch factor Kt
3.0) of the SAE 1045 steel in cut-milled and shot peened
conditions with different surface residual stresses. Source:
Ref 30, 31

Fig. 18

specimens and in Fig. 18 for quenched-and-tempered specimens of SAE 1045 with a notch factor Kt 3.0 that were investigated in push-pull
tests (Ref 30, 31). Figure 17(a) compares the cyclic deformation curves of material states with
very low residual stresses (produced by milling
and subsequent annealing) with rather high residual stresses (produced by shot peening). In
both conditions plastic deformation occurs in the
first cycle at nominal stress amplitudes rn,a
above 150 MPa. The annealed state yields continuous cyclic softening for all investigated rn,a
values. Contrarily, the initial plastic strain amplitudes of the shot peened condition that are
much less dependent on the stress amplitude
compared to the milled and annealed state are
reduced during the first cycles. This is a result
of residual stress relaxation that is the more pronounced the higher the nominal stress amplitude
is (see the following section Evaluation of Experimental Results). The consequences are
smaller effective stress amplitudes and smaller
ea,p values in the following cycles. Hence, the
observed cyclic work hardening is fictitious.
However, at higher numbers of cycles, cyclic
softening occurs with lower ea,p values than in
the unpeened condition.
On the other hand, as can be seen from Fig.
17(b), the shot peening notch root residual
stresses have an influence on the initial parts of
the mean strain curves that were determined for
different nominal stress amplitudes (Ref 30, 31).
As a consequence of the cyclic deformation induced relaxation of the compressive residual
stresses in the notch, the initial parts of the em /
log N curves show negative plastic mean strains,
where the magnitudes of em increase with increasing nominal stress amplitude. The minima
of em are shifted to lower numbers of cycles with
increasing rn,a values. However, from the minima of em at approximately 5% of the number of
cycles to failure, the em /log N curves show increasing mean strains for all nominal stress amplitudes. Simultaneously, a decrease of the tensile compliance in the hysteresis loops that
intensifies with an increasing number of cycles
is observed. By that, the compressive compliance is approximately constant. This finding
proves that this increase of the mean strains with
the number of cycles is caused by an increase of
the number of microcracks and/or the growth of
microcracks. These sections are indicated by
dashed lines in the em /log N curves of Fig. 17(b).
The true plastic strain amplitudes in the cyclic
deformation curves of the quenched-and-tempered (400 C/2 h) specimens in Fig. 18(a) are
influenced by the magnitude of the notch root
residual stresses, but not by their sign (Ref 30,
31). All investigated conditions in Fig. 18(a)
show for the initial cycles fictitious cyclic-workhardening effects due to the relaxation of residual stresses rrs of the surface layers (rrs 460
MPa and 320 MPa produced by upcut milling
as well as 710 MPa by shot peening). Up to
N 20 cycles, increasing magnitudes of the
compressive residual stresses cause higher negative mean strains. However, positive milling re-

Residual Stresses and Fatigue Behavior / 33

0
n,a = 220 MPa
200
n,a = 150 MPa

400

Plastic strain amplitude (a,p), %

0.20
250

a in MPa

0.16
125
0.12
150
0.08
350
0.04

200

100

0
1

102

10

(a)
3

10

102

103

104

105

106

Number of cycles, N
(a)

105

106

125

250
9
350
15

a in MPa

102

10

400
450
103

104

105

106

Number of cycles, N

Plastic strain amplitude (a) and plastic mean


strain versus number of cycles (b) for axial
stress-controlled cyclic loading with mean stress rm 300
MPa for specimens of the normalized SAE 1045 steel.
Source: Ref 32

Fig. 20

n,a = 200 MPa

rs

300

100

300
600
900
0

(b)

10

102

103

104

105

106

Number of cycles, N

Residual stress relaxation during stress-controlled push-pull loading of notched specimens


(notch factor Kt 3.0) for (a) normalized SAE 1045 in shot
peened conditions and (b) quenched-and-tempered (400
C/2 h) SAE 1045 in cut-milled and shot peened conditions.
Source: Ref 30, 31

Fig. 19

Ratio of cycles N/Nf, %

A Untreated
B Deep rolled

80
60
40
20
0

C Shot peened
Macrocracks
(>750 m)
Short cracks
(200-750 m)
Microcracks
(>200 m)
No damage
marks

Influence of mechanical surface treatments on


the damage evolution of the push-pull loaded
austenitic steel AISI 304. ra 320 MPa; R 1; A, Nf
3859; B, Nf 4445; C, Nf 20,265. Source: Ref 28

Fig. 21

plastic strain of notched specimens (see Fig.


18b).
A study of the residual stress effects on the
cyclic deformation behavior seems to be possible by their simulation by applied mean stresses,
which are homogeneously distributed over the
cross section of the specimen. An example is
presented in Fig. 20 for smooth, normalized SAE
1045 steel specimens that were stress-controlled
loaded in push-pull tests with a constant mean
stress of rm 300 MPa (Ref 32). The cyclic
deformation curves at stress amplitudes ra
125 and 150 MPa in Fig. 20(a) show first of all
a quasi-elastic incubation interval, followed by
considerable cyclic softening within some cycles. Subsequently, within some further cycles,
cyclic work hardening occurs to such an extent
that extremely low plastic strain amplitudes result. For ra 200 MPa, the compressive stress
peak | rm ra |, which is higher than the yield
strength, induces plastic deformation in the first
cycle. This procedure is followed by a rapid cyclic work hardening in such a manner that at cycles above 100 extremely low ea,p values are observed. As a consequence of the constant
compressive mean stress, the specimens shorten
for all investigated stress amplitudes, as proved
by Fig. 20(b). For ra 200 MPa, the mean
strains increase considerably and continuously.
However, for ra  200 MPa the ea,p values saturate above 20 cycles.

200

3

(b)
1

104

150

100

21

600

103

Number of cycles, N

Residual stress (l ), MPa

stricted mean free path of the mobile dislocations


in the work-hardened surface layers.
In hardened as well as in quenched-and-tempered at low temperature conditions, the changes
in the cyclic deformation behavior result not
only from surface hardening or softening effects,
but also from the more stable residual stresses as
can be seen in Fig. 19(b) for notch root residual
stresses of the quenched and tempered (400 C/
2 h) SAE 1045 steel under the loading conditions of Fig. 18(a) (Ref 30, 31). The consequences are a considerable influence of the
residual stresses on the cyclic behavior, especially with regard to the development of mean

Plastic mean strain (m), %

rs

Residual stress (l ), MPa

sidual stresses yield to positive mean strains. The


increases in the plastic mean strains that appear
for the numbers of cycles above the minimum of
em are caused also by microcracks as described
previously.
Evaluation of Experimental Results. In relatively soft material states, as for example in normalized as well as in quenched-and-tempered at
high-temperature conditions, the consequences
of mechanical surface treatments by shot peening or deep rolling on the cyclic deformation behavior are mainly caused by near-surface micro
residual stresses, that is, work hardening of the
surface layers, because the macro residual
stresses are relaxed very soon by cyclic plastic
deformation (see the article Stability of Residual Stresses in this Handbook). This relaxation
of macro residual stresses is shown in Fig. 19(a)
for shot peening residual stresses in the notch
root of a normalized SAE 1045 steel (Ref 30,
31) under the loading conditions of Fig. 17(a).
The dislocation structures in the ferrite after the
mechanical surface treatment are not stable and
change during cyclic loading in energetically
more favorable arrangements. The formation of
typical fatigue-induced dislocation structures is
combined with cyclic softening effects as presented in Fig. 13(a), 14(a), and 17(a). The small
plastic strain amplitudes of the shot peened or
deep-rolled conditions and the resulting increase in fatigue life are caused by the re-

Influence of Residual Stresses on


the Crack Initiation
Characteristic Examples. Crack initiation
occurs as a consequence of microstructural
changes in metallic materials during cyclic loading. Different mechanisms are responsible for
their formation (e.g., Ref 33). If it is accepted
that for given materials states at comparable load
amplitudes increasing amounts of plastic strain
amplitudes lead to decreasing numbers of cycles
to crack initiation Ni, it follows that residual
stresses may extend, shorten, or leave unchanged
the number of cycles to crack initiation. However, experimental investigations concerning the
influence of macro and micro residual stresses
on crack initiation are scarce. This is due to the
difficulties connected with the observation of the
formation and the propagation of small cracks.
A recently published investigation gives a report on the influence of mechanical surface treatments on crack initiation and crack propagation
in push-pull loading of steels (Ref 28). In untreated materials, crack initiation normally takes
place at positions of high localized slip, for example, at extrusions and intrusions that are connected with persistent slip bands. However, as
shown in Fig. 21 in shot peened and deep-rolled
conditions of the austenitic steel AISI 304, crack
formation occurs later than in the untreated state
due to the consequence of numerous obstacles
for slip (dislocations, grain and twin boundaries)
in the work-hardened surface layer that impede
localized slip. In these surface-work-hardened

34 / Effect of Materials and Processing


conditions, no persistent slip band is observed at
all. Furthermore, crack propagation is slower
than in the untreated state due to the effect of
microstructure and compressive residual
stresses. Similar results are found in Ref 28 for
the normalized steel SAE 1045, in Ref 34 for the
plain carbon steel SAE 1080, and in Ref 35 and
36 for a high-strength spring steel AISI 6150
(German grade 50 CrV 4).
However, as reported in Ref 1, the crack initiation time of shot peened specimens is sometimes shorter than that of unpeened specimens
despite increased lifetimes. One example is
given in Fig. 22 for the quenched-and-tempered
steel SAE 1045 in the case of bending fatigue
tests in seawater (Ref 37). With the exception of
high stress amplitudes, the cracks are formed
earlier in shot peened specimens than in ground
ones. This finding is attributed to an enhanced
crack initiation at micronotches resulting from
shot peening, which is obviously supported by
corrosion pittings in the case of seawater environment. However, the number of cycles to failure of the shot peened conditions are higher than
that of ground conditions.
Residual stresses may have a remarkable influence on the location of crack initiation. In the
case of bending fatigue, for example, cracks may
start below the surface, if the magnitude of the

Stress amplitude (a), MPa

800

compressive residual stresses near the surface is


high enough. Characteristic results are presented
in Fig. 23 for shot peened bending specimens in
a hardened state of the steel SAE 1045 with the
bending fatigue strength Rf 960 MPa after
shot peening with the shot size d 0.3 mm and
Rf 1050 MPa with d 0.6 mm, respectively
(Ref 38, 39). For stress amplitudes ra  Rf, crack
initiation was observed directly at the surface.
However for ra  Rf, subsurface cracks occurred
as shown in Fig. 24. If the centers of the rosettes
on the scanning electron micrographs (SEMs)
are considered to be the crack initiation points,
it is obvious that the crack initiation depth increases with decreasing stress amplitude (Ref
39). At stress amplitudes that result in a fatigue
life Nf above 107 cycles, crack initiation is located approximately 0.3 mm below the surface.
These findings stimulated the development of a
concept of the local fatigue strength (see the paragraphs on experimental results in the section
Influence of Residual Stresses on S-N curves
in this article).
Evaluation of Experimental Results. The definition of the number of cycles to crack initiation
Ni and particularly the establishment of the affiliated crack length is of decisive meaning for
the evaluation of the influence of different parameters on Ni from experimental results from

(a)
Fracture
Crack initiation

600

200 m
(b)

400

200 m

200
Shot peened

(c)

Ground
0
104

105
106
Number of cycles, N

107

Stress amplitude versus number of cycles to


crack initiation and to failure of a quenchedand-tempered SAE 1045 steel under bending fatigue loading in seawater. Source: Ref 37

200 m
(d)
200 m

200 m

Distance from surface, mm

0.4

200 m

0.2
Site of
macrocrack
initiation

However, decreasing Ni values are observed,


if in the surface area:
The roughness is increased, for example, by

shot peening

Tool marks or material overlaps from shot

peening or deep rolling exist

Tensile macro residual stresses occur that fa-

vor interface cracks between matrix and hard


second phases.
In this context, an investigation of the fatigue
crack initiation in double-edge notch specimens
of SAE 1080 steel with a fine-grained, spheroidized microstructure shows considerable effects
of the macro residual stresses (Ref 34). In a
press-fitted condition, large compressive residual macro stresses occur that remained relatively
stable throughout the fatigue life and thus greatly
increased the numbers of cycles to crack initiation Ni. With reference to the Basquin-relationship, Ni is estimated from the following relationship:

r (r

2ra
a

1/b;

r )
rs

(Eq 13)

where Ni is defined as the number of cycles at a


crack length of a 0.1 mm, rrs is the macro
residual stress, ra is the applied stress amplitude,
and ri 2315 MPa and bi 0.197 are material constants.

200 m

Shot size 0.3 mm


Shot size 0.6 mm

1000
1200
Stress amplitude (a), MPa

Influence of Residual Stresses on


the Crack Propagation

(g)

1400

Crack initiation sites versus stress amplitude of


a hardened SAE 1045 steel in different shot
peened conditions. Source: Ref 38, 39

Fig. 23

stresses

Is smoothed by polishing or deep rolling

(f)

0.3

800

face treatment

Bears additional compressive macro residual

(e)

Becomes work hardened by mechanical sur-

Ni

Fig. 22

0.1

literature. If the corresponding crack length


amounts to only some grain diameters, then Ni
is determined by shear-stress-controlled processes in stage I of crack propagation. However,
if crack lengths of several hundred microns appear (stage II of crack propagation), Ni will be
determined by normal-stress-controlled processes. In the last case, stable macro residual
stresses may have a considerable influence on
the Ni values.
The experimental results achieved until now
for crack formation directly at the surface with
an appropriate consideration of the Ni definition
show increasing numbers of cycles to crack initiation, if the surface area:

Scanning electron micrographs of fractured


specimens of hardened and shot peened SAE
1045 at different stress amplitudes. (a) ra 1300 MPa. (b)
ra 1250 MPa. (c) ra 1200 MPa. (d) ra 1100 MPa.
(e) ra 1000 MPa. (f) ra 950 MPa. (g) ra 900 MPa.
Source: Ref 39

Fig. 24

Characteristic Examples. During cyclic


loading, the lifetime of components is determined significantly by the stage of crack propagation, especially in the case of existing macro
residual stresses. Crack propagation as a consequence of fatigue loading without mean stresses
can be described, with the exception of the crack
initiation and the near-fracture stage, by the Paris
equation (see Eq 1):

Residual Stresses and Fatigue Behavior / 35


da
C(DK)m
dN

(Eq 14)

where DK is the stress intensity range and C and


m are constants. Mean stresses are taken into account by the Foreman equation:
da
C(DK)m

dN
(1 R)Kc DK

200

rs

200

Crack propagation rate (da/dN), mm/cycle

(Eq 15)

where R is the stress ratio and Kc is the stress


intensity factor for plane stress. Both relationships are valid for crack propagation in materials
states without macro residual stresses. However,

Residual stress (y ), MPa

these descriptions imply that a propagating crack


itself is surrounded by a typical residual stress
field as shown in Fig. 25(a) for a high-strength
structural steel of the European grade S690QL1
(Ref 40, 41). A crack was produced by cyclic
loading up to a stress intensity range of DKI
47.4 MPa m. The distribution of the macro re-

104

As received

105

106

Autofrettaged

0.2

0.4

0.6

0.8

1.0

Ratio a/W

Crack propagation rate da/dN versus ratio a/


W in an autofrettaged and an untreated tube of
the SAE 4337 steel. Source: Ref 43

Fig. 27

400
12

8
4
0
4
8
Distance from crack tip, mm

(a)

12
10
150 bar
Crack propagation rate
(da/dN), mm/cycle

rs

Residual stress (y ), MPa

200

200

8
4
0
4
8
Distance from crack tip, mm

Residual stress component rrsy versus distance


from crack tip of the steel S690QL1 after a
mode I base load of DK 47.4 MPa m (a) and an overload of DK 94.8 MPa m (b). Source: Ref 40, 41

102
=2

103
104

=3

105

8

4

75 bar

225 bar

4
2
300 bar
0

1
2
Crack length (a), mm

Fatigue crack propagation rates da/dN in annealed and with different rolling pressures
deep-rolled steel AISI 304. Stress amplitude ra 320 MPa,
surface residual stresses rrs  200 MPa (75 bar), 350
MPa (150 bar), 400 MPa (225 bar), and 300 MPa (300
bar). Source: Ref 29

Fig. 28

Fig. 25

12

103

104

Stress relieved

As welded

105

106

10

20

30 40

60 80

Stress intensity range (K), MPa m

Crack length (aol), mm

Crack propagation rate da/dN versus crack


length a0l for overloads with k 2 and 3.
Source: Ref 40, 41

Fig. 26

Untreated

12

Crack propagation rate (da/dN), mm/cycle

12
(b)

106

400

Crack propagation rate (da/dN), mm/cycle

103

Crack propagation rate da/dN versus stress intensity range DK in the heat-affected zone of
EN S355 steel in an as-welded condition and after a stressrelief heat treatment. Source: Ref 44

Fig. 29

sidual stress component rrs


y , which acts perpendicular to the crack flanks, shows maximum
compressive residual stresses of approximately
350 MPa at the crack tip. The alteration of this
distribution after application of 20 overload cycles with an overload ratio k 2 (DKI 94.8
MPa m) is given in Fig. 25(b). In front of the
crack tip, a larger maximum value and a larger
area with compressive residual stresses compared with Fig. 25(a) are developed. The influence of different overload cycles on the crack
propagation rate is shown in Fig. 26 for k 2
and 3. For both overload ratios, a delayed retardation of crack propagation occurs that is more
pronounced for k 3 than for k 2 due to the
effect of overload-induced compressive residual
stresses. Thus, by sufficiently high overloads,
crack arrest can be produced (Ref 42).
If a crack propagates into a macro residual
stress field, the crack propagation behavior can
be considerably influenced by magnitude and
distribution of the residual stresses. This is
shown in Fig. 27, in which crack propagation
rates within an autofrettaged tube of SAE 4337
steel are compared with those in the untreated
material (Ref 43). As a consequence of the autofrettaging process, a triaxial residual stress
state is created with very high compressive residual stresses near the inner surface of the tube.
In agreement with Eq 14, loading with a constant
stress intensity range results in a constant crack
propagation rate in residual-stress-free tubes.
However, if the crack propagates through the
compressive residual stress field at the inner surface of the autofrettaged tube, considerably
smaller crack propagation rates occur.
Crack velocity is influenced by work hardening as well as by residual stresses in mechanically treated surface layers. This is demonstrated
in Fig. 28 for differently surface-rolled AISI 304
stainless steel. The crack propagation rate da/
dN, which was determined by analyzing striations on cracked surfaces, increases with increasing rolling pressure and is considerably
diminished compared with untreated specimens
(Ref 29).
Finally, the influence of welding residual
stresses at the steel European grade EN S355 is
compared with the behavior of the same steel in
a stress-relieved condition in Fig. 29 in a da/dN
versus DK diagram (Ref 44). The crack propagation rate in the heat-affected zone is considerably lower in the as-welded state than in the
stress-relieved state due to the distribution of
welding compressive residual stresses.
Evaluation of Experimental Results. For
practical purposes, it is very important to know
that crack propagation through residual stress
fields can be modeled quantitatively by introducing an effective stress intensity range DKeff.
This can be seen in Fig. 30, which shows crack
propagation rates da/dN in welded and unwelded specimens of SAE 1019 steel (Ref 45).
Exactly at the welding seam, tensile residual
stresses of about 340 MPa exist that disappear
in a distance of 10 mm. Figure 30 presents crack
propagation rates in the unwelded base material

103
0.1 0 0.5 1.0
104

0.2

R = 0.4

2.0

105

106

10

20

30

40 50 60

Stress intensity range (K), MPa m

Crack propagation rate (da/dN), mm/cycle

(a)

103

R=0
0.5

104

105

106

1.0

2.0

6

10

20

30

40 50 60

(Eq 16)

If the influence of the residual stress contribution


on Kop is known, DKeff can be estimated and used
to predict the fatigue crack growth. Thus, the
double logarithmical plot of the crack propagation rate da/dN versus the effective stress intensity range DKeff in Fig. 30(c) shows that all measured data points form a narrow scatter band. In
welded specimens, the (da/dN)/DK relations are
completely changed (e.g., Fig. 30a and b). Obviously, for small DK values the crack propagation behavior is entirely controlled by the
welding residual stress state, which leads to
identical crack propagation rates irrespective of
the R ratio. However, outside the volume areas

Nominal stress amplitude (n,a), MPa

Crack propagation rate (da/dN), mm/cycle

DKeff Kmax Kop

Stress intensity range (K), MPa m

(b)

103

104

105

106

10

20

30

300
Upcut milling
rs = 234 MPa
200

Downcut milling
rs = 242 MPa

kt = 2.5
=5
0
104

105
106
107
Number of cycles to failure, Nf

Alternating bending S-N curves of notched


specimens of normalized plain carbon SAE
1045 steel after annealing, downcut milling and upcut milling. Source: Ref 4749

Fig. 31

103

104

105

106

10

20

30

40 50 60

Effective stress intensity range (Keff), MPa m

(d)
Crack propagation rate da/dN versus stress intensity range DK in the base material (a) and (c)
and versus effective stress intensity range DKeff (b) and (d)
in the welded zone of the SAE 1019 steel. Source: Ref 45

Fig. 30

300

200

=1

= 0.4
= 0.1

kt = 1.7, = 2

kt = 2.5, = 5
100

kt = 4.4, = 15

kt = 2.5, = 2

Smooth
Notched
0
750 500 250

250

500

750

rs

Surface residual stress (l ), MPa


Alternating bending fatigue strength of milled
smooth and notched specimens of normalized
plain carbon SAE 1045 steel versus surface residual stress.
Source: Ref 4752

Fig. 32

with high tensile residual stresses, individual


curves for each R value used are observed. For
R (pulsating compression), first increasing and then decreasing crack propagation rates
are observed indicating the influence of the tensile residual stress field. However, also in this
case, the crack propagation behavior can be described by a single, relatively narrow scatter
band if the influence of the welding residual
stresses on the DK values appearing at the crack
tip is taken into account (see Fig. 30d). In this
way, the crack propagation behavior in macro
residual stress fields can be quantitatively described if crack propagation data of macro residual-stress-free materials are available and the residual stress distribution is known.

Influence of Residual Stresses on


S-N Curves
Characteristic Examples
Low-Strength Steel. Figure 31 shows S-N
curves for alternating bending of normalized
SAE 1045 steel (German grade Ck 45) (Ref 47
49). The notched specimens had a stress-concentration factor kt 2.5. The stress gradient at
the notch root dr/dz related to the maximum
stress r (normalized stress gradient g):
g

Annealed
rs = 0 MPa

100

40 50 60

Effective stress intensity range (Keff), MPa m

(c)
Crack propagation rate (da/dN), mm/cycle

as a function of DK for different stress ratios R.


As expected, with increasing R values, higher
crack propagation rates are observed. It is important to note that a mode I crack can only grow
during that portion of loading cycle where the
crack is open. This portion is influenced by the
R value itself and the near-crack-tip residual
stress distribution. The effective stress intensity
range can be determined quantitatively as the
difference between the maximum stress intensity
and the stress intensity where the crack opens:

Bending fatigue strength (Rf), MPa

Crack propagation rate (da/dN), mm/cycle

36 / Effect of Materials and Processing

1 dr
r dz

z0

(Eq 17)

was 5 mm1. All data are nominal stress amplitudes and are valid for a failure probability of
50%. The bending fatigue strength was evaluated at an ultimate number of cycles Nu 107.
By downcut milling and upcut milling, surface
residual stresses of 242 and 234 MPa, respectively, were generated. The corresponding S-N
curves are almost identical. A third batch of
specimens was annealed 2 h at 700 C after
downcut milling. The annealing results in a reduction of bending fatigue life and bending fatigue strength (Ref 4749).
The alternating bending fatigue strengths of
milled smooth and notched specimens with different geometries are plotted in Fig. 32 as a function of the surface residual stresses (Ref 4652).
Again, all data are given for a failure probability
of 50%, and the bending fatigue strengths are
nominal stress amplitudes at Nu 107. With
increasing stress-concentration factor and decreasing stress gradient, the bending fatigue
strength decreases. The influence of the stressconcentration factor is clearly visible from a
comparison of the specimens with the same
value g 2 mm1, but different values kt
1.7 and 2.5, respectively. On the other hand, the
increase of g from 2 to 5 mm1 at specimens
with kt 2.5 results in a significant increase of
bending fatigue strength. It is also interesting to
note that specimens with kt 4.4, g 15 have
a somewhat higher strength than specimens with

Residual Stresses and Fatigue Behavior / 37

300
280
260

Deep rolled

240
220
200
180
104

105
106
107
Number of cycles to failure, Nf

320
280

rs

260
240
220
200
180
104

1000

Deep rolled

300

Residual stress (l ), MPa

Nominal stress amplitude (n,a), MPa

(a)

2 20 m, 30 m/s
500
2 6 m, 15 m/s
0
2 3 m, 15 m/s

500

105
106
107
Number of cycles to failure, Nf

0.1
0.2
Distance from surface, mm

(b)

Nominal stress amplitude (n,a), MPa

600
500
400
rs = 221 MPa

300
200
100
0
104

19

kt = 1.7
=2

602

105
106
107
Number of cycles to failure, Nf

Alternating bending S-N curves of notched


specimens made from quenched-and-tempered (60 C/2 h) plain carbon SAE 1045 steel after different grinding processes. Source: Ref 4749

Fig. 34

Depth distribution of the residual stress in


notched specimens made from quenched-andtempered (600 C/2 h) plain carbon SAE 1045 steel by different grinding processes with the indicated two steps of
final feed and cutting speed. Source: Ref 4749

Fig. 35

Bending fatigue strength (Rf), MPa

S-N curves of specimens made from normalized plain carbon SAE 1015 steel in the asheat-treated state and after an additional deep rolling for
(a) push-pull loading and (b) rotating bending. After Ref 53

Fig. 33

with rather small and with compressive surface


residual stresses yield identical fatigue behavior.
It should be noted that the surface layer bearing
compressive residual stresses is rather small.
Contrarily, the generation of tensile residual
stresses causes a significant decrease of bending
fatigue strength and a rather small decrease of
finite fatigue life, which obviously vanishes at
high stress amplitudes.
The alternating bending fatigue strength of
ground smooth and notched specimen with different geometries made of the same material
state is plotted in Fig. 36 as a function of the
surface residual stresses (Ref 4752). Similar to
Fig. 32, the influences of the stress-concentration
factor and the normalized stress gradient are
clearly discernible. However, contrary to the
normalized steel there is distinct reduction of the
bending fatigue strength with increasing tensile
residual stress, this being more pronounced in
the comparison of smooth specimens with
notched specimens. In the range of compressive
residual stresses covered, the influence of residual stress on bending fatigue strength is rather
small.
The S-N curves of smooth specimens in the
ground state and after an additional shot peening
are compared in Fig. 37(a) (Ref 4749). There
is a distinct increase of the bending fatigue
strength by shot peening, but a rather small in-

400
= 0.4

Smooth
300

= 0.1

kt = 1.7
Notched

=2

200

100

0
500

=5

kt = 2.5

250

=2
0

250

500

750

rs
Surface residual stress (l ), MPa

Alternating bending fatigue strength of ground


smooth and notched specimens made from
quenched-and-tempered plain carbon SAE 1045 steel versus surface residual stress. Source: Ref 4752

Fig. 36

Nominal stress amplitude (n,a), MPa

320

machining procedures. After correction of the


data points given in Fig. 32 to the same hardness,
it turns out that the bending fatigue strength is
hardly changed or slightly diminished at most,
if the residual stresses change from compressive
to tensile (Ref 49).
S-N curves for push-pull loading of smooth
specimens made from normalized SAE 1015
steel (German grade Ck 15) in the as-heat-treated
state and after an additional deep rolling are
shown in Fig. 33(a) (Ref 53). Again, the fatigue
behavior in the range of finite fatigue life and the
fatigue strength do not differ much. Figure 33(b)
shows a plot of S-N curves for rotating bending
of the same material states. Now, by deep rolling
finite fatigue life is increased by one order of
magnitude or more, and the bending fatigue
strength is increased significantly.
Medium-Strength Steel. Figure 34 shows SN curves that were determined in alternating
bending on notched specimens of quenched-andtempered (600 C/2 h) SAE 1045 steel (Ref 47
49). All data are valid for a failure probability
of 50% and Nu 107. Three batches of specimens were manufactured by grinding. The grinding parameters (final feed, cutting speed) and the
resulting depth distributions of residual stresses
are given in Fig. 35 (Ref 4749). Specimens

600
500

Shot peened

400
300
Ground
200
100
0
104

105
106
107
Number of cycles to failure, Nf

(a)
Nominal stress amplitude (n,a), MPa

Nominal stress amplitude (n,a), MPa

kt 2.5, g 2 mm1. However, with regard


to the residual stress state, there is no significant
influence on the bending fatigue strength, even
though the range of residual stresses covered
comes to more than 1000 MPa regarding specimens with kt 4.4, g 15 mm1. Careful
inspection of the hardness of the specimens
tested shows that a positive slope of the lines in
Fig. 32 is not related to the changing (macro)
residual stress state, but to different hardness of
the specimen and, hence, differences in the micro residual stress state produced by different

600
500

Shot peened

400
Milled
300
200

Ground

100
0
104

(b)

105
106
107
Number of cycles to failure, Nf

Alternating bending S-N curves of specimens


made from quenched-and-tempered (600 C/2
h) plain carbon SAE 1045 steel. (a) Smooth specimens after
grinding and after additional shot peening. (b) Notched
specimens after milling, grinding, and grinding with additional shot peening. Source: Ref 4749

Fig. 37

38 / Effect of Materials and Processing

400
Smooth
= 0.4

300

grinding. Again, in the range of finite fatigue life,


the influence of the different manufacturing processes is almost negligible. The relative increase
of the bending fatigue strength by shot peening
is more pronounced compared with smooth
specimens. It is interesting to note that milled
specimens have a higher bending fatigue
strength than ground ones, even though they
have lower compressive residual stresses (159

200
Notched
kt = 1.7
=2

100
0
750 500 250

250

500

750

rs

Maximum residual stress (l ), MPa


Alternating bending fatigue strength of
quenched-and-tempered (600 C/2 h) plain
carbon SAE 1045 steel versus surface residual stress evaluated from Fig. 36 and 37

Fig. 38

Bending fatigue strength (Rf), MPa

Bending fatigue strength (Rf), MPa

fluence on finite fatigue life. Figure 37(b) compares S-N curves of notched specimens which
were milled, ground, and shot peened after

800

600

400

Notched

Smooth

200

1000

2000
3000
4000
Deep-rolling force, N

5000

Bending fatigue strength smooth and notched


specimens made from quenched-and-tempered SAE 5135 steel versus deep-rolling force. After Ref
57

Alternating bending S-N curves of smooth


specimens made from blank-hardened AISI
5115 steel in the as-blank-hardened and with different conditions shot peened states including one with a subsequently electropolished surface. 1, as-blank-hardened; 2,
shot velocity v 23 m/s, coverage c 100%, mean diameter of the shot d 0.6 mm; 3, v 53 m/s, c 100%,
d 0.3 mm; 4, v 53 m/s, c 100%, d 0.6 mm; 5,
v 81 m/s, c 600%, d 0.6 mm; 6, v 53 m/s, c
100%, d 0.6 mm, 100 lm surface layer electrolytically removed. After Ref 5456

Fig. 39

Nominal stress amplitude (n,a), MPa

Fig. 41

600
rs = 400 MPa

500
400

300
200

923

100
0
104

kt = 1.7
=2
105
106
107
Number of cycles to failure, Nf

Alternating bending S-N curves of notched


specimens made from quenched plain carbon
SAE 1045 steel after different grinding processes. Source:
Ref 4749

Fig. 42

1000

200

Residual stress (l ), MPa

0
200

rs

rs

Residual stress (l ), MPa

400

3
4
5

400
600
800

1000
0

0.1
0.2
0.3
0.4
Distance from surface, mm

0.5

Depth distribution of the residual stress in specimens made from blank-hardened AISI 5115
steel in the as-blank-hardened (1) and in different conditions of the shot peened state (3), (4), and (5) corresponding
to Fig. 39. After Ref 5456

Fig. 40

2 15 m, 30 m/s
500
2 9 m, 15 m/s
0
2 3 m, 15 m/s
500

0.1
0.2
Distance from surface, mm

Depth distribution of the residual stresses in


notched specimens made from quenched plain
carbon SAE 1045 steel and ground with the two steps of
final feed and cutting speed indicated. Source: Ref 4749

Fig. 43

MPa) at the surface than the latter ones (221


MPa).
Figure 38 shows plots of the alternating bending fatigue strengths evaluated from Fig. 37 as a
function of the surface residual stresses. The arrows mark the shift in bending fatigue strengths
and surface residual stresses produced by shot
peening. Additionally, data points of ground
specimens with negligible or tensile residual
stresses shown in Fig. 36 are included. In the
case of notched specimens, all data points lie on
a common line with the slope 0.154 except for
ground specimens with compressive residual
stresses at the surface. Regarding smooth specimens, the influence of tensile residual stresses
on the bending fatigue strength is much more
pronounced than the influence of compressive
residual stresses.
Figure 39 shows S-N curves for alternating
bending of smooth specimens (g 1 mm1) of
blank-hardened AISI 5115 steel (German grade
16 MnCr 5) determined in the unpeened and
various shot peened states including one with
electrolytically removed surface layer (Ref 54
56). The corresponding depth distributions of residual stresses are given in Fig. 40. From the
comparison of both figures, it becomes clear that
the surface residual stress is not a suitable parameter for the assessment of the influence of the
various treatments on the fatigue behavior.
The influence of the deep-rolling force on the
rotating bending fatigue strength of smooth and
notched specimens made from quenched-andtempered SAE 5135 steel (German grade 37 CrS
4) is shown in Fig. 41 (Ref 57). In both cases, a
maximum of the fatigue strength occurs at a certain force. However, the increase of the fatigue
strength of notched specimens (kt 2) by deep
rolling is much more pronounced than that of
smooth specimens. In the end, the optimal bending fatigue strength of notched specimens
which is a nominal stress amplitudeis higher
than the bending fatigue strength of smooth
specimens.
High-Strength Steels. S-N curves for alternating bending of notched specimens of
quenched SAE 1045 steel in differently ground
conditions are compared in Fig. 42 (Ref 4749).
The grinding parameters (final feed, cutting
speed) and the resulting depth distributions of
residual stresses are given in Fig. 43. Similar to
ground quenched-and-tempered specimens, the
fatigue behavior of specimens with very small or
compressive residual stresses (which have a very
small penetration depth) hardly differ. Tensile
residual stresses, however, not only cause a
strong reduction of bending fatigue strength, but
also of finite fatigue life.
In Fig. 44, the bending fatigue strength that
was evaluated from Fig. 42 and corresponding
data from tests on smooth specimens are plotted
as a function of the surface residual stresses. The
negative influence of tensile residual stresses on
the bending fatigue strength of smooth specimens is even more pronounced compared to
notched specimens. The influence of compressive residual stresses generated by grinding is

Residual Stresses and Fatigue Behavior / 39

800
Smooth
= 0.4

600
400

200
0
1000

Notched
kt = 1.7
=2

500

500

fatigue life, which comes up to one and a half


orders of magnitude at high stress amplitudes.
The S-N curves of notched specimens of the
same steel state after grinding and after additional shot peening with shot of different hardness are shown in Fig. 45(b). Compared to
smooth specimens, shot peening produces a
much stronger increase of the bending fatigue
strength. Again, there is also a remarkable in-

Bending fatigue strength (Rf), MPa

Bending fatigue strength (Rf), MPa

much smaller than the influence of tensile


stresses.
In Fig. 45(a), the S-N curves of smooth specimens in the ground state and after additional
shot peening are compared (Ref 4649). Similar
to quenched-and-tempered specimens (see Fig.
37), shot peening produces a significant increase
of the bending fatigue strength. Contrary to the
results of the medium-strength steel, however,
there is also a very pronounced increase of finite

1000

800
Smooth
= 0.4

600

400

Evaluation of Experimental Results

Ground
Shot peened
Milled

200

Notched
kt = 1.7
=2

0
1000

500

500

1000

rs

Surface residual stress (l ), MPa

1500

rs

Surface residual stress (l ), MPa

Fig. 44 Alternating bending fatigue strength of ground


smooth and notched specimens of quenched
plain carbon SAE 1045 steel versus surface residual stress.
Source: Ref 4749

Alternating bending fatigue strength of smooth


and notched specimens made from quenched
plain carbon SAE 1045 steel with different surface conditions versus surface residual stress

Fig. 46

Ratio Rf/Rh, MPa/m

900

Shot peened

800
700
Ground
600

10
4

1
500
400
104

200

400

600

800

Surface hardness, HV 5
105
106
107
Number of cycles to failure, Nf

(a)

Changes of alternating bending fatigue strength


DRf relative to changes of surface roughness
height DR h versus surface hardness HV 5 for different heat
treated SAE 1045 and AISI 5115 steels. After Ref 58, 59

Fig. 47

1000
900

0
5458 HRC

800
700
600
500
400
104

4650 HRC Shot peened


Milled
Ground
105
106
107
Number of cycles to failure, Nf

(b)
Alternating bending S-N curves of specimens
made from quenched plain carbon SAE 1045
steel. (a) Smooth specimens after grinding and after additional shot peening. (b) Notched specimens after grinding,
milling, and grinding with additional shot peening with
shot of the indicated hardness. Source: Ref 4649

Fig. 45

rs

Nominal stress amplitude (n,a), MPa

SAE 1045
AISI 5115

40

Surface residual stress (l ), MPa

Nominal stress amplitude (n,a), MPa

100
1000

crease of finite fatigue life. Additionally, the SN curve of milled specimen is included. Finite
fatigue life and bending fatigue strength of these
specimens are lower compared to shot-peened
ones, even though they contain very high surface
compressive residual stresses of 1200 MPa.
In Fig. 46, the bending fatigue strength data
already plotted in Fig. 44 are complemented by
data evaluated from Fig. 45. The arrows mark
the shift in bending fatigue strengths and surface
residual stresses produced by shot peening.
Similar to the discussion of Fig. 39 and 40, it
becomes obvious that the magnitude of surface
residual stress is not a suitable parameter for the
assessment of the influence of shot-peening-induced residual stresses on the fatigue strength.

n,a = 360 MPa

200

325

400

280
200

600

185
0

10

102 103 104 105 106 107


Number of cycles, N

Surface residual stress in notched (kt 1.7)


upcut milled specimens made from normalized plain carbon SAE 1045 steel during alternating bending fatigue at different nominal loading amplitudes versus
number of cycles. Source: Ref 49

Fig. 48

Low-Strength Steels. Regarding Fig. 31 to


33, different machining processes or process parameters were applied to the specimens investigated producing different surface topographies.
In Fig. 47, the ratio DRf /DRh, which gives the
change of the alternating bending fatigue
strength of differently treated SAE 1045 and
AISI 5115 steels per lm increase of roughness
height is plotted as a function of the hardness at
the surface (Ref 58, 59). As is well known from
many other investigations, the susceptibility of
the bending fatigue strength to roughness increases with increasing hardness. For the steel
states investigated, a value of 2 MPa/lm can
serve as a guide. Since the roughness height of
the specimens did not vary by more than 2 lm,
the results plotted in Fig. 31 and 32 are hardly
influenced by surface topography. It should be
kept in mind, however, that a large increase of
surface roughness will significantly decrease the
fatigue strength even in low-strength material
states.
In low-strength steels, cyclic plasticity appears if the cyclic loading approaches the fatigue
strength. Extensive cyclic plasticity occurs in the
range of finite fatigue life, being combined with
cyclic softening and/or hardening processes (see
the section Some Aspects of Fatigue of Steels
in this article). Cyclic plasticity results in macro
residual stress relaxation, which is more rapid
the higher the cyclic loading is (see the article
Stability of Residual Stresses in this Handbook). Figure 48 (Ref 49) shows the relaxation
of the surface residual stresses during alternating
bending fatigue in notched upcut milled specimen made from normalized SAE 1045 steel
starting at an initial value of 590 MPa. The
bending fatigue strength amounts to 190 MPa
(see Fig. 32). During loading with an amplitude
of 185 MPa just below the fatigue limit, the
change of the macro residual stresses is almost
negligible. During all other fatigue loadings, extensive residual stress relaxation occurs. Therefore, almost no influence of macro residual
stresses is found in Fig. 31 and 32 (Ref 4752).
The amount of cyclic plasticity, which occurs
during cyclic loading resulting in a given finite

40 / Effect of Materials and Processing


fatigue life or which corresponds to the fatigue
limit, decreases with increasing strength of the
steel. This results in decreasing residual stress
relaxation rates and finally in stable macro residual stresses. On the other hand, in a low-strength
steel, local work hardening may occur during
manufacturing processes such as milling, turning, or grinding. In other words, the micro residual stress state is changed, mainly by the increase
of the dislocation density. As discussed in the
article Stability of Residual Stresses in this
Handbook, the resistance of the micro residual
stress state against cyclic relaxation is much
higher compared to that of the macro residual
stress state. So one may assume that macro residual stresses that exist in a part of a component
that has undergone work hardening by manufacturing processes are more stable than expected
from the basic strength of the material. Looking
at Fig. 32, this relationship may be responsible
for the small reduction of the bending fatigue
strength by tensile residual stresses that occurs
in some of the specimen series investigated.
A much more important consequence of the
change of the micro residual stress state by local
work hardening is the direct alteration of the local fatigue behavior. Increasing strength reduces
the local cyclic plastic deformation at a given
external loading, thus increasing the number of

cycles to crack initiation and decreasing the


propagation rate of short cracks (Ref 60, 61).
Therefore, as is evident from Fig. 31, a residual
stress relief heat treatment that eliminates the
machining-induced work hardening reduces the
finite fatigue life and the fatigue limit of normalized SAE 1045 steel. At rather high loading
amplitudes, this influence vanishes because then
the cyclic deformation behavior is almost entirely governed by cyclic softening and hardening processes during the fatigue loading (see the
section Cyclic Deformation Behavior in this
article) irrespective of initial variations of the local work-hardening state.
A smaller influence of the micro residual
stress state appears in Fig. 33(a) and a somewhat
stronger one in Fig. 33(b). Here, it becomes clear
that even in a rather soft material state such as
normalized SAE 1015 steel, an improvement of
the fatigue behavior can be obtained from local
work hardening. In this case, the depth of the
surface layer of the smooth specimens influenced
by deep rolling is relatively large compared to
the milled specimens, which are the basis of Fig.
31 and 32. Nevertheless, push-pull loading (Fig.
33a) without stress gradient reveals a small influence of deep rolling, because of two major
effects. First, the crack initiation site may be
shifted below the work-hardened zone. Then, the
small increase of the fatigue strength mirrors the
benefit of omitting surface effects (such as spec-

600

Smooth
= 0.4
n,a = 200 MPa

rs

Surface residual stress (l ), MPa

Ratio Rf/HV 5

Rh = 3 m
1

Rh = 15 m
0
0

200

400

600

800

1000

Surface hardness, HV 5

400
Notched
kt = 1.7, = 2
n,a = 150 MPa

200

kt n,a = 255 MPa

0
0

Changes of alternating bending fatigue strength


DRf relative to changes of surface hardness
DHV 5 versus surface hardness HV 5 for different surface
roughnesses heights Rh. After Ref 58, 59

10

Fig. 49

600

=2

kt = 2.5

=5

1.5

kt = 1.7

400

200

200

0
0

200

400

106 107

10

102

103

104

105

106

Number of cycles, N
600

rs
Surface residual stress (l ), MPa

Fatigue notch factor kf of specimens made from


quenched-and-tempered (600 C/2 h) plain
carbon SAE 1045 steel versus surface residual stress.
Source: Ref 4749

Fig. 50

105

Notched
n,a = 350 MPa
kt = 1.7
=2
kt n,a = 595 MPa

=2
1.0
400

104

Smooth
= 0.4
n,a = 400 MPa

rs

2.0

103

(a)
Surface residual stress (l ), MPa

Fatigue notch factor, kf

2.5

102

Number of cycles, N

(b)
Surface residual stress in smooth and notched
specimens made from quenched-and-tempered (600 C/2 h) plain carbon SAE 1045 steel versus
number of cycles (a) at a lifetime of approximately 2 106
cycles and (b) at a lifetime of approximately 105 cycles.
Source: Ref 4749

Fig. 51

imen roughness, corrosion, oxidation) on the


crack initiation. Secondly, if crack initiation
occurs at the surface, the influence of the workhardened zone may also be rather small, becauseat a given loadinga stress redistribution may occur from the softer core region of the
specimen to the work-hardened zone. This effect
is combined with cyclic plasticity in the core region and, hence, with rather high loading amplitudes. The first effect predominates in the range
of the fatigue limit, where cyclic plasticity is
small, and the second one in the range of finite
fatigue life. On the other hand, during rotating
bending (Fig. 33b) with a distinct stress gradient,
crack initiation always occurs at the surface. If
the work-hardened zone is thick enough, there is
no stress redistribution from the core region
which is then purely elastically loadedto the
work-hardened zone. Again, at very high loading
amplitudes, the cyclic softening/hardening processes produced by the cyclic loading itself push
the influence of the initial work-hardening state
into background.
In Fig. 49, the ratio DRf /DHV, which gives
the increase of alternating bending fatigue
strength per unit hardness, is plotted as a function of the hardness itself (Ref 58, 59). The diagram shows that the sensitivity of the fatigue
strength to the hardness decreases with increasing hardness. Taking into account that the resistance of the macro residual stress state against
cyclic relaxation is low in low-strength steels, it
becomes clear from Fig. 49 thatbesides the
change of the surface topographythe alteration of the micro residual stress state is the governing parameter for the assessment of the influence of machining on the fatigue behavior of
components made from low-strength steel.
Medium-Strength Steel. A detrimental effect
of tensile grinding residual stresses appears in
Fig. 34 and 36 on alternating bending fatigue
strength, being more pronounced for smooth
than for notched specimens. Therefore, the fatigue notch factor decreases with increasing residual stress, as shown in Fig. 50. There is relaxation of surface residual stresses during
bending fatigue loadings of smooth and notched
specimens, which results in lifetimes of approximately 2 million cycles, as shown by Fig. 51(a).
The stress relaxation in the notched specimens
is significantly more pronounced than in smooth
ones. An elastic estimation of the stress in the
notch root r* kt rn,a yields a much higher
value than the stress amplitude in the smooth
specimens. Even though r* is only an upper
bound for the true stress at the notch root, it becomes clear that it is the higher local cyclic loading in notched specimens that causes the stronger
cyclic stress relaxation and, hence, the smaller
residual stress sensitivity of the bending fatigue
strength compared to smooth specimens. Regarding the relatively high loadings indicated in
Fig. 51(b), stress relaxation in notched specimen
is almost complete at the end of fatigue life, and
therefore only a small influence of initial macro
residual stresses on fatigue life at high loadings
is expected. This is shown in Fig. 34.

Residual Stresses and Fatigue Behavior / 41


According to Fig. 34 and 36, the beneficial
effect of compressive grinding residual stresses
on bending fatigue strength is small regarding
smooth and almost negligible regarding notched
specimens. Again, the smaller effect in notched
specimens is due to a stronger residual stress relaxation. However, during cyclic loading corresponding to the fatigue limit, a considerable portion of the initial residual stresses is retained to
the ultimate number of cycles even in notched
specimens. Therefore, and in view of the small
penetration depth of the compressive residual
stresses (
20 lm, see Fig. 35), one would assume that the crack initiation site is shifted below the surface. However, this could not be
proved unambiguously in either smooth or
notched specimens (Ref 49). Therefore, it is
thought that very thin surface zones bearing
compressive residual stresses do not retard the
initiation and early growth of microcracks.
Regarding smooth specimens in Fig. 38, a
smaller effect of compressive residual stresses
produced by grinding or shot peening appears
compared to tensile grinding residual stresses,
even though crack initiation occurred always at
the surface. The small influence of compressive

rs

Residual stress (l ), MPa

n,a = 400 MPa

500

N 107
Before loading

1000

1500
0

0.1

0.2

0.3

0.4

0.5

Distance from surface, mm


Residual stress of smooth ground and additional shot peened specimens made from
quenched-and-tempered (600 C/2 h) plain carbon SAE
1045 steel before loading and after 107 cycles versus distance from surface. Source: Ref 4649

Fig. 52

rs

Residual stress (l ), MPa

0
n,a = 460 MPa

N 103
n,a = 300 MPa

500

N 107
Before loading

1000

1500
0

0.1

0.2

0.3

0.4

0.5

Distance from surface, mm


Residual stress of notched ground and additional shot peened specimens made from
quenched-and-tempered (600 C/2 h) plain carbon SAE
1045 steel before and after loading at different nominal
stress amplitudes versus distance from surface. Source: Ref
4649

Fig. 53

grinding residual stresses was already attributed


to their small penetration depth. Contrarily, as
evident from Fig. 40, compressive residual
stresses generated by shot peening have a rather
large penetration depth. However, since the
bending fatigue strength of shot peened specimens (414 MPa) is much higher than that of
specimens with tensile grinding residual stresses
(195 MPa), also the residual stress relaxation is
more pronounced during corresponding cyclic
loadings. As shown in Fig. 52, residual stresses
relax significantly during fatigue loading, which
does not cause failure within 107 cycles. Furthermore, the surface roughness of the shot
peened specimens (Rh 35.6 lm) was much
higher than that of ground specimens (Rh 6.5
lm) (Ref 4649). If the bending fatigue strength
of shot peened specimens is corrected for the
high surface roughness by means of Fig. 47 and
is plotted in Fig. 38 at the surface residual stress
remaining after 107 cycles (see Fig. 52), one gets
a data point that lies significantly above the extrapolated relationship between bending fatigue
strength and tensile residual stresses.
The relationship between bending fatigue
strength and surface residual stresses of notched
specimens (Fig. 38) is described by a straight
line except for specimens with compressive
grinding residual stresses as already discussed.
In all cases, crack initiation occurred at the surface of the specimens that failed. Since the extrapolated straight lines for smooth and notched
specimens cross at a tensile surface residual
stress of approximately 750 MPa the fatigue
notch factor approaches a unity with increasing
tensile residual stress (see the paragraphs on medium-strength steel in the section Haigh Diagram in this article). Again, different residual
stress relaxation plays an important role. As
shown in Fig. 51(a) (Ref 49), and discussed previously, the residual stress relaxation in a smooth
specimen loaded just above the fatigue limit,
which failed after 2 million cycles, is significantly smaller than in a notched specimen that
failed after approximately the same number of
cycles. Figure 53 proves that residual stress relaxation is even more significant up to a distance
of 100 lm from the surface during fatigue loading of shot peened notched specimens, which did
not fail up to 107 cycles. During fatigue loading
resulting in a fatigue life of approximately 104
cycles, residual stress relaxation is almost complete. Therefore, no influence of macro residual
stresses appears in the low-cycle fatigue range
(see Fig. 37b).
From Fig. 51(a) and 53, it follows that at the
fatigue limit, the residual stress relaxation in shot
peened notched specimens with compressive
tensile residual stresses is more pronounced than
in ground notched specimens with tensile residual stresses. For this comparison, in Fig 51(a)
only the stress relaxation up to the appearance
of a crack (arrow at 5 105 cycles) may be
regarded, because the final relaxation is influenced by crack propagation and does not exist
in Fig. 53. Furthermore, the surface roughness
of shot peened notched specimens (Rh 18.0

lm) was larger than that of ground specimens


(Rh 5.5 lm). Therefore, the question arises
why the data points of shot peened specimens do
not fall below the solid line drawn in Fig. 38.
Careful inspection of microhardness distribution
and the width of x-ray interference lines as a
function of the distance from the surface show
thatcompared to grindingshot peening produces a more intense work hardening which, in
addition, penetrates into a larger depth. The same
holds for smooth specimens. The work hardening reduces cyclic plasticity and retards crack
initiation. Hence, it becomes clear that the bending fatigue strength of medium strength steel is
strongly influenced by both the micro and the
macro residual stress states and their stability
against cyclic loading. However, in the low-cycle fatigue (LCF) range, the macro residual stress
state is unstable and the influence of the (more
stable) micro residual stress state vanishes in
view of the strong cyclic-softening processes occurring during this loading condition (see the
section Cyclic Deformation Behavior in this
article). Therefore, neither an influence of the
macro residual stress state nor an influence of
the micro residual stress state appears in Fig. 37
at high loading amplitudes.
Figure 40 shows that rather different depth
distributions of compressive residual stresses
can be produced in smooth specimens (g 1
mm1) made from a medium-strength steel by
shot peening with different peening parameters
(see also the article Stability of Residual
Stresses in this Handbook). To understand the
influence of the surface roughness, the work
hardening and the macro residual stress state
produced by shot peening on the resulting bending fatigue behavior it is necessary to discuss
crack initiation and crack propagation separately, and this is done in the section Propagating and Nonpropagating Cracks in this article.
Some relationships, however, should be pointed
out here. The highest bending fatigue strength
(curve 6 in Fig. 39) was determined for specimens shot peened according to curve 4 in Fig.
40. Then, a 100 lm thick surface zone was electrolytically removed, shifting the maximum
compressive residual stresses to the surface and
reducing the surface roughness. The smooth surface (Rh 3 lm) and the work hardening of the
surface zone oppose crack initiation. The relatively high compressive residual stresses at the
surface hinder microcrack propagation. This influence outweighs the larger maximum and the
greater penetration depth of the compressive residual stresses according to curve 5. In this shot
peening state, a higher surface roughness (Rh
43.8 lm) and a lower work hardening at the surface result in crack initiation at a lower stress
amplitude, and the lower surface compressive residual stresses offer less resistance to early microcrack propagation. On the other hand, in the
range of finite fatigue life, there is almost no advantage of the electropolished state 6 compared
to the initial shot peened state 4 and a clear disadvantage compared to state 5. Here, the higher
number of cycles to crack initiation (state 6) is

42 / Effect of Materials and Processing


almost compensated (state 4) or overbalanced
(state 5) by a slower rate of crack propagation
due to the larger penetration depth andin the
case of state 5the higher amount of the compressive residual stresses. Since the surface
roughness, the surface work-hardening state and
the surface residual stress state (see Fig. 40) of
the material states 2 to 5 do not differ much, they
have similar resistance to crack initiation at the
surface. Hence, the different fatigue strength appearing in Fig. 39 is determined by the appertaining resistance to microcrack propagation,
which, in turn, depends on the depth distribution
of the compressive residual stresses. This is discussed in more detail in the section Propagating
and Nonpropagating Cracks in this article.
A very interesting example of the interaction
of changing surface roughness as well as micro
and macro residual stress state in a mediumstrength steel are given in Fig. 41. The strong
increase of the rotating bending fatigue strength
of notched specimens by deep rolling is clearly
attributed to the generation of micro- and macrostresses with a rather large penetration depth. It
is interesting to note that the resistance to microcrack initiation is hardly influenced by the deeprolling treatment (Ref 57), and the increasing
bending fatigue strength is almost entirely combined with an increasing resistance to micro-

Fatigue notch factor, kf

2.5

2.0

kt = 1.7
=2
1.5

1.0
500

500

1000

rs

Surface residual stress (l ), MPa


Fatigue notch factor kf of specimens made from
quenched plain carbon SAE 1045 steel versus
surface residual stress. Source: Ref 4749

Fig. 54

rs

Residual stress (l ), MPa

0
Before loading
n,a = 750 MPa

500

N 107
1000

1500
0

0.1
0.2
0.3
0.4
Distance from surface, mm

0.5

Residual stress of smooth ground and additional shot peened specimens made from
quenched plain carbon SAE 1045 steel before loading and
after 107 cycles versus distance from surface. Source: Ref
4649

Fig. 55

crack propagation, resulting in crack arrest at


stress amplitudes increasing with the deep-rolling force. At relatively high forces, however, the
bending fatigue strength decreases, because then
the impairment of the surface zone by the strong
increase of the surface roughness and microcracking as well as an unfavorable residual stress
state at the surface enhances the formation of
rather long initial cracks that only arrest at reduced stress amplitudes. The bending fatigue
strength of smooth specimens increases only
weakly with increasing rolling force. Regarding
the maximum value, an increase of only 15% is
achieved, compared to 120% in the case of
notched specimens. Since the stress gradient is
rather small in smooth specimens, there is some
speculation (Ref 56, 57) that the crack initiation
site is shifted below the surface by deep rolling.
Then, the small increase of the fatigue strength
mirrors the benefit of omitting surface effects
(such as specimen roughness, corrosion, oxidation) and the effect of a decreasing local stress
amplitude on the crack initiation. This mechanism is discussed in detail in the section Haigh
Diagram in this article. At high rolling forces,
however, the crack initiation occurs at the surface again, where it is enhanced by a high roughness, an increased roughness sensitivity due to
work hardening, and an unfavorable residual
stress state. Therefore, the bending fatigue
strength decreases, and there is no improvement
by deep rolling at all. In the end, the maximum
bending fatigue strength of the notched specimens (remember, evaluated as nominal stress
amplitude) is higher than the maximum bending
fatigue strength of the smooth specimens. These
results show that the gradients of the micro and
the macro residual stresses in comparison to the
gradient of the loading stresses are of great importance, and this will become even more evident in the next section.
High-Strength Steels. The S-N curves of
quenched SAE 1045 steel in Fig. 42 prove that
there is a detrimental effect of tensile grinding
residual stresses on bending fatigue strength and,
contrary to the quenched-and-tempered state (see
Fig. 34) also on finite fatigue life. This is because
there is only a weak relaxation of the initial residual stresses even at rather high stress amplitudes. Similar to quenched-and-tempered specimens, compressive grinding residual stresses
with very small penetration depth (
20 lm, see
Fig. 43) have almost no effect on bending fatigue
strength and finite fatigue life. Therefore and
since in the quenched state the compressive residual stresses are retained to the largest part during fatigue loading in the range of the fatigue
limit, one may expect that crack initiation occurs
below the surface. Analogous to the quenchedand-tempered state, this could not be proved definitely (Ref 49). Since in the quenched state a
grinding-induced work softening occurs close to
the surface and an influence of the surface roughness always exists, it is very probable that crack
initiation occurs at the surface. Then, one must
conclude that compressive residual stresses with
a penetration depth of only 20 lm do not retard

or stop the propagation of microcracks existing


after crack initiation.
The extrapolated bending fatigue strength versus surface residual stresses curves in Fig. 44
cross in the range of tensile residual stresses at
1125 MPa. This means that the fatigue notch
factor becomes unity. On the other hand, at negligible or compressive residual stresses, the fatigue notch factor approaches the stress-concentration factor, as shown in Fig. 54. Qualitatively
similar to the quenched-and-tempered state, the
influence of tensile residual stresses is stronger
than that of compressive residual stresses and the
residual stress sensitivity of notched specimens
is smaller than that of smooth ones. However,
this cannot be explained by different residual
stress relaxation so far. Careful x-ray analyses
prove that in notched specimens loaded with a
stress amplitude of 450 MPa and in smooth specimens loaded with 750 MPa (corresponding to
the fatigue limit of each) the relaxation of compressive residual stress is small and of equal
amount. This is not astonishing, since an elastic
estimation of the stress amplitude in the notch
root (r* kt r 765 MPa) yields almost the
same value as the stress amplitude of the smooth
specimen. No corresponding measurements for
smooth and notched specimens with tensile residual stresses were performed. However, since
the fatigue limit is reduced by tensile residual
stresses, it is clear that their relaxation is negligible.
A comparison of Fig. 45(a) with Fig. 37(a)
shows that the shot peening treatments performed increases the bending fatigue strength of
smooth specimens in the quenched state less
than in the quenched-and-tempered state. In
view of the different residual stress states and
their different stabilities in both heat treating
states, this finding is rather astonishing. In Fig.
55, the depth distribution of residual stresses in
shot peened quenched specimens before fatigue
loading and after 107 cycles at the fatigue limit
is shown. In the range of specimen scatter, the
results are identical. Contrary to quenched-andtempered specimen (see Fig. 52), the shot peening residual stresses are much higher and much
more stable in quenched specimens. However,
in the former, crack initiation occurs at the surface. Hence, the shot-peening-induced changes
of the micro and the macro residual stress state
have a strong influence on the measured bending
fatigue strength. In smooth quenched specimens,
cyclic loading at the fatigue limit initiates cracks
below the surface. Therefore, the influence of the
shot-peening-induced changes of the residual
stress state have a rather small influence on the
measured bending fatigue strength. This is
treated more detailed in the section Modeling
the Influence of Residual Stresses on Fatigue Behavior. Contrarily, in the range of finite fatigue
life, crack initiation in both states occurs at the
surface, and there is a strong influence of shot
peening regarding quenched specimens and a
very small one regarding quenched-and-tempered specimens. This finding is directly correlated with very different residual stress relaxa-

Residual Stresses and Fatigue Behavior / 43

Bending fatigue strength (Rf), MPa

tion. While surface residual stress relaxation is


definitely complete in quenched-and-tempered
specimens after 50,000 cycles at a stress amplitude of 570 MPa, there is hardly any change of
the surface residual stresses in the quenched
specimens after the same number of cycles even
at 1200 MPa (Ref 49), although there may be
some residual stress relaxation below the surface.
Regarding notched specimens, a different picture appears (compare Fig. 45b with Fig. 37b).
Now, shot peening of quenched specimens pro-

800
Smooth
= 0.4

600

400
Ground
Shot peened
Milled

200

Notched
kt = 1.7
=2

0
1500 1000 500
0
500
1000
rs
Maximum residual stress (max ), MPa

Alternating bending fatigue strength of smooth


and notched specimens made from quenched
plain carbon SAE 1045 steel with different surface conditions versus surface residual stress

Fig. 56

Before loading
500

rs

Residual stress (l ), MPa

n,a = 640 MPa

N 107
1000

1500
0

0.1
0.2
0.3
0.4
Distance from surface, mm

0.5

(a)

after
107 cycles. In this case, a significant relaxation and redistribution of the residual
stresses occur.

Modeling the Influence of Residual


Stresses on Fatigue Behavior
Haigh Diagram
Basic Relationships. One obvious way to account for the influence of (macro) residual
stresses on the fatigue behavior is to treat them
as local mean stresses. In doing so, one has to
realize that there are several important differences between (loading) mean stresses and residual stresses, as discussed in the section Comparison between Loading Stresses and Residual
Stresses in this article. Figure 58 shows a Haigh
diagram for smooth and notched specimens
made from a medium-strength steel (Ref 1, 62).
The Goodman approximation (see Eq 9) is used
to account for the influence of residual stresses
on the fatigue strength. If the amount of the minimum stress or the maximum stress in smooth
specimens does not exceed the critical stress amplitude ra,crit, which is a function of the cyclic
yield strength (see the article Stability of Residual Stresses in this Handbook), the residual
stresses do not relax, and the line AB gives the
influence of the residual stress on the fatigue
strength. Then, all combinations of residual
stress and stress amplitude inside the shaded area
result in neither residual stress relaxation nor fatigue failure. However, if the amount of the minimum stress or the maximum stress exceeds the
critical stress amplitude, it is assumed that the
residual stresses relax to the value given by the
points A and B, respectively, and the fatigue
strength remains constant at the value given by
these points. In the case of notched specimens,
the cyclic yield strength and the notch fatigue
strength (both in terms of nominal stress amplitudes) are less than the respective values of
smooth specimens. However, the ultimate tensile
strength of notched specimens is larger than that
of smooth ones in such a material state because

Before loading
500

rs

Residual stress (l ), MPa

duces a strong increase of both fatigue limit and


finite fatigue life. Milling, which may generate
rather high compressive residual stresses, is less
effective. In all cases, crack initiation was observed at the surface. However, as evident from
Fig. 46, the surface residual stress state is not an
appropriate parameter to assess bending fatigue
life in the presence of compressive residual
stresses. In Fig. 56, the bending fatigue strength
is plotted as a function of the maximum residual
stresses. Again, the arrows mark the shift of the
residual stresses and bending fatigue strength
obtained from shot peening. Now, the data of
ground notched specimens with tensile or negligible residual stresses and of notched specimens peened with shot of different hardness are
described satisfactorily by one common line
covering a residual stress range of 2350 MPa.
The data points of ground and those of milled
specimens bearing compressive residual stresses
fall below this line, which has the slope 0.214.
In Fig. 56, it becomes even more clear than in
Fig. 46 that the improvement of the bending fatigue strength of smooth specimens by shot
peening is very limited. As a consequence, the
notch effect on bending fatigue strength is completely eliminated at high compressive residual
stresses.
It is interesting to look at the stability of very
high compressive residual stresses. As already
shown in Fig. 55, fatigue loading of smooth
specimens in the range of the fatigue limit does
not cause any significant change of the residual
stress state. As shown in Fig. 57(a), the same
holds for notched specimens peened with shot of
relatively low hardness, which have a very similar depth distribution of the residual stresses.
Elastic calculation of the minimum stress 70 lm
below the notch root (at the maximum of the
compressive stresses) yields the value 2035
MPa. This stress value may occur more than 107
times without effecting any change of the residual stress state. As shown in Fig. 57(b), notched
specimens peened with shot of higher hardness
bear even higher compressive residual stresses.
In this case, elastic calculation of the minimum
stress in a depth of 100 lm yields the initial
value 2475 MPa and the value 2130 MPa

n,a = 760 MPa

1000

N 107
1500
0

0.1
0.2
0.3
0.4
Distance from surface, mm

0.5

(b)
Residual stress of notched ground and additional shot peened specimens made from
quenched plain carbon SAE 1045 steel before loading and
after 107 cycles versus distance from surface. (a) Shot
peened with shot of a hardness of 46 to 50 HRC. (b) Shot
peened with shot of a hardness of 54 to 58 HRC. Source:
Ref 4649

Fig. 57

Fig. 58

Haigh diagram. Bending fatigue strength Rf of smooth and notched specimens made from a medium-strength
steel versus residual stress. After Ref 1, 62

44 / Effect of Materials and Processing


of the triaxial stress state in the interior of the
notched specimens. Now, the Goodman relationship holds between points C and D, and residual
stress relaxation occurs outside the lightly
shaded area. From these relationships, it is expected that the residual stress sensitivity of
notched specimens is less than that of smooth
specimens.
Application to a High-Strength Steel. Figure 59 shows Haigh diagrams for smooth and
notched specimens made from a high-strength
steel, which are adopted to Fig. 44, 46, and 56.
In Fig. 59(a), the data points give the correlation

(a)

(b)
Haigh diagrams. Bending fatigue strength Rf of
smooth and notched specimens made from
quenched plain carbon SAE 1045 steel with different surface conditions. (a) Versus surface residual stress. (b) Versus
maximum residual stress

Fig. 59

between the residual stress and the nominal


stress amplitude at the surface. In Fig. 59(b), the
data points give the correlation between the maximum residual stress and the nominal stress amplitude at the locus of the maximum residual
stress. In the range of tensile residual stresses,
this is always the surface. The arrows connecting
some data points illustrate the residual stress relaxation that occurs during bending fatigue loading. The ultimate tensile strength Rm 1910
MPa of smooth specimens was taken from Ref
49. The ultimate tensile strength Rm,no of notched
specimens is unknown, but the value 2000 MPa,
which results from an extrapolation of the data
given in Fig. 44, is reasonable for this highstrength material state and the stress-concentration factor 1.7 (Ref 63, 64). The cyclic yield
stresses of smooth and notched specimens are
also unknown. However, using the results shown
in Fig. 55 (smooth specimens) and Fig. 57(a)
(notched specimens), a border line for residual
stress relaxation can be obtained. As obvious
from the figures and discussed in the previous
section, the cyclic loading indicated there causes
no residual stress relaxation. The corresponding
data points are given in Fig. 59(b) as open square
(smooth specimens) and open circle without arrow (notched specimens). Assuming that residual stress relaxation starts at a limiting amount
of the minimum or maximum stress irrespective
of the fractions of the mean stress and the stress
amplitude, one gets the shaded areas in which
the residual stresses are stable. These areas are
transferred to Fig. 59(a), too. The two open circles connected with an arrow illustrate the
change of the amount of the surface residual
stress (Fig. 59a) and of the minimum stress (Fig.
59b) during the cyclic loading of notched specimens indicated in Fig. 57(b). In milled specimens, there is also some residual stress relaxation as indicated by the triangles connected with
an arrow. Obviously, the amounts of the minimum stresses are reduced to a value corresponding to the border line, confirming this estimation.
Extrapolation of this line to zero residual stress
yields the values 1785 MPa (smooth specimens)
and 1660 MPa (notched specimens), respectively, which are higher than the corresponding
yield strengths (Ref 49).
Even though there are some uncertainties in
establishing the relationships in Fig. 59, the following conclusions can be drawn for a highstrength steel with tensile residual stresses:
Residual stress relaxation has no influence re-

Haigh diagram. Bending fatigue strength Rf of


smooth and notched specimens made from
quenched-and-tempered (600 C/2 h) plain carbon SAE
1045 steel with different surface conditions versus surface
residual stress

Fig. 60

garding the effect of tensile residual stresses


on the fatigue strength.
The influence of tensile residual stresses on
the bending fatigue strength corresponds satisfactorily to the Goodman approximation.
This means that the residual stress sensitivity
and the mean stress sensitivity of the bending
fatigue strength is almost the same.
In this formalism, the different residual stress
sensitivity of smooth and notched specimens,
which causes a reduction of the fatigue notch
factor with increasing residual stress (see Fig.
54), is related to the different stress states in

smooth and notched specimens via the different ultimate tensile strengths.
In the range of compressive residual stresses,
the relationships are much more difficult. The
fatigue strength of smooth specimens is much
less than expected by the Haigh diagrams, regardless of whether the existing surface or the
maximum residual stresses are concerned. In the
case of shot peened specimens, this finding is
clearly related to subsurface crack initiation.
Therefore, neither the stress state at the surface
(Fig. 59a) nor the stress state in the depth of the
maximum compressive stresses (Fig. 59b) is directly relevant for the fatigue strength. Regarding notched specimens, the fatigue strength of
shot peened states is much higher than expected
by the Haigh diagram, if it is correlated with the
surface residual stresses (Fig. 59a). This simply
means that crack initiation must occur at the surface, if the specimens are loaded with amplitudes
in the range of the fatigue limit. On the other
hand, the stress amplitude in the depth of the
maximum of the compressive stresses of shot
peened specimens is somewhat lower than expected from the Haigh diagram (Fig. 59b). The
same holds for the data points of milled specimens, which have maximum compressive residual stresses at the surface. These findings can
only be understood if crack initiation and crack
propagation or crack arrest are treated separately.
This is done in the sections Concept of Local
Fatigue Strength and Propagating and Nonpropagating Cracks.
Application to a Medium-Strength Steel. In
medium-strength steels, residual stress relaxation definitively influences the relationships in
the Haigh diagram. Since the manufacturing process, which generates the macro residual
stresses, almost always changes the micro residual stresses and, hence, the local cyclic yield
strength, there are no unique borderlines for the
onset of residual stress relaxation as assumed in
Fig. 58. Nevertheless, Fig. 60 tries to bring the
results shown in Fig. 36 and 38 into a Haigh
diagram.
Regarding smooth specimens, the ultimate
tensile strength was taken from Ref 49. In
ground specimens with tensile residual stresses,
stress relaxation (marked by an arrow) brings the
corresponding data point rather close to the
Goodman line. The border lines for the onset of
residual stress relaxation are constructed using
this data point. However, these border lines are
not valid for shot peened specimens, in which
the residual stresses relax to a much smaller extent than expected as shown by the lower arrow.
Obviously, the resistance to residual stress relaxation is enhanced by shot peening, which results
in a stronger work hardening and, hence,
stronger increase of micro residual stresses than
grinding. Work hardening, however, should also
increase the fatigue strength at zero residual
stress and, hence, should produce deviations
from the solid line established with the aid of the
data points of ground specimens. However, as
already discussed in the paragraphs on medium-

Residual Stresses and Fatigue Behavior / 45

SWT parameter, MPa

strength steel in the section Evaluation of Experimental Results, there is a detrimental effect
of a very high surface roughness. The second
arrow illustrates the correction, which is based
on Fig. 47, for this effect. Compressive residual
stresses produced by grinding show some relaxation, although their initial value is inside of the
border lines. This finding is consistent with the
consideration given so far, since work hardening
is almost negligible in this specimen. Moreover,
the data point corresponding to the relaxed compressive grinding residual stresses still falls below the Goodman line. This finding was already
attributed to the low penetration depth of the
grinding residual stresses.
The residual stress sensitivity of notched specimens with kt 1.7 is significantly smaller than
that of smooth specimens, even though the consideration is based on the relaxed residual
stresses. Therefore, the different residual stress
sensitivity of the fatigue strength of smooth and
notched specimens, respectively, is not solely
based on different residual stress relaxation, and
the third conclusion given previously for a highstrength steel is also valid for a medium-strength
steel to some extent. From an extrapolation of
the Goodman line plotted, an ultimate tensile
strength of 1133 MPa is obtained, which is significantly higher than the ultimate tensile
strength of smooth specimens. Again, this finding results from the triaxial stress state in the
interior of the notched specimens and is consistent with data measured on other mediumstrength steels (Ref 63, 64). Regarding material
states with compressive residual stresses, the
data point of milled specimens comes rather
close to the Goodman line. No measurements
were done concerning the stability of this residual stress state. Since the data point falls inside
of the border lines, the residual stress state
should be stable. Contrarily, there is some relaxation of the compressive residual stresses in
ground specimens even though their initial value
falls inside of the border lines, too. Again, this
is a consequence of different micro residual
stress states. Similar to smooth specimens, the
data points fall below the Goodman line as a
result of the small penetration depth of the com-

1000
800

SAE 1045

600
400

AISI 304

300
200

100
103

Untreated
Shot peened
Deep rolled
105
106
104
Number of cycles to failure, Nf

107

Smith-Watson-Topper parameter versus number of cycles to failure for specimens made


from normalized SAE 1045 and AISI 304 steel in different
surface conditions. After Ref 28

Fig. 61

pressive grinding residual stresses. The same


holds for notched specimens with kt 2.5, g
5. The data points of shot peened specimens fall
above the expected relationship as a consequence of pronounced work hardening, which
increases both fatigue strength and resistance of
the compressive residual stresses against relaxation.
From the relationships shown in Fig. 60, the
following conclusions can be drawn for a medium-strength steel:
Residual stress relaxation plays a significant

role for the influence of residual stresses on


fatigue behavior.
Residual stress relaxation depends on the
work hardening produced by the various
manufacturing processes. Not only the
amount, but also the penetration depth of
work hardening is important. As a consequence, even for one initial material state
there are no unique borderlines for the onset
of residual stress relaxation.
If comparable work-hardening states (or micro residual stress states) are concerned, the
influence of relaxed residual stresses on the
bending fatigue strength corresponds satisfactorily to the Goodman approximation. This
means that the relaxed residual stress sensitivity and the mean stress sensitivity of the
bending fatigue strength is almost the same.
In this formalism, the different initial residual
stress sensitivity of smooth and notched specimens, which causes a reduction of the fatigue
notch factor with increasing residual stress, is
related to both the different residual stress relaxation and the different stress states in
smooth and notched specimens via the different ultimate tensile strengths.

Damage Parameters
As discussed in the section Lifetime Behavior, there are some so-called damage parameters
that correlate fatigue life with loading parameters. The well-known Manson-Coffin relationship (Eq 5) is not applicable to material states
with residual stresses, because it does not account for mean stresses. On the other hand, the
Ostergren parameter (Eq 12) and the Smith-Watson-Topper parameter (Eq 11) include the mean
stress via the maximum stress. With the limitations already outlined in the previous section,
one could try to use these parameters to assess
the effect of residual stresses on fatigue life.
However, the application of the Ostergren parameter requires the knowledge of the local
plastic strain amplitude, which in turn needs sophisticated modeling of the local cyclic elasticplastic deformation behavior. This seems not to
be a useful approach. However, the Smith-Watson-Topper parameter (SWT parameter) is easier
to evaluate and was used in Ref 28 to account
for surface treated steels (normalized SAE 1045
and AISI 304). Figure 61 shows characteristic
plots of the SWT parameter versus the number
of cycles of stress-controlled tests on both steels

in shot peened as well as deep-rolled conditions


in comparison with the untreated states. For each
steel, a unique curve is obtained for the three
material conditions. Hence, concerning these
steels the quantities that are included in the SWT
parameterthe maximum stress and the total
strain amplitude (see Eq 11)form a reliable
basis for the assessment of the influence of residual stresses on the fatigue life.

Concept of Local Fatigue Strength


Basic Relationships. As shown in the previous section, in the paragraphs on applying the
Haigh diagram to high-strength and mediumstrength steels, the use of the Haigh diagram for
the assessment of the effect of residual stresses
on the fatigue behavior has serious limitations.
Regarding, for example, smooth specimens
made from a high-strength steel, shot peening
produces much less improvement in fatigue
strength than expected from the residual stress
state. This finding is combined with subsurface
crack initiation. Regarding medium-strength
steel, difficulties arise because not only the
(macro) residual stress state is changed by manufacturing processes, but also the micro residual
stress state. The latter influences the fatigue
strength at zero mean stress and the resistance of
the residual stress state against relaxation. Therefore, it is appropriate to look for the local fatigue
strength, which depends on the local micro residual stress state, the local macro residual stress
state, and (as far as crack initiation at the surface
is concerned) on the topography of the surface
of the component. Strictly speaking, the local resistance against fatigue crack initiation is concerned, as shown later.
The concept of the locally effective fatigue
strength, which has its origin in Ref 38 and 39,
enables quantitative predictions of the effect of
the depth distributions of residual stresses on the
locus of crack initiation as well as on the fatigue
strength. The basic assumption of the concept is
that a crack can only be initiated at or below the
surface if the local loading stress exceeds the local fatigue strength. Especially in the case of
relatively hard materials, for example, hardened
steels, this concept yields to a good estimation
of the corresponding properties. For that purpose, it is necessary to have a good knowledge
of the depth distributions of the fatigue strength
in the residual-stress-free condition R0f (z) as well
as of the macro residual stress rrs(z) and the residual stress sensitivity m(z). The locally effective fatigue strength Rf(z) as a function of the
distance x from the surface is calculated by the
relationship:
Rf(z) R0f (z) m(z) rrs(z)

(Eq 18)

where the residual stress sensitivity m(z) approaches the mean stress sensitivity M of the
Goodman relationship (e.g., Eq 9) if the residual
stresses are stable. Then, the residual stress sensitivity is determined approximately by m(z)

46 / Effect of Materials and Processing


R0f (z)/Rm(z) (Ref 39). However, if residual stress
relaxation occurs, the residual stress sensitivity
m is smaller than the mean stress sensitivity M,
if the initial residual stress distribution is used
(Ref 39, 62). However, if the relaxed residual
stress distribution is used, the residual stress sen-

sitivity m again approaches the mean stress sensitivity M as shown in the paragraphs applying
the Haigh diagram to medium-strength steels in
the previous section. The depth distributions of
the tensile strength Rm(z) and of R0f (z) can be
estimated from appropriate correlations with

Notched
Residual stress, rs

Residual stress, rs

Smooth
0

(a)

(b)

Distance from surface, mm

Distance from surface, mm

N f = 104

a and Rf,loc

a and Rf,loc

N f = 104

Nf =

Local fatigue strength


Loading stress
0

Nf =
0

Distance from surface, mm

(c)

Distance from surface, mm

(d)

Distribution of residual stress (a) and (b) and distribution of loading stress and local fatigue strength for loadings
in the range of finite and infinite lifetime (c) and (d) for smooth (a) and (c) and notched (b) and (d) specimens
made from a high-strength steel

Fig. 62

Propagating crack
Not propagating
crack

d = 0.6 mm
1000

Rf
800

Loading
stresses

1200

d = 0.3 mm

a and Rf,loc, MPa

a and Rf,loc, MPa

1200

1000

Rf
800

600

600
0

0.1
0.2
0.3
Distance from surface, mm

0.4

400

rs

400
800

1200

0.4

0.1
0.2
0.3
Distance from surface, mm

0.4

400
800

1200
0

(c)

0.1
0.2
0.3
Distance from surface, mm

(b)
400
l, MPa

rs

l, MPa

(a)

0.1
0.2
0.3
Distance from surface, mm

0.4
(d)

Distribution of the local fatigue strength and the loading stress for different stress amplitudes with marking of
the locus of crack initiation (a) and (b) and distribution of residual stress (c) and (d) for specimens made from
a quenched SAE 1045 steel after shot peening with shot with a mean diameter of 0.6 mm (a) and (c) and 0.3 mm (b) and
(d), respectively. After Ref 38, 39

Fig. 63

measured depth distributions of the hardness


(see e.g., Ref 65).
Application to High-Strength Steels. In a
high-strength steel, manufacturing processes
may generate very large (macro) residual
stresses, as shown by several examples in the
paragraphs on high-strength steel in the section
Characteristic Examples. Contrarily, the
changes of the micro residual stress state are
rather small. Therefore, the local fatigue strength
is almost entirely determined by the initial material state and the depth of the (macro) residual
stresses. Figure 62(a) and (b) show depth distributions of compressive residual stresses in
smooth and notched bending specimens produced by shot peening. In Fig. 62(c) and (d), the
resulting depth distributions of the fatigue
strength for infinite life and for one arbitrarily
chosen finite life are plotted together with the
corresponding depth distributions of the loading
stress. It becomes clear from this figure that the
improvement of the fatigue strength for infinite
life of smooth specimens (low stress gradient)
by shot peening is small because crack initiation
occurs below the surface in a depth with vanishing residual stresses. Contrarily, with increasing
stress amplitude, the crack initiation site is
shifted toward the surface, and the shot-peeninginduced stresses influence crack initiation and
early fatigue crack growth. Hence, there is a
large influence of shot peening on the low-cycle
fatigue strength and a small influence on the
high-cycle fatigue strength. This is shown in Fig.
45(a). Contrarily, crack initiation in notched
specimens (high stress gradient) occurs always
at the surface and both low-cycle and high-cycle
fatigue strength are improved, as shown in Fig.
45(b).
The assumption of the local fatigue concept
that fatigue life determining cracks initiate only
in places where the loading stress exceeds the
local fatigue strength is verified by several experimental results (e.g., Ref 38, 39, 62). Figure
63 presents for two series of specimens of the
aforementioned investigations from Ref 38 and
39 on hardened and shot peened SAE 1045 steel
graphs of the depth distribution of the local fatigue strength (see Fig. 63a and b). Rf,loc can be
calculated from the compressive residual stress
distributions (Fig. 63c and d) under the assumption of uniform hardness. The closed circles
mark the initiation sites of damage-relevant
cracks for each load stress distribution, which
could be taken from SEM of the fracture surface
(see Fig. 24). As can be seen, the actual crack
initiation positions agree with the expectation after the concept of the local fatigue strength. For
high stress amplitudes (1150 MPa), the cracks
that determine fatigue life can initiate at the surface. For lower stress amplitudes (1150 MPa),
the damage behavior is changed. Of course,
stress amplitudes between 1000 and 1100 MPa
are able to initiate cracks at the surface (open
circles in Fig. 58a and b). However, these stress
amplitudes are too small to propagate the cracks
into greater depth because of the high compres-

Residual Stresses and Fatigue Behavior / 47


sive residual stresses existing there. This aspect
is treated in more detail in the paragraphs on
high-strength steel in the section Propagating
and Nonpropagating Cracks. At stress amplitudes below 1000 MPa, cracks are generated
only beneath the surface as expected by the plotted straight lines of the loading stress in comparison with the courses of the local fatigue
strength (Fig. 62a and b). The fatigue strength Rf
is determined by the maximum loading stress
amplitude, which in any cross section of the
specimen exceeds the local fatigue strength and
initiates a crack that is able to propagate. The
corresponding distributions are plotted in Fig.
63(a) and (b) as dashed straight lines.
Figure 64 applies the concept of local fatigue
strength to shot peened smooth and notched
specimens according to Fig. 55 to 57. In notched
specimens peened with shot of low (Fig. 64a)
and high (Fig. 64b) hardness and loaded in the
range of the fatigue strength, crack initiation is
predicted at the notch root, as actually observed.
However, the surface loading stress amplitude is
much higher than the surface fatigue strength.
The same result is obtained from the Haigh diagram in Fig. 59(a). Therefore, one has to assume that loading stress amplitudes below the
fatigue limit can initiate a crack at the surface

and that the fatigue strength is not determined


by the maximum loading stress amplitude that
does not initiate a crack. This shows the limitations of the concept of the local fatigue strength
and is treated in more detail in the paragraphs on
high-strength steel in the section Propagating
and Nonpropagating Cracks.
Figure 64(c) shows that crack initiation in
smooth specimens is expected below the surface
where the local fatigue strength is hardly increased by compressive residual stresses. This is
verified by metallographical investigation (Ref
46, 49). This is the reason for the small improvement of the bending fatigue strength by shot
peening shown in Fig. 45, 46, 56, and 59. However, from Fig. 45(a) it becomes clear that shot
peening significantly improves finite fatigue life.
This finding cannot be understood on the basis
of the local fatigue strength alone and is treated
in the paragraphs on high-strength steel in the
section Propagating and Nonpropagating
Cracks.
In the case of notched specimens finished by
milling (see Fig. 45b, 46, and 56), the concept
of local fatigue strength predicts crack initiation
below the surface, because the gradient of the
residual stresses is steeper than the gradient of
the loading stresses. Hence, the increase of the

a and Rf,loc, MPa

1500
Local fatigue
strength
1000

Propagating and Nonpropagating


Cracks

Loading stress

500

0
0

0.2

0.4

Distance from surface, mm


(a)

0.2

0.4

Distance from surface, mm

0.2

0.4

Distance from surface, mm

(b)

(c)

Distribution of the local fatigue strength and the loading stress with marking of the locus of crack initiation
for notched (a) and (b) and smooth (c) specimens made from a quenched SAE 1045 steel after shot peening
with a shot of a hardness of 46 to 50 HRC (a) and 54 to 58 HRC (b) and (c), respectively. Source: Ref 49

Fig. 64

a and Rf,loc, MPa

1500

1000

Local fatigue
strength

500
Loading stress
0
0

0.2

0.4

Distance from surface, mm


(a)

0.2

0.4

Distance from surface, mm


(b)

Distribution of the local fatigue strength and the loading stress with marking of the locus of crack initiation
for notched (a) and smooth (b) specimens made from a quenched-and-tempered (600 C/2 h) SAE 1045 steel
after shot peening with a shot of a hardness of 46 to 50 HRC

Fig. 65

bending fatigue strength is smaller than expected


in view of the high surface compressive residual
stresses (see the Haigh diagram in Fig. 64a), but
larger than in shot peened smooth specimens
with subsurface crack initiation. This is a consequence of the higher loading stress gradient in
notched specimen compared to smooth specimens. However, crack initiation below the surface of these specimens was not proved (Ref 49).
Application to Medium-Strength Steels.
The application of the concept of the local fatigue strength to medium-strength steels is somewhat more difficult compared to high-strength
steels because residual stress relaxation becomes
important. Therefore, a knowledge of the initial
depth distribution of the residual stresses and of
the residual stress sensitivity m or of the relaxed
residual stress distribution is necessary. In Fig.
65, the concept is applied to shot peened smooth
and notched specimens made from quenchedand-tempered SAE 1045 steel. The initial and
relaxed residual stress distributions are given in
Fig. 52 and 53. Again, in notched specimens
(Fig. 65a) crack initiation at the surface is predicted as actually observed. Regarding smooth
specimens, crack initiation below the surface is
expected. Contrarily, crack initiation at the surface is found in metallographical investigation.
In both cases, the difference between local loading stress and local fatigue strength is small
throughout the depth range under consideration.
Therefore, factors such as the high roughness of
the smooth specimens may well account for the
deviation of the expected locus of crack initiation.

General Remarks. It is well known that in


notched specimens or components that are
loaded in the range of the fatigue strength, cracks
may initiate in the root of sharp notches and may
arrest in a certain depth, where the driving force
for crack propagation falls below its threshold
value because of the steep drop of the loading
stress. In most cases, such sharp notches are not
relevant for components (Ref 57, 66). However,
in the presence of high residual stresses, the interaction of loading stresses and residual stresses
may produce strong gradients of the driving
force for crack propagation making crack arrest
possible even in notches with low stress-concentration factor. On the other hand, in smooth specimens with large compressive residual stresses
at and below the surface, crack initiation may
occur below the surface because the gradient of
loading stresses is small. Then, the question
arises whether or not the crack can propagate
toward the surface where its propagation is hindered by the residual stress field. Both situations
are illustrated by the Haigh diagrams in Fig. 59.
In the case of shot peened smooth specimens,
the fatigue strength falls below the values expected by the Haigh diagrams, because crack initiation occurs below the surface, where no ben-

48 / Effect of Materials and Processing


eficial effect of the compressive residual stresses
is effective. Contrarily, the combination of surface loading stress and surface residual stress occurring at the notch root of specimens with kt
1.7 (Fig. 59a) falls significantly above the corresponding Goodman line. As already men-

Keff, MPa m

(Eq 19)

DKeff kt rn,a p aY Kmax


Without residual stresses

4
Range of Kth,eff
0
0

0.05

0.1

0.15

0.2

0.25

0.3

Distance from surface, mm


Range of effective stress intensity factor DKeff
with and without consideration of the residual
stresses versus distance from surface for ground notched
specimens made from a quenched SAE 1045 steel at cyclic
bending loading (kt 1.7, g 2)

Fig. 66

(Eq 20)

for smooth (kt 1) and notched (kt  1) specimens without residual stresses. rn,a is the nominal stress amplitude. If residual stresses are
present, the maximum stress is increased or decreased, depending on the sign and the amount
of the residual stresses, leading to:
DKeff (kt rn,a
rrs) p aY Kmax

(Eq 21)

if Kmin  0 or:
DKeff kt Drn p aY

60
Without residual stresses
Keff, MPa m

DKeff Kmax Kop  DKth,eff

(see Eq 16). The following concerns alternating


bending fatigue (R 1). In a rough approximation, it is assumed that Kop equals zero, resulting in:

With residual stresses

12

tioned, this means that cracks initiate at the notch


root, but are arrested below the surface.
A crack is arrested when the driving force for
its propagation falls below the threshold value:

Initial residual
stresses

40

Relaxed
residual
stresses

20
Range of Kth,eff
0
0

0.1

0.2

0.3

0.4

0.5

0.6

Distance from surface, mm


Range of effective stress intensity factor DKeff
without and with consideration of the initial
and the relaxed, respectively, residual stresses versus distance from surface for shot peened (shot of a hardness of
54 to 58 HRC) notched specimens made from a quenched
SAE 1045 steel at cyclic bending loading (kt 1.7, g
2)

Fig. 67

(Eq 22)

if Kmin 0.
Application to a High-Strength Steel. In a
high-strength SAE 1045 steel, literature data for
the threshold value Kth,eff fall in the range from
2.2 to 3.2 MPa m (Ref 67). As a first example,
ground notched specimens (kt 1.7, g 2)
with tensile residual stresses (see Fig. 43, curve
with open circles) are treated. Concerning the
half-elliptical cracks that initiate at the notch
root, a geometry factor Y 1 is assumed. The
interaction of this residual stress state with the
loading stress amplitude ra 230 MPawhich
corresponds to the fatigue strengthresults in
the depth distribution of DKeff shown in Fig. 66.
In a depth of 5 lm only, DKeff exceeds the
threshold value. Therefore, a crack that initiates
at the notch root will always propagate into the
interior of the specimen, and the fatigue limit

40

Without residual stresses

30

Relaxed residual
stresses

Keff, MPa m

Keff, MPa m

15

Initial
residual
stresses
Range of Kth,eff

20
10

Initial residual
stresses

Relaxed
residual
stresses

10

Range of Kth,eff

Crack initiation

0
0

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.2

0.4

0.6

0.8

1.0

1.2

Distance from surface, mm

Distance from surface, mm


Range of effective stress intensity factor DKeff
without and with consideration of the initial
and the relaxed, respectively, residual stresses versus distance from surface for shot peened (shot of a hardness of
4650 HRC) notched specimens made from a quenched
SAE 1045 steel at cyclic bending loading (kt 1.7, g 2)

Fig. 68

Range of the effective stress intensity factor


DKeff for subsurface crack initiation without and
with consideration of the initial and the relaxed, respectively, residual stresses versus distance from surface for
ground and additional shot peened smooth specimens
made from a quenched SAE 1045 steel at cyclic bending
loading (g 0.4)

Fig. 69

corresponds to the maximum stress amplitude


that does not initiate a crack. This is in full agreement with the Haigh diagram in Fig. 59(a). It is
interesting to note that DKeff has a local maximum in a depth of 80 lm. Up to a depth of 150
lm, the residual stress field increases significantly DKeff as compared with the loading of a
specimen without residual stresses (dotted line).
A completely different situation prevails regarding shot peened notched specimens (kt
1.7, g 2) with the initial and the relaxed residual stress distributions shown in Fig. 57(b).
As shown in paragraphs on high-strength steel
in the sections Haigh Diagram and Concept
of Local Fatigue Strength, a crack will initiate
at the notch root if a specimen is loaded in the
range of the fatigue strength. Hence, if the crack
has the length a, the loading stress and the residual stress in the distance a from the surface determines if the crack will continue propagation
or not. In Fig. 67, DKeff is plotted versus the
distance from the surface. If the initial residual
stress distribution is concerned, DKeff becomes
zero at a depth of 40 lm, and higher stress amplitudes than the measured bending fatigue
strength would be necessary to propagate a crack
into the interior of the specimen. However, if the
relaxed residual stress distribution is concerned,
DKeff approaches the threshold range between 50
and 200 lm below the surface, and cracks may
or may not propagate. Hence, the fatigue limit
corresponds to the boundary between propagating or nonpropagating cracks that initiate at the
notch root. It is interesting to note that at depths
greater than 0.3 mm, DKeff becomes much larger
than in Fig. 66 as a consequence of the high loading amplitude. This means that the strong increase of finite fatigue life by shot peening is
entirely based on very small crack propagation
rates at small distances from surface.
Regarding shot peened notched specimens
with the residual stress distributions shown in
Fig. 57(a), almost no residual stress relaxation
occurs during loading in the range of the fatigue
limit. Therefore, the depth distributions of DKeff
corresponding to the initial and the relaxed residual stress distributions shown in Fig. 68 are
almost the same. At a depth of 40 lm, DKeff
becomes zero and exceeds the threshold value
only at a depth of 170 lm. Therefore, one would
expect that cracks could still be arrested at stress
amplitudes higher than the measured bending fatigue strength. This shows the limitations of the
simple estimation given here.
Figure 64(c) shows that crack initiation in shot
peened smooth specimens with the residual
stress distribution plotted in Fig. 55 occurs below the surface at a depth of 500 lm. For the
following rough estimation, it is assumed that
the crack has a circular shape (see Fig. 24), that
it propagates with the same rate toward the surface as well as into the interior of the specimen,
and that the residual stresses in depths greater
than 500 lm are zero. If the loading stress in the
depth of the crack front and a geometry factor Y
0.63 (Ref 68) are used for the calculation of
DKeff, one finds the relationship shown in Fig.

Residual Stresses and Fatigue Behavior / 49

8
With residual
stresses
Keff, MPa m

Without residual stresses

2
Range of Kth,eff
0
0

0.05

0.1

0.15

0.2

0.25

0.3

Distance from surface, mm


Range of the effective stress intensity factor
DKeff with and without consideration of the residual stresses versus distance from surface for ground
notched specimens made from a quenched-and-tempered
(600 C/2 h) SAE 1045 steel at cyclic bending loading (kt
1.7, g 2)

Fig. 70

Keff, MPa m

20

Without residual
stresses

15

Initial residual
stresses
Relaxed
residual
stresses

10

Range of Kth,eff

5
0
0

0.1

0.2

0.3

0.4

Distance from surface, mm


Range of the effective stress intensity factor
DKeff without and with consideration of the initial and the relaxed, respectively, residual stresses versus
distance from surface for ground and additional shot
peened notched specimens made from a quenched-andtempered (600 C/2 h) SAE 1045 steel at cyclic bending
loading (kt 1.7, g 2)

Fig. 71

Relaxed residual
stresses
Without residual
stresses

Keff, MPa m

16
12
8

Initial residual
stresses
4
Range of Kth,eff

0
0

0.1
0.2
0.3
0.4
Distance from surface, mm

0.5

of the effective stress intensity factor


Fig. 72 Rangewithout
and with consideration of the iniDKeff
tial and the relaxed, respectively, residual stresses versus
distance from surface for ground and additional shot
peened smooth specimens made from a quenched-andtempered (600 C/2 h) SAE 1045 steel at cyclic bending
loading (g 0.4)

between 300 and 200 lm below the surface and


finally crack arrest. However, in this stage of
crack propagation the crack extension will be
large compared to the depth of the compressive
residual stresses. Therefore, loading stress redistribution will occur, as will relaxation of the
compressive residual stress field in the process
zone ahead of the crack front. The crack will
continue to propagate toward the surface causing
final failure. Then, the fatigue limit is correlated
with the maximum stress amplitude that does not
cause crack initiation below the surface. Therefore, the improvement of the fatigue strength of
smooth specimens by shot peening is limited, as
shown in Fig. 45(a). However, Fig. 69 illustrates
that the overall value of DKeff and, hence, the
driving force for crack propagation is rather
small (compare, for example, Fig. 69 with Fig.
66 in view of the very different loading stress
amplitudes). Therefore, already in the transition
from infinite to finite fatigue lifewhere crack
initiation still occurs below the surface (see Fig.
63)there is a strong influence of shot peening
on fatigue life, as shown also in Fig. 45(a). The
influence of shot peening on finite fatigue life
increases with increasing stress amplitude, because the crack initiation site is shifted toward
the surface, as discussed in the paragraphs on
high-strength steel in the section Concept of
Local Fatigue Strength.
Application to a Medium-Strength Steel. In
a medium-strength SAE 1045 steel, the threshold
value Kth,eff ranges from 2.3 to 3.3 MPa m (Ref
67). In the following, ground notched specimens
(kt 1.7, g 2) with tensile residual stresses
(see Fig. 35, curve with open circles), are treated.
The interaction of this initial residual stress state
with the loading stress amplitude ra 143
MPawhich corresponds to the fatigue
strengthresults in the depth distribution of
DKeff shown in Fig. 70. At a depth of 15 lm,
DKeff exceeds the range of threshold values.
Therefore, a crack that initiates at the notch root
will always propagate into the interior of the
specimen, and the fatigue limit corresponds to
the maximum stress amplitude that does not ini-

Nominal stress amplitude (n,a), MPa

69. It becomes obvious that a crack, once initiated, can propagate. Regarding crack propagation toward the surface, the simple model predicts very slow crack growth in the range

1200

R m = 1650 MPa

Final failure
1000

Transition range
1400 MPa
No final
failure
1150 MPa

800

600
400
0

0.2

0.4

0.6

0.8

1.0

1.2

Length of arrested crack, mm


Influence of the nominal bending stress amplitude on the length of arrested cracks in deeprolled notched specimens (kt 2) made from SAE 5135
steel with the ultimate tensile strengths indicated. After Ref
66

Fig. 73

tiate a crack. Again, similar to the quenched state


(see Fig. 66), DKeff has a local maximum in a
depth of 70 lm, and up to a depth of 150 lm the
residual stress field increases significantly DKeff
as compared with the loading of a specimen
without residual stresses (dotted line).
The corresponding relationships for shot
peened notched specimens with the residual
stress distributions shown in Fig. 53 are given in
Fig. 71. The loading stress amplitude ra 310
MPa which corresponds to the bending fatigue
strength, is considered. From Fig. 60 it becomes
clear that crack initiation at the notch root occurs
under such loading conditions because the corresponding data point (open square) falls significantly above the Goodman relationship. On the
basis of the initial residual stress distribution,
there is no driving force for the propagation of
such a crack up to a distance of 170 lm from
the notch root. However, if the distribution of
the relaxed residual stresses is considered, DKeff
comes close to the threshold in depths ranging
from 30 to 200 lm. Therefore, the fatigue limit
corresponds to a loading condition where cracks
that initiate at the notch root may propagate or
not.
In the case of smooth shot peened specimens
with the residual stress distributions shown in
Fig. 52, it is difficult to determine on the basis
of the concept of the local fatigue strength
whether crack initiation occurs below or at the
surface, if the loading approaches the fatigue
limit (see paragraph on medium-strength steel in
the section Concept of Local Fatigue Strength
and Fig. 65b). Actually, crack initiation at the
surface is observed that is favored by the rather
high roughness. Figure 72 gives the depth distribution of DKeff for the loading stress amplitude
ra 414 MPa, which corresponds to the bending fatigue limit. Regarding the initial residual
stress state, the longest nonpropagating crack
would have a length of 250 lm. Due to stress
relaxation, this value is reduced by 50%. Still,
rather long cracks have to be formed during the
initiation stage to be able to propagate into the
interior of the specimen.
The difference between the minimum loading
stress amplitude for crack initiation at the root
of notched specimen and the maximum loading
stress amplitude at which crack arrest is observed may be very large in deep-rolled specimens, which have rather small or even tensile
residual stresses at the notch root, but large compressive residual stresses down to a relatively
great depth below the surface. Then, the length
of an arrested crack will strongly depend not
only on the depth distribution of the residual
stresses, but also on the loading stress amplitude.
This is proved by Fig. 73 concerning notched
specimens (kt 2) made from differently heat
treated SAE 5135 steel with the ultimate tensile
strength indicated (Ref 66). The diagram gives
the length of arrested cracks as a function of the
nominal stress amplitude. The higher the
strength of the material state, the less the crack
length increases with increasing stress amplitude
because the amount and the stability of the com-

50 / Effect of Materials and Processing


pressive residual stresses produced by deep rolling increase with increasing hardness. From this
figure it becomes clear that the large increase of
the notch fatigue strength visible in Fig. 41 is
almost entirely based on an increase of the resistance against crack propagation produced by
deep rolling.
In view of the relationships presented so far,
it is interesting to look again at Fig. 39 and 40.
The lowest bending fatigue limit was determined
with ground specimens having low tensile residual stresses at the surface (curve 1 in both figures). An increase of 324 MPa (curve 6 in Fig.
39) was achieved with specimens that were shot
peened according to curve 4 in Fig. 40 and from
which a surface layer of 100 lm thickness was

removed electrolytically, resulting in a shift of


the maximum compressive residual stresses into
the surface. Taking the slope of the straight line
for smooth specimens, g 0.4, in the range of
(initial) tensile residual stresses in Fig. 36 as a
rough approximation, one would expect a difference of both specimen series of 316 MPa
caused by the different residual stresses. On the
other hand, on the basis of the Goodman line in
Fig. 60, one would expect a much higher difference of 429 MPa. This shows that stress relaxation occurs in this medium-strength steel state,
as expected. Anyway, the bending fatigue limit
of both specimen series is determined by crack
initiation at the surface. The specimens corresponding to curve 5 in both figures have a lower

30
Without residual stresses
n,a = 640 MPa
n,a = 600 MPa

Keff, MPa m

Initial residual
stresses
20
Condition 6
n,a = 640 MPa

Condition 5
n,a = 600 MPa

Assumed
relaxed
residual
stresses

10

Range of Kth,eff

0
0

0.1

0.2
0.3
0.4
Distance from surface, mm

0.5

0.6

Range of the effective stress intensity factor DKeff without and with consideration of the initial and the relaxed,
respectively, residual stresses versus distance from surface for (corresponding to Fig. 39) shot peened smooth
specimens made from a blank-hardened AISI 5115 steel at cyclic bending loading

Fig. 74

rs

ls
Smooth

Fatigue strength, Rf

kt
Rh

fatigue limit, but higher finite fatigue lives than


the specimens corresponding to curve 6. Applying the same procedure with the aid of Fig. 36
as above to the surface residual stress state, an
increase of the bending fatigue limit of 220 MPa
compared to ground specimens is expected. Actually, a difference of 279 MPa is determined.
This finding means that the bending fatigue limit
is not determined by crack initiation at the surface, but by the largest amplitude at which crack
arrest below the surface occurs. These relationships become more clear in Fig. 74. There, DKeff
for surface cracks developing in conditions 5 and
6 at nominal stress amplitudes corresponding to
the bending fatigue strength are plotted as a
function of the distance from the surface. Regarding the initial residual stress distribution,
cracks smaller than 320 lm (condition 5) or 100
lm (condition 6) would not be able to propagate.
Actually, stress relaxation occurs, the extent of
which, however, is not given in Ref 54 to 56. If
it is assumed that the residual stresses relax by
50%, one gets the curves for the relaxed residual
stress state in the figure. Now, one can see that
a crack once initiated in condition 6 will continue propagation. Hence, the bending fatigue
strength of this condition is determined by the
resistance against crack initiation, as stated
above. Contrarily, in condition 5 the bending fatigue strength corresponds to the maximum
nominal stress amplitude at which crack arrest
below the surface occurs.
From these considerations, it becomes clear
that the depth distribution of relaxed residual
stresses must be taken into account to gain a reliable assessment of the influence of residual
stresses on the fatigue behavior of mediumstrength steel. Frequently, these data are not
known. Then, the use of initial tensile residual
stresses for the dimensioning of components is
a conservative procedure, because their detrimental effect on the fatigue behavior of a component will be reduced by residual stress relaxation. However, the beneficial effect of
compressive residual stresses will be overestimated by using the initial values, and this results
in a nonconservative dimensioning. Therefore, it
is important to estimate the remaining residual
stresses conservatively, for example with the article Stability of Residual Stresses in this
Handbook.

rs

micro

rs

ls

Summary and Recommendations


kt
Rh

Notched
Smooth
rs

micro

ls

Notched

Smooth

kt

Rh

Notched

Some important conclusions that can be


drawn from this article are illustrated by Fig. 75,
in which the fatigue strength Rf is plotted as a
function of the macro residual stress. The following parameters are regarded:
The material strength; a low-strength steel

Residual stress ,rs


Influence of the macro residual stress and some other parameters on the fatigue strength of smooth and notched
specimens made from a low-strength steel (lower band), a medium-strength steel (middle band), and a highstrength steel (upper band)

Fig. 75

(for example, a normalized steel, lower band


in the figure), a medium-strength steel (for
example, a quenched-and-tempered (at a medium temperature) steel, middle band in the
figure), and a high-strength steel (a quenched

Residual Stresses and Fatigue Behavior / 51

or quenched-and-tempered (at a low temperature) steel, upper band in the figure) are concerned
The depth distribution of the macro residual
stress rrs, characterized by its sign, magnitude, and gradient grs
The depth distribution of the micro residual
stress rrs
micro
The notch factor kt (smooth and mildly
notched specimens are regarded in the figure)
The gradient of the loading stress gls
The surface topography, characterized by the
roughness height Rh

Low-Strength Steel
In a low-strength steel, there will be no or very
little influence of the macro residual stress, because it is relaxed more or less completely if the
cyclic loading approaches the fatigue strength
(see Fig. 32 and 48). A change of the micro residual stress state by work hardening may significantly increase Rf since the resistance against
cyclic plastic deformation and hence, crack initiation increases (see Fig. 31). Then, also the resistance against macro residual stress relaxation
is raised resulting in a certain sensitivity of the
work-hardened zone to macro residual stress.
This may be detrimental or beneficial (see Fig.
33b) for Rf depending on the sign and the magnitude of the macro residual stress. The influence
of the surface topography is rather small in a
low-strength steel (see Fig. 47). The fatigue
notch factor kf is significantly smaller than the
notch factor kt because cyclic plastic deformation and stress redistribution occur in the notch
root (see Fig. 32). With increasing loading stress
gradient at a given kt, the fatigue strength increases due to the decrease of the highly stressed
volume of the component or specimen (compare
kt 2.5, g 2 with kt 2.5, g 5 in Fig.
32).

Medium-Strength Steel
In a medium-strength steel, there is a significant influence of the macro residual stress on Rf
since only a small part of rrs relaxes during cyclic loading in the range of the fatigue limit (see
Fig. 36, 38, 51a, and 52). However, in the lowcycle fatigue range, relaxation becomes more
complete with increasing amplitude, and the influence of the macro residual stress vanishes (see
Fig. 34, 37, 39, 51b). Tensile residual stresses
are always detrimental to Rf. Therefore, in the
presence of large tensile residual stresses a medium-strength steel may have equal or even
lower fatigue strength than a low-strength steel
(compare Fig. 32 with 36). If relaxed tensile residual stresses are concerned, the residual stress
sensitivity m of the fatigue strength approaches
the mean stress sensitivity M (see Fig. 60). In
material states of similar hardness loaded in the
range of the fatigue limit, compressive residual
stresses relax stronger than tensile ones simply
because of the different corresponding stress am-

plitudes. However, the micro residual stress state


may be changed significantly by the processes
that generate the macro residual stress. Hence,
in a medium-strength steel the local resistance to
residual stress relaxation may be quite different
even if the same initial material state is concerned, and unique borderlines for the onset of
residual stress relaxation do not exist (see Fig.
60). A significant benefit from compressive residual stresses is only obtained if their penetration depth is sufficiently high and/or their gradient is sufficiently low as compared to the
gradient of the loading stresses. Therefore, compressive residual stresses produced by grinding
do hardly influence the fatigue strength of
smooth and notched specimens (see Fig. 36).
Contrarily, compressive residual stresses generated by deep rolling or shot peening increase the
notch fatigue strength (see Fig. 38, 39, and 41).
The fatigue strength of smooth specimens is less
improved, if subsurface crack initiation occurs.
Then, the fatigue strength of notched specimens
may even be higher than corresponding values
of smooth specimens (see Fig. 41). This means
that the fatigue notch factor kf becomes less than
unity. At vanishing residual stresses, kf comes
rather close to kt if the loading stress gradient is
small, but is significantly less at higher gls values. With increasing tensile residual stresses, the
fatigue notch factor is reduced again, as expected
on basis of the Goodman relationship for smooth
and notched specimens (see Fig. 58 and 60).
There is a considerable influence of the surface
roughness on the fatigue strength, as shown in
Fig. 47.

High-Strength Steel
In high-strength steels, stress relaxation during cyclic loading in the range of the fatigue
strength only occurs in notched specimens bearing very high compressive residual stresses.
Then, the resulting fatigue strength is also high,
and during corresponding cyclic loading very
high magnitudes of the minimum stress occur,
which leads to some residual stress relaxation
(see Fig. 57b). Contrarily, in the range of high
tensile residual stresses and cyclic loadings that
lead to infinite life or to technically relevant lifetimes the occurring maximum stresses are much
lower and no residual stress relaxation is observed even in the range of low-cycle fatigue.
Consequently, the residual stress sensitivity and
the mean stress sensitivity of Rf are identical (see
Fig. 59), and the fatigue strength is strongly reduced with increasing tensile residual stress.
This is also true for the finite fatigue life (see
Fig. 42). In the range of compressive residual
stresses, complex relationships exist. A strong
effect of rrs will only occur if cracks are initiated
at the surface. However, in thick smooth specimens or components cyclically loaded in the
range of the fatigue strength, the loading stress
gradient is usually lower than the gradient of the
local fatigue strength, which depends on the
depth distribution of the residual stresses. Consequently, cracks are initiated below the surface

(see Fig. 62a and c, 63, and 64c) and the improvement of the fatigue strength even by high
compressive residual stresses is limited (see Fig.
45a, 46, and 56). With increasing stress amplitude and hence loading stress gradient, the crack
initiation site is shifted to the surface. Therefore,
the finite fatigue life is much more improved
than the fatigue strength (see Fig. 62a and c, 63,
and 45d). In notched specimens, the loading
stress gradient is larger than in smooth ones. If
high compressive residual stresses with a sufficient depth and/or a small gradient exist, crack
initiation will occur in the notch root resulting
in a strong improvement of the fatigue strength
(see Fig. 64a and b, 45b, 46, and 56). However,
if the penetration depth of the compressive residual stresses is low and therefore the gradient
of the local fatigue strength highfor example,
after grindingthe improvement of the notch
fatigue strength will be small (see Fig. 44 and
59). Frequently, after shot peening or deep rolling, maximum compressive residual stresses occur below the notch root. Then, the fatigue
strength may not be determined by the maximum
cyclic loading, which does not result in crack
initiation, but by the maximum cyclic loading at
which crack arrest below the surface is possible
(see Fig. 67, 68, and 73). As a consequence of
all these relationships, the fatigue notch factor
may vary strongly in the presence of compressive residual stresses regarding one-notch geometry and may take values ranging from less
than unity to the notch factor kt, as sketched in
Fig. 75. In fact, regarding smooth and notched
specimens, Fig. 44 and 54 prove that kf of ground
specimens with compressive residual stresses
produced by grinding approaches kt. From Fig.
56 it can be deduced that kf of shot peened specimens approaches unity. At vanishing residual
stresses, kf comes close to kt as expected in a
high-strength steel. With increasing tensile residual stress, the fatigue notch factor is reduced
and finally approaches unity (see Fig. 44 and 54).
Again, this finding is in correspondence with the
Goodman relationship for smooth and notched
specimens (see Fig. 58 and 59). The surface
roughness has principally a large influence on
the fatigue strength of high-strength steel, as
shown in Fig. 47. On the other hand, in practice
the roughness height of hard steel is rather low
even after mechanical surface treatments such as
shot peening.

Recommendations
From all of these relationships, some recommendations may be deduced. In medium- and
high-strength steels tensile macro residual
stresses must strictly be avoided since they always promote crack initiation and crack propagation and are detrimental to the fatigue strength
andat least in higher strength steelto finite
fatigue life. In a low-strength steel, the influence
of tensile macro residual stresses is usually small
or negligible, and the change of the micro residual stress state by work hardening is much more
important. In most cases, work hardening will be

52 / Effect of Materials and Processing


beneficial for fatigue strength andif the micro
residual stress state is sufficiently stablefor finite fatigue life. However, processes must be
avoided in which strong work hardening is combined with the generation of tensile residual
stresses. Then, residual stress relaxation during
cyclic loading will be incomplete, and the detrimental influence of tensile residual stresses on
the fatigue strength will appear.
The beneficial influence of compressive
macro residual stresses ranges from almost nil to
the full extent expected after the mean stress sensitivity of the material state under consideration
and depends on the wholeness of the parameter
considered in Fig. 75. In a low-strength steel, it
is recommended to use processes in which the
generation of compressive residual stresses is
combined with strong work hardening. Then,
there will be on one hand the beneficial influence
of work hardening on the reduction of cyclic
plasticity and the increase of the materials resistance against crack initiation. On the other hand,
the resistance against residual stress relaxation
will also increase, resulting in some beneficial
influence of the compressive macro residual
stresses on crack initiation and crack propagation. One example is deep rolling (see Fig. 33).
Another example would be shot peening. However, shot peening of a low-strength steel will
produce a rather rough surface, and this will
eventually counterbalance a large part of the
beneficial influence on the fatigue behavior.
In medium-strength steels it is also recommended to utilize processes that produce both
work hardening and compressive macro residual
stresses. Again, deep rolling (see Fig. 41) or shot
peening (see Fig. 37 and 39) can serve as examples. Now, for a significant effect it is necessary to adjust the penetration depth of the
macro residual stresses and the width of the
work-hardened zone to the depth distribution of
the loading stress. Clearly, this is impossible in
push-pull loading of smooth specimens or components (gls 0) and very difficult in any cyclic
loading of thick smooth specimens or components, where gls is rather small. Then, some improvement of the fatigue behavior will be obtained because crack initiation at the surface is
retarded or crack initiation is shifted below the
surface (see Fig. 37a and Fig. 41, smooth specimens). However, in notched or rather thin
smooth specimens or components with higher gls
it is possible and recommended to produce a
depth distribution of the macro and micro residual stresses and, hence, of the local fatigue
strength that ensures that crack initiation can
only occur at the surface. Then, a strong improvement of the fatigue strength and finite fatigue life is possible by one of two mechanisms.
On one hand, the resistance of the steel against
crack initiation and growth of very short cracks
may strongly be increased by a smooth surface,
strong work hardening, and maximum compressive residual stresses at the surface. Examples
are given in Fig. 37(b) (shot peened notched
specimens) and Fig. 39 (condition 6, shot peened
smooth specimens). In these cases, the fatigue

strength is determined by the maximum loading


stress amplitude that does not initiate a crack at
the notch root. On the other hand, by high compressive stresses below the surface, the resistance against propagation of cracks with lengths
up to the distance of the residual stress maximum
from the surface may strongly be increased.
Again, strong work hardening, is essential to prevent the compressive residual stresses as much
as possible from relaxation. Examples are given
in Fig. 39 and 40 (condition 5, shot peened
smooth specimens) and Fig. 41 and 73 (deeprolled notched specimens). In these cases, the
fatigue strength is determined by the maximum
loading stress amplitude at which crack arrest
below the surface is obtained.
In high-strength steels, residual stress relaxation is negligible except for extreme conditions
that were discussed above (see Fig. 57b). Apart
from this item, principally the same relationships
as just discussed are valid. Again, it is difficult
to achieve a large improvement of Rf in thick
smooth specimens or components, since crack
initiation will occur below the surface (see Fig.
56, 63, and 64c). However, a significant improvement of finite fatigue life in the low-cycle
fatigue range may be obtained, since crack propagation toward the surface is hindered by the
high compressive residual stresses (see Fig. 69)
and/or the crack initiation site is shifted to the
surface as a result of the increasing loading stress
gradient (see Fig. 63). In thin smooth or in
notched specimens and components it is recommended to produce high compressive residual
stresses with a large penetration depth by processes such as shot peening, deep rolling, and
others described in this Handbook. Then, crack
initiation and early crack growth is forced to occur at the surface, and enormous increases of the
fatigue strength are obtained by one of the two
mechanisms described previously. However, by
application of the aforementioned processes to a
high-strength steel, frequently rather small surface compressive residual stresses are obtained
and maximum values of |rrs| occur below the
surface. Then, the second mechanism (crack arrest below the surface) is solely effective (see
Fig. 67 and 68), and the fatigue strength may be
much higher than expected by the surface residual stress and loading state plotted in a Haigh
diagram (see Fig. 59a).
Regarding surface roughness, numerous investigations exist. The interaction of an increased roughness with tensile residual stresses
will always be detrimental to the fatigue strength
and to finite fatigue life, especially in highstrength steels. In the presence of compressive
residual stresses, different influences exist. If the
fatigue strength is entirely and the finite fatigue
life mainly determined by crack initiation at the
surface, there will be decisive influence of the
surface roughness. A much smaller influence of
surface roughness, however, may exist if the fatigue strength is determined by crack initiation
or crack arrest below the surface or if the finite
fatigue life is mainly influenced by crack propagation below the surface. If compressive resid-

ual stresses are utilized in the dimensioning of


components, it is of utmost importance to account for their stability. The reader is referred to
the article Stability of Residual Stresses in this
Handbook.
REFERENCES
1. B. Scholtes, Residual Stresses Generated by
Mechanical Surface Treatments, DGM Informationsgesellschaft, Oberursel, 1990, in
German
2. V. Hauk, Ed., Structural and Residual
Stress Analysis by Nondestructive Methods:
Evaluation, Application, Assessment, Elsevier Science B.V., Amsterdam, 1997
3. H. Ishigami, M. Matsui, Y. Jin, and K.
Ando, Proc. ICRS-6, IOM Communications
Ltd., London, 2000, p 667673
4. Fatigue and Fracture, Vol 19, ASM Handbook, ASM International, 1996
5. E. Macherauch, Laboratory Course on Material Science and Engineering, ViewegVerlag Braunschweig, Germany, 1992, in
German
6. D. Eifler, Dr.-Ing. Thesis, University Karlsruhe (TH), 1981
7. D. Eifler and E. Macherauch, Defects, Fracture and Fatigue, G.C. Sih and J.W. Prowan, Ed., Martinus Nijhoff, The Hague,
1983, p 171182
8. D. Eifler and E. Macherauch, Int. J. Fatigue,
Vol 12 (No. 3), 1990, p 165174
9. M. Becker, Dr.-Ing. Thesis, University
Karlsruhe (TH), 1987
10. M. Becker, D. Eifler, and E. Macherauch,
Mater.wiss. Werkst.tech., Vol 24, 1993, p
5764
11. P.C. Paris, M.P. Gomez, and W.E. Anderson, The Trend in Engineering, Vol 13,
1961, p 914
12. P.C. Paris and F. Erdogan, J. Basic Eng.
(Trans. ASME ), Vol 85, 1963, p 528534
13. NASA/FLAGRO: Fatigue Crack Growth
Computer Program, National Aeronautics
& Space Administration, p 5
14. R.C. McClung, K.S. Chan, S.J. Hudak, Jr.,
and D.L. Davidson, Behavior of Small Fatigue Cracks, Fatigue and Fracture, Vol 19,
ASM Handbook, ASM International, 1996,
p 153158
15. A. Wohler, Z. Bauwesen, Vol XX, 1870, p
73106
16. O.H. Basquin, Proc. ASTM, Vol 10, 1910,
p 625630
17. L.F. Coffin, Trans. ASME, Vol 76, 1954, p
931950
18. S.S. Manson, NACA Report 1170, Lewis
Flight Propulsion Laboratory, 1954
19. J.H. Smith, J. Iron Steel Inst., Vol 82, 1910,
p 246318
20. B.P. Haigh, No. 85, Report Brit. Association, 1915, p 163170
21. J. Goodman, Mechanics Applied to Engineering, Longmans-Green, London, 1899
22. H. Gerber, Z. Bayr. Architekt. Ing.-Ver., Vol
6, 1874, p 101110

Residual Stresses and Fatigue Behavior / 53


23. K.N. Smith, P. Watson, and T.H. Topper, J.
Mater., Vol 5, 1970, p 767778
24. W.J. Ostergren, J. Test. Eval., Vol 4 (No.
5), 1976, p 327339
25. A. Ebenau, Dr.-Ing. Thesis, University
Karlsruhe (TH), 1989
26. A. Ebenau, D. Eifler, O. Voehringer, and E.
Macherauch, Proc. ICSP 4, K. Iida, Ed., The
Japan Society of Precision Engineering, Tokyo, 1990, p 327336
27. F. Hoffmann, Dr.-Ing. Thesis, University Bremen, 1987; Fortschritts-Berichte VDI Reihe 5
Nr. 123, VDI-Verlag, Dusseldorf, 1987
28. I. Altenberger, Dr.-Ing. Thesis, University
GH Kassel, 2000
29. I. Altenberger, B. Scholtes, U. Martin, and
H. Oettel, Mater. Sci. Eng. A, Vol 264,
1999, p 116
30. G. Kuhn, Dr.-Ing. Thesis, University Karlsruhe (TH), 1991
31. G. Kuhn, J.E. Hoffmann, D. Eifler, B.
Scholtes, and E. Macherauch, ICRS 3, H.
Fujiwara, T. Abe, and K. Tanaka, Ed., Elsevier Applied Science, London, 1991, p
12941301
32. A. Glaser, D. Eifler, and E. Macherauch,
Mater.wiss. Werkst.tech., Vol 22, 1991, p
266274
33. H. Mughrabi, Fatigue Behavior of Metallic
Materials, D. Munz, Ed., DGM-Informationsgesellschaft Verlag, Oberursel, 1985, p
738, in German
34. J.D. Almer, J.B. Cohen, and B. Moran, Mater. Sci. Eng. A, Vol 284, 2000, p 268279
35. H. Berns and L. Weber, Residual Stresses
in Science and Technology, E. Macherauch
and V. Hauk, Ed., DGM-Informationsgesellschaft Verlag, Oberursel, 1987, p 751
758
36. H. Berns and L. Weber, Shot Peening, ICSP
3, H. Wohlfahrt, R. Kopp, and O. Voehringer, Ed., DGM-Informationsgesellschaft
Verlag, Oberursel, 1987, p 647654
37. R. Herzog, Dr.-Ing. Thesis, University
Braunschweig, 1997
38. E. Macherauch and H. Wohlfahrt, Fatigue
Behavior of Metallic Materials, D. Munz,

39.
40.
41.
42.

43.
44.
45.
46.

47.

48.

49.
50.
51.

Ed., DGM-Informationsgesellschaft Verlag,


Oberursel, 1985, p 237283, in German
P. Starker, E. Macherauch, and H. Wohlfahrt, Fatigue Eng. Mater. Struct., Vol 1,
1987, p 319327
S. Jaegg, Dr.-Ing. Thesis, University GH
Kassel, 1999
S. Jaegg and B. Scholtes, ICRS 5, Linkoping, Sweden, 1997, p 10781083
E. Welsch, D. Eifler, B. Scholtes, and E.
Macherauch, Proc. Sixth European Conf. on
Fracture (ECF 6), V. Elst and A. Bakker,
Ed., Eng. Mater. Adv. Services Ltd
(EMAS), Amsterdam 1986, p 13031320
A. Stacey and G.A. Webster, Materials Research Society Proc., Vol 22, 1984, p 215
H.P. Lieurade, Advances in Surface Treatments, A. Niku-Lari, Ed., Pergamon Press,
Oxford, 1987, p 455482
K.J. Kang, J.H. Song, and Y.Y. Earmme,
Fatigue Fract. Eng. Mater. Struct., Vol 12,
1989, p 363376
J.E. Hoffmann, D. Loehe, and E. Macherauch, Proc. ICSP3, H. Wohlfahrt, R. Kopp,
and O. Voehringer, Ed., Garmisch-Partenkirchen, DGM Verlag, Oberursel, Germany,
1987, p 631638
J.E. Hoffmann, D. Loehe, and E. Macherauch, Residual Stresses in Science and
Technology, Proc. ICRS 1, E. Macherauch
and V. Hauck, Ed., Vol 1, DGM Verlag, Oberursel, Germany, 1987, p 801808
J.E. Hoffmann, D. Eifler, and E. Macherauch, Residual Stresses: Origination, Measurement, Assessment, E. Macherauch and
V. Hauck, Ed., Vol 2, DGM Verlag, Oberursel, Germany, 1983, p 287300, in German
J.E. Hoffmann, Dr.-Ing. Thesis, University
Karlsruhe (TH), 1984
B. Syren, H. Wohlfahrt, and E. Macherauch,
Arch. Eisenhuttenwes., Vol 46, 1975, p
735739
B. Syren, H. Wohlfahrt, and E. Macherauch,
Proc. Second Int. Conf. Mech. Behavior of
Materials (ICM2) (Boston), 1976, American Society for Metals, p 807811

52. B. Syren, Dr.-Ing. Thesis, University Karlsruhe (TH), 1975


53. H. Traiser and K.-H. Kloos, Z. Werkstofftechn., Vol 16, 1985, p 135143
54. R. Schreiber, H. Wohlfahrt, and E. Macherauch, Arch. Eisenhuttenwes., Vol 49,
1978, p 207210
55. R. Schreiber, H. Wohlfahrt, and E. Macherauch, Arch. Eisenhuttenwes., Vol 48,
1977, p 653657
56. H. Wohlfahrt, Mechanical Surface Treatments, H. Wohlfahrt and P. Krull, Ed., Wiley-VCH, Weinheim, Germany, 2000, p
5584, in German
57. K.-H. Kloos, B. Fuchsbauer, and J. Adelmann, J. Fatigue, Vol 9, 1987, p 3542
58. P. Starker and E. Macherauch, Z. Werkstofftechn., Vol 14, 1983, p 109115
59. P. Starker, Dr.-Ing. Thesis, University
Karlsruhe (TH), 1981
60. A. Navarro and E.R. de los Rios, Philos.
Mag. A, Vol 57 (No. 1), 1988, p 1536
61. K.J. Miller and E.R. de los Rios, The Behavior of Short Fatigue Cracks, Mechanical
Engineering Publications Limited, London,
UK, 1986
62. B. Scholtes, Structural and Residual Stress
Analysis by Nondestructive Methods: Evaluation, Application, Assessment, V. Hauk,
Ed., Elsevier Science B.V., Amsterdam,
1997
63. W. Backfisch and E. Macherauch, Arch. Eisenhuttenwes., Vol 50, 1979, p 167171
64. K. Wellinger and D. Dietmann, Calculation
of Strength, A. Kroener Verlag, Stuttgart,
1969, in German
65. B. Winderlich, Mater.wiss. Werkst.tech.,
Vol 21, 1990, p 378389
66. K.H. Kloos, Notches and Structural Durability, H. Nowack, Ed., Deutscher Verband
fur Materialforschung und-pruefung e.V.,
1989, p 740, in German
67. J. E. Hoffmann and D. Loehe: to be published
68. T. Fett and D. Munz, Stress Intensity Factors and Weight Functions, Computational
Mechanics Publications, 1997

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p54-69
DOI: 10.1361/hrsd2002p054

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Stability of Residual Stresses


D. Lohe and O. Vohringer, Institut fur Werkstoffkunde I, Universitat Karlsruhe, Germany

RESIDUAL STRESSES are generated in


structural components during manufacturing
processes such as forging, machining, heat treating, shot peening, and many others. These
stresses are always a consequence of inhomogeneously distributed dimensional changes due
to inhomogeneous plastic deformations, thermochemical treatments, and/or phase transformations. They can be either beneficial or detrimental to component behavior in service,
depending on the materials state as well as the
sign, magnitude, and stability of the residual
stresses, mechanical loading, and environmental
conditions (Ref 115). Compressive macro residual stresses in the surface region of materials
with medium and high hardnesses increase the
fatigue life and the fatigue limit at cyclic loading
compared to materials states that are free of residual stresses. This improvement is caused by
an increased resistance against crack initiation
and, to a certain extent, against crack propagation if the residual stresses are sufficiently stable
in the areas of highest loading (in most cases, the
component surface areas). Moreover, compressive macro residual stresses can increase the resistance of certain materials against corrosion fatigue and stress corrosion cracking. They can
also improve the wear resistance. Therefore, in
many cases compressive macro residual stresses
are intentionally generated in near-surface regions by controlled heat treatments or by posttreatments such as shot peening or deep rolling.
The stability or relaxation behavior of these
residual stresses at purely thermal or mechanical
loadings as well as superimposed thermal and
mechanical loadings is thus of decisive importance for the service behavior of components and
hence of great interest from a scientific as well
as a practical point of view (Ref 1618). If residual stresses are relaxed by annealing or mechanical treatment, they naturally have little if
any influence on subsequent component failure.
Residual stresses can be reduced or completely relaxed by the application of mechanical
and/or thermal energy. The elastic residual
strains, ee, that are associated with the residual
stresses via Hookes law can be converted into
micro plastic strains, ep, by suitable deformation
processes. For example, this transformation can
be achieved by dislocation slip, dislocation
creep, grain-boundary sliding, and/or diffusion

creep (Ref 1617). If these processes occur to


only a limited extent or not at all, relaxation of
residual stresses is also conceivable by crack formation and propagation. Relaxation of residual
stresses in the real case occurs by complex interaction of a large number of factors. It depends
not only on the residual stress state itself but also
on the material state, loading condition, geometry, and environment of the component under
consideration.
The best known and most important techniques for inducing residual stress relaxation are
annealing (tempering), uniaxial deformation
(drawing, stretching), and cyclic deformation
(Ref 1618). Relaxation can also be caused by
thermal cycling, quenching, neutron bombardment, the effect of alternating magnetic fields (in
the use of ferromagnetics), vibration, and partial
damage. As discussed in other articles in this
Handbook, residual stresses can be differentiated
by their technological origins (i.e., whether they
were produced during forming, machining, heat
treatment, joining, coating, or casting).
Before the effect of applied thermal or mechanical energy on the residual stress state can
be assessed, the latter must be determined quantitatively. In the case of mechanical methods, the
necessarily large number of measurements involve considerable expenditure of both time and
money. Most investigations have thus employed
x-ray methodsfor example, the sin2w method
(Ref 19) with CrK radiation on the {211}-interference plane of the ferrite, as well as
interference-profile analyses (e.g., Ref 20, 21).
In this article, current knowledge about residual stress relaxation of steels is presented in a
condensed and, as far as possible, systematic
form. The abundance of available experimental
data nevertheless necessitates a limited scope.
Thus, the findings of residual stress relaxation
by annealing, tensile loading, compressive loading, and cyclic loading will be taken as examples
to demonstrate universally valid rules. Shot
peening residual stresses are particularly suitable
for such investigations due to their sign and their
relatively large absolute value. Possible formulas
that can be used to quantify residual stress relaxation are discussed, and the underlying microstructural as well as micromechanical mechanisms are considered.

Relaxation of Residual Stresses by


Annealing
If a pure metal is annealed for several hours
at a temperature of
Ta  0.5 Tm [K]

(Eq 1)

where Tm is the melting or solidus temperature,


and then cooled slowly to room temperature, almost total relaxation of the residual stresses arising from forming, machining, heat treatment, or
joining operations can be achieved. The necessary annealing time depends essentially on the
workpiece dimensions and the material state.
Since an annealing temperature of 0.5 Tm corresponds to the recrystallization temperature of
ferritic steels, complete relaxation of macro residual stresses can be expected. In the case of
pearlitic steels, macro residual stresses are completely eliminated during the pearlite-austenite
transformation above the A1 temperature. Micro
residual stresses are considerably reduced but
not entirely removed, since lattice defects (particularly dislocations) and, in the case of heterogeneous materials, the different expansion coefficients of the various phases are always
responsible for some micro residual stresses. Residual stress relaxation by annealing is brought
about by so-called thermally activated processes
for which the annealing temperature and the annealing time are interchangeable within certain
limits. In order to achieve comparable residual
stress relaxation at a lower annealing temperature, the annealing time must be increased correspondingly (Ref 16, 17).
Thermal residual stress relaxation is fundamentally affected by the residual stress state itself and by the material state. This is convincingly demonstrated by the findings presented in
Fig. l and 2 (Ref 16, 17), which show the effect
of 1 h anneals on surface macro residual stresses
of a variety of origins in a variety of steels investigated using x-ray methods. Figure 1 is a bar
chart showing the macro residual stresses present
in the original state and after the normal industrial stress relieving for hardened components of
1 h at 200 C (390 F). The materials deformed
in tension have undergone negligible residual
stress relaxation. The greater the original compressive residual stresses, the greater the residual

Stability of Residual Stresses / 55


stress relaxation produced by annealing of
surface-machined specimens: 11% for a plain
carbon steel with 0.45 wt% Carbon, SAE 1045
(Fig. 1C); 17% for SAE 1045 (Fig. 1D); and
19% for the rolling bearing steel AISI 52100
(Fig. 1E). The most pronounced relative residual
stress relaxation, however, is observed in the
hardened materials: 25% for plain carbon steel
with 0.22 wt% C, SAE 1023 (Fig. 1F); and 37%
for SAE 1045 (Fig. 1G).
Figure 2 shows the relative residual stresses
rrs(Ta)/rrs(293 K) for the same steels as a function of the homologous annealing temperature,
Ta /Tm. It can be seen that a 1 h anneal at 0.5 Tm
(about 600 C, or 1110 F) results in complete
relaxation of the macro residual stresses in every
case. Characteristic, material-specific stresstemperature curves are obtained. Clearly, residual stresses produced by hardening are relaxed
at lower temperatures, machining residual
stresses at medium temperatures, and deformation residual stresses at higher temperatures. In
order to achieve the same degree of residual
stress relaxation, (e.g., 50%) in 1 h in hardened
and in deformed SAE 1045, a temperature difference of 150 C (270 F) is necessary. Furthermore, according to Fig. 1, larger initial residual stresses result in a shift of the rrs versus
T curve to lower annealing temperatures in the
case of hardened or machined materials.
Micro residual stresses and their relaxation behavior can be determined directly from the halfwidth values (HW) of the x-ray interference
lines, which are a measure of the microstructural
work-hardening state as well as the dislocation
density. The change of the HW values are approximately proportional to the changes of the
micro residual stresses (DHW  Drrs
micro).
Again, the micro residual stresses in steels are
more effectively relaxed at a higher temperature
and, like macro residual stresses (see Fig. 2),

hardening micro residual stresses are relaxed at


lower temperatures, such as those produced by
machining; likewise, machining micro residual
stresses are relaxed at lower temperatures, such
as those produced by deformation in tension. In
comparable steel states, micro residual stresses
are relaxed only after a longer period or at a
higher temperature than macro residual stresses
(Ref 16, 17). For example, in order to achieve
50% micro residual stress relaxation in tensiledeformed SAE 1045 annealed for 1 h, the annealing temperature must be 100 C (180 F)
higher than that needed to achieve the same degree of macro residual stress relaxation.
Data pertaining to the thermal relaxation of
micro residual stresses for hardened steels are
available by analyzing x-ray interference lines of
different steels in (Ref 2224). The mean lattice
distortions e21 / 2, which are proportional to the
micro residual stresses according to
2 1/2
rrs
micro E110 e 

(Eq 2)

E110 is the Youngs modulus of iron in the 110


direction, increase with growing carbon content
mainly due to the increasing dislocation density
and, secondarily, to the increasing number of
solute carbon atoms in octahedral sites of the
tetragonal martensite lattice. Usually, it can be
assumed that with increasing carbon content the
temperature for the onset of residual stress relaxation is lowered and the recovery rate increases.
However, it must be remembered that with the
annealing-induced formation of carbides on dislocations, which leads to their pinning, the relaxation procedure slows.
The relaxation behavior of residual stresses
depends for a similar surface treatment, as for
example by shot peening, as shown in Fig. 3 for
the plain carbon steel SAE 1045 with 0.45 wt%
C, from the heat treatment before shot peening

400
Ck22
F
Ck45
A

Ck22
B

Ck45
C

Ck45
D

Ck45
G

and thus, from the hardness of the material (Ref


25). The macro residual stresses (left-hand side
of Fig. 3) and the micro residual stresses (righthand side; characterized by the half-width ratio
DHW/DHW0), yield with the hardness increasing
from the normalized to the quenched and tempered and to the hardened condition, greater relaxation rates. Figure 3 also illustrates the delay
of the relaxation of the micro residual stresses,
which was mentioned above in comparison with
the macro residual stresses. Further experimental
investigations of the macro and micro residual
stress relaxation of metallic materialstempering steel AISI 4140 (Ref 2628), titanium alloy
Ti-6Al-4V (Ref 29), and several CuZn alloys and
AlMg alloys (Ref 30, 31)lead to similar results.
Materials with surface residual stresses produced by grinding, milling, deep rolling, or shot
peening frequently show different residual stress
relaxation behavior when comparing the surface
layer with subsurface areas. In most cases, this
behavior is connected with work-hardening gradients, where an increasing work hardening due
to increasing dislocation densities causes growing mean strains and greater half-width values.
Characteristic examples of depth distributions
of the macro residual stresses and the half-width
values in SAE 1045, which were produced after
different heat treatments and electrolytical removal of thin material layers, are shown Fig. 4
and 5 (Ref 25). Normalized SAE 1045 samples
in the ground, milled, or shot-peened condition
result in typical depth distributions that depend
on the heat treatment (Fig. 4). In this case, the
relaxation behavior of the macro residual
stresses (left-hand side), and of the half-widths
(right-hand side), is more pronounced at the surface than in deeper layers. After selected annealing treatments, this behavior holds to completely relaxed shot peening residual stresses in
the near-surface layers in contrast to the interior,
where remaining residual stresses exist (see Fig.
4, bottom row, left, after annealing 30 min at 550
C, or 1020 F).
The relaxation behavior of normalized SAE
1045 in the shot-peened condition is reproduced

100Cr6
E

Ta, C

400

Normalized
+5% deformed
in tension
Normalized
+ ground

800
Untempered
1200

Fig. 1

200 C/1 h tempered

1.0
rs(Ta)/rs(293 K)

rs, MPa

Hardened

Hardened +
shot peened

Macro residual stress (rrs) of different steels before and after annealing for 1 h at 200 C (390 F). Ck 45 SAE
1045; Ck 22 SAE 1023; 100 Cr 6 AISI 52100

100

200

300

400

C
B

0.6

0.2
0
0.15 0.20

600

ta = 1 h

0.8

0.4

500

E
F
G

0.25 0.30 0.35

0.40 0.45 0.50

Ta/Tm, K
Macro residual stress ratio versus homologous
annealing temperature of different steels (see
Fig. 1) after annealing for 1 h

Fig. 2

56 / Effect of Materials and Processing


again in Fig. 5 (top row) in comparison with
shot-peened conditions of quenched and tempered as well as hardened states of SAE 1045.
The shot peening residual stresses in the hardened state (bottom row) are reduced by annealing of 1 h at 300 C (570 F) almost 50% at the
surface and in the depth of the maximum of compressive residual stresses. The corresponding
HW values in the area close to the surface are
smaller than those measured in deeper regions.
This finding is well known from relatively hard
materials, such as hardened steels (500 HV)
with relatively high dislocation densities, where
the shot peening process causes decreasing HW
values due to work-softening effects (Ref 32).
The higher the hardness, the more effective the
work softening. In these material states, decreases in HW values are caused by rearrangements of the dislocations with high density in
low-energy structures. However, in the
quenched and tempered condition, and in the
normalized condition, enlarged relaxation effects are observed in the surface layer due to a
dislocation-induced higher driving force and the
relaxation rate decreases with increasing distance from the surface. These findings must be

taken into account for the valuation of residual


stress relaxation of corresponding components.

can be linearized with the help of a Zener-WertAvrami function (Ref 33) to:
m
rrs/rrs
0 exp [(Ata) ]

Thermally Activated Relaxation


Processes
In order to analyze the thermally activated
processes responsible for residual stress relaxation, the influence of annealing time, ta, and of
annealing temperature, Ta, must be known. If the
time and temperature ranges are sufficiently restricted and if rrs(Ta,ta)/rrs(293 K) constant
(abbreviated in the following to rrs/rrs
0 ), the relationship
ta t0 exp (DH/kTa)

(Eq 3)

should hold between annealing time and temperature. In Eq 3, t0 is a time constant, k is the
Boltzmann constant, and DH is the activation enthalpy for the relaxation process in the ta versus
Ta range under consideration. Data in the form
of rrs versus 1og ta plots as in Fig. 3 are not
suited to a direct evaluation using Eq 3.
The effect of time on residual stress relaxation

where m is a numerical term dependent on the


dominant relaxation mechanism and A is a function dependent on material and temperature according to
A B exp (DH/kTa)

Ta, C

Ta, C

200
250
300
350
400
450
500
550

200
250
300

0.5

350
400
450
500
550

0.5

Normalized
0

1.0

150
150

1.0

250

250

0.5

350

350
450

1.0

Quenched and tempered

0.5

HW/HW0

rs, (Ta , t a )/rs(293 K)

1.0

450

1.0

150

150
250

250

0.5

0.5
350

350
450

0
0

10

100

1000

450

Hardened

0
0

10

100

1000

Anealing time, ta, min


Macro residual stress ratio versus annealing time, and ratio of the changes of the half-width values DHW/DHW0
versus annealing time, at different annealing temperatures at the surface of shot-peened SAE 1045 samples in
normalized, quenched and tempered, and hardened states

Fig. 3

(Eq 5)

where B is a constant. It follows from Eq 4 that


rs
log ln (rrs
0 /r ) m log ta m log A

(Eq 6)

with the result that for a constant annealing temperature Ta, a plot of log ln(r0rs/rrs) versus log ta
gives a straight line (see Fig. 6).
Corresponding annealing temperatures and
times for rrs/r0rs constant can be derived from
such plots, as schematically illustrated in Fig. 6
in the log ta versus l/kTa plot. The values for
rrs/rrs0 constant lie on a straight line. Based
on Eq 5 and 6, this line results in
log ta constant

1.0

(Eq 4)

DH
ln 10 kTa

(Eq 7)

which can be rearranged to give Eq 3. The slope


of the log ta versus l/kTa plot in Fig. 6 yields to
the activation enthalpy for residual stress relaxation.
An experimental verification of Eq 6 is presented in Fig. 7, with results for shot-peened turbine blades made from a 12% Cr steel that was
annealed for residual stress relaxation in the temperature range from 150 to 600 C (300 to 1110
F) up to 104 h (Ref 34, 35). The measuring data
are fitted very well by straight lines with a slope
of approximately m 0.10. The evaluation of
these lines by Eq 7 leads to the activation enthalpy DH  3.5 eV. This finding will be discussed together with the following results.
A complete quantitative description of the
thermal relaxation of residual stresses is of practical importance. This was achieved by an extensive study on the steel AISI 4140, which was
quenched, tempered 2 h at 450 C (840 F), and
then shot-peened. The shot peening was carried
out with cast steel shot S170 (44 to 48 HRC), at
a pressure of 1.6 bar and a coverage of 98% (Ref
26, 27). The absolute values of the macro residual stresses are plotted in Fig. 8(a) versus the
logarithm of the annealing time for different annealing temperatures. They decrease with increasing time and temperature. Because the conventional way for determining the Avrami
quantities DH, m, and B in Eq 4 and 5 requires
strong extrapolations to very high and low annealing times, a new iterative method was used
that allows determination of these parameters by
a nonlinear minimization of the residual sum of
squares (Ref 26, 27). To achieve this, all measurements in the evaluation are taken into account with the same weight. The algorithm

Stability of Residual Stresses / 57


yields DH 3.3 eV, m 0.122, and B 1.22
1021 min 1 for the surface values of the macro
residual stresses. The curves in Fig. 8(a) were
calculated using these constants in Eq 4 and 5.
They describe the time and temperature dependence of the relaxation process very well.
The alterations of surface half-widths by annealing at temperatures between 250 and 450 C
(480 and 840 F) for different annealing times
are shown in Fig. 8(b) for the same steel. The
reductions of the HW values are similar to the
relaxation of the macro residual stresses presented in Fig. 8(a). Again, the Avrami approach
is used to describe the relaxation of half-widths,
applying the new iterative method to the differences between the half-widths after annealing
and the value HW 1.65 of a normalized specimen related to their starting values instead of
the ratio rrs(Ta,ta)/rrs
0 in Eq 4. The curves in Fig.
8(b) were calculated by means of the material
properties DH 2.48 eV, m 0.116, and B
1.09 1013 min 1 and agree well with the measured values. The dependence of the relaxation
of mean strains e21 / 2 and micro residual
stresses rrs
micro, respectively, on the annealing
time was measured in the temperature range between 250 and 450 C (480 and 840 F). The
data can also be modeled by the Avrami approach using e2(Ta, ta)1/2/e201/2 instead of

rrs(Ta, ta)/rrs0 in Eq 4. The new algorithm was


used to determine the material parameters DH
2.64 eV, m 0.096, and B 5.32 1012
min1, which allow description of the measured
values as shown in Fig. 8(c).
Comparison of these relaxation data in Ref 26
and 27 shows that the relaxation of macro residual stresses at the surface of the shot-peened state
is a bit faster than the relaxation of half-widths
and mean strains; this corresponds with the fact
that the constant B differs by nearly nine orders
of magnitude (see Table 1). The reason for the
differences in the relaxation rate is that for the
relaxation of macro residual stresses, dislocation
movement is sufficient. For a distinct relaxation
of micro residual stresses, however, additional
dislocation annihilation is necessary. The exponents m show no significant alterations, and the
activation enthalpies approach the values of the
activation enthalpy of self-diffusion of iron: DHS
 2.8 eV. Additional data for the same steel in
a normalized and another quenched and tempered state in Table 1 (with exception of the
hardened state) confirm this statement (Ref 28).
Accordingly, volume diffusion-controlled dislocation creep in the residual stress field that is
dominated by climbing of edge dislocations
should be the rate-controlling process for the relaxation of the shot peening residual stresses of

3.0
400

Ground
350 C/22 min
2.5

200
2.0
0
0
2.4

350 C/22 min


100

2.2

HW, 2

rs, MPa

Milled

2.0

200

0.1

0.2

0.1

0.2

0
Shot peened
3.0

400 C/30 min


500 C/30 min

200

550 C/30 min

2.5

400

2.0
0

0.1

0.2

0.3

0.1

0.2

0.3

Distance from surface, mm

Fig. 4

Macro residual stress and half-width of the x-ray interference line versus distance from surface before and after
annealing of ground, milled, and shot-peened SAE 1045 samples in a normalized state

AISI 4140 in normalized as well as quenched


and tempered states.
In order to obtain some information about the
relaxation behavior of macro residual stresses in
subsurface layers, the values of the residual
stresses after the annealing processes of AISI
4140 mentioned above were measured for different distances from the surface (Ref 26, 27).
The evaluated parameters of the Avrami approach vary only at the surface itself from those
values measured below the surface. There, the
evaluated data show no significant tendency and
amount to the mean values DH 2.99 eV, m
0.172, and B 6.09 1017 min1. As
shown in Fig. 9, the dependence of the macro
residual stresses after different annealing times
at 450 C (840 F) on the distance from surface
can be described quantitatively using the surface
material properties and the mean values for all
subsurface layers. The agreement between measured values and modeled curves is very good.
During heating to sufficiently high annealing
temperatures, a distinct relaxation of macro residual stresses occurs. An attempt was made to
model this behavior by extending the Avrami approach to nonisothermal stress relaxation (Ref
27). The transient relaxation of the macro residual stresses was calculated for specimens that
were immersed up to 90 s in a salt bath at 450
C (840 F). The temperature at the specimen
surfaces developed according to the T(t) curve
shown in Fig. 10. For calculation of the relaxation of residual stresses, the real T(t) relationship
was partitioned into a staircase curve with small
equidistant steps and isothermal sections. Calculation of the relaxation was determined by a
numerical integration after the so-called stresstransient method (Ref 27) under the application
of the Avrami quantities DH, m, and B (Table 1)
which are valid for the surface state. The magnitudes of the macro residual stresses measured
after interruption of the heating and cooling
down to room temperature are marked with triangles. It becomes evident that the stresstransient method describes the measured values
in an excellent manner. Figure 10 also shows the
isothermally calculated relationship between
heating time and absolute values of residual
stresses. While there are distinct differences during the first stage of heating time of approximately 20 s, the curves calculated isothermally
and by the stress-transient method closely approach at long annealing times.
A similar discussion of thermally activated
processes was carried out for other steels (Ref
16, 17, 28, 30, 31) and for AlMg alloys and
CuZn alloys (Ref 30, 31) with surface residual
stresses originating from a variety of processes.
In the case of these steels, the activation enthalpy
DH depends on the state of the material and lies
in the range of 1.1 to 2.6 eV. DH is lowest for
relaxation of residual stresses due to hardening
and highest for those due to deformation in soft
annealed states. That proves unequivocally that
residual stress relaxation in steels can occur by
several processes.
Characteristic structural changes occur during

58 / Effect of Materials and Processing


the deformation, machining, or hardening of
steels. Typically, an increase in the dislocation
density and a change in the dislocation arrangement are observed. In the case of hardening, the
concentration of solute interstitial atoms differs
from the equilibrium value, and this, together
with the presence of dislocations, has a decisive
influence on the residual stress fields. If a single
heat treatment is carried out in the temperature
range corresponding to recovery (T  0.5 Tm),
the dislocations adopt arrangements of lower energy by elementary processes such as glide and
cross slip by screw dislocations and glide and
climb by edge dislocations. In the case of hardened material, diffusion of carbon atoms dependent on the annealing temperature and accompanied by the formation of characteristic
carbides is superimposed on these processes.
The rate-determining process, with the exception
of the early stages in the annealing of hardened
steels (Ref 36), is clearly the thermally activated
climb of edge dislocations (Ref 17). If diffusion
of matrix atoms occurs along the edge dislocations to the dislocation core, the activation enthalpy should be DH  0.5 DHS, where DHS is
the activation enthalpy of self-diffusion. If volume diffusion predominates, the DH value for
climb is determined by DHS. In the real case,
both processes occur simultaneously but to different degrees. The dislocation density and arrangement are of considerable importance. In the

case of randomly distributed dislocations or tangles of extremely high density, qt for example,
in hardened steels where qt  1012 cm2 (Ref
23, 24)residual stress relaxation is expected
to involve dislocation-core diffusion-controlled
climb by edge dislocations. Predominantly volume diffusion will determine recovery if the dislocation configurations are relatively stable and
consist of plane arrangements, cell walls, or subgrain or low-angle grain boundaries. This recovery process probably occurred in the investigated
shot-peened quenched and tempered steels and
normalized steels as well as nonferrous alloys.
In view of the activation energies, the residual
stress relaxation in the hardened steels can probably be classified between these two extremes
and occurred by two recovery mechanisms in
competition with each other.
Up to now, discussion has centered on residual stress relaxation at temperatures Ta  0.5
Tm that is, those temperatures brought about
by typical recovery processes. In this case, mechanical parameters such as hardness and yield
strength are not significantly altered. During a
recrystallization anneal at Ta  0.5 Tm, the dislocation density rapidly takes very small values
as a result of the growth of new grains. This leads
to complete removal of macro residual stresses
and to small micro residual stresses, but is associated with pronounced changes in mechanical
properties. If extensive residual stress reduction

is required in a component without significant


change in yield strength or tensile strength, the
annealing temperature and time must be chosen
to correspond with the recovery stage and not
with recrystallization.

Resistance to Residual Stress Relaxation


Residual stress relaxation by heat treatment is
fundamentally impossible if in a predominantly
uniaxial residual stress state rrs is smaller than
the creep yield strength. This resistance for the
onset of plastic creep deformation, designated in
the following by Rce, is characterized by the
creep strain limit at vanishingly small plastic deformation. As shown schematically in Fig. 11,
Rce decreases with increasing temperature and
load time. In contrast to relevant times regarding
the creep condition of a high-temperature component at service, only very short times are necessary for residual stress relaxation by creep processes (dislocation creep, grain-boundary glide,
or diffusion creep). With increasing T and/or t,
Rce approaches a localized residual stress peak
of magnitude rrs. For T Tti, rrs equals Rce
associated with the localized onset of creep deformation. Further increases in temperature or
time result in an increasing and measurable microplastic creep strain. As illustrated in Fig. 11,
residual stress relaxation begins at higher temperatures (Ti) the smaller the load time (ti) or the
residual stress (rrs) and the greater the creep resistance (Rce) of the material. Changes in Rce can
be achieved by deliberate alterations in the state

0
Shot peened
400 C/30 min

400

500 C/30 min


550 C/30 min

800

Ta,4

Normalized

Ta,3

6
300 C/60 min

400

Quenched and tempered

HW, 2

Shot peened
rs, MPa

Ta,2

rs

log ln (0/rs)

Normalized

Ta,1 < Ta,2 < Ta,3 < Ta,4

800
Quenched and tempered

Ta,1

log ta

400

5
4
Shot peened

800
Hardened
0

Fig. 5

log ta

0.2

Hardened
0.4 0
Distance from surface, mm

300 C/60 min


0.2

Ha/ln10

3
0.4

Macro residual stress and half-width versus distance from surface before and after annealing of shot-peened
SAE 1045 samples in normalized, quenched and tempered, and hardened states

1/kTa

Fig. 6

Schematic of conventional determination of


Avrami approach parameters

Stability of Residual Stresses / 59

Loading stresses acting in opposition to the residual stresses in a given region of the material
delay the onset of microplastic deformation in
this region. Since the distribution of residual
stresses in a component is always inhomogeneous, there will be other regions in which superposition of rrs and rls in the same sense promotes plastic deformation. Therefore, it depends
on the superposition of rls and rrs at any point
in the component if a loading stress retards (|rls
rrs|max  |rrs|max) or enhances (|rls rrs|max
 |rrs|max) residual stress relaxation.
At this stage it is interesting to explain the
result shown in Fig. 1 and 2, where residual
stress relaxation in a deformed sample of SAE
1045 is delayed in comparison to that in SAE
1023 despite the compressive residual stresses
having double the magnitude. Due to its higher
carbon content, SAE 1045 contains a greater number of ferrite-cementite phase boundaries than
SAE 1023. These phase boundaries represent stable obstacles to dislocation slip. Apparently, the
restricted possibilities for recovery-driven movement of dislocations to arrangements of lower en-

(rrs1 r2rs)2 (r2rs r3rs)2 (r3rs r1rs)2

(Eq 9)

on the basis of the von Mises hypothesis. The


equivalent residual stress is thus dependent on
the differences between the principal residual
stress components. These in their turn are proportional to the shear stresses acting on dislocations in the slip systems. Residual stress relaxation therefore does not occur for
rrs
eq  Rce

ep C(T)rntm

but does occur for

can be established for constant loading stress rls


(Ref 16, 17). C is a quantity depending on the
temperature and the condition of the material;
the exponents have the values n  1 and 0  m
 1; and the creep strain increases with applied
stress and with time. Residual stress relaxation

600

rrs
eq Rce

0.4

10

500

250 C

400

300 C

300

350 C

200

400 C

100

The case rrs


eq  Rce obtains only briefly, since the
immediate onset of creep deformation attempts
to restore the condition rrs
eq Rce.

10
102
ta, min

103

104

(a)

Effect of the Magnitude of Residual


Stresses on Relaxation Behavior

3.2

Ta = 250 C

The data in Fig. l and 2 clearly show that the


residual stress relaxation in machined or hard-

102

300 C
2.8 350 C
400 C
450 C
2.4

103

104
600 C

90

525 C

0
10

80

410 C

60

375 C
340 C
320 C

50

250 C

20

40
30

0.8

200 C

10

150 C

102
10
ta, min

103

104

(b)
1.4

200
Ta = 250 C

< 2 >1/2 in 103

70

rs

0.4

50 C
Ta = 4

Residual stress relaxation, %

log ln (0/rs)

Ta = 450 C

0
10

Annealing time, h
1

(Eq 10)

1.2

300 C
350 C
400 C

1.0

250

450 C

200

rs
, MPa
micro

(Eq 8)

rrs
eq

ened steels occurs more quickly or at lower temperatures, the greater the magnitude of the residual stresses themselves. This is a consequence of
temperature, time, and stress-dependent processes similar to those observed in so-called primary microcreep. Empirical relationships of the
form

|rs|, MPa

|rls rrs| Rce

ergy outweigh the increasing driving force of the


larger residual stress. The delayed residual stress
relaxation in deformed SAE 1045 is thus explained by an increase in the creep resistance Rce.
In the case of a multiaxial residual stress state,
the residual stress rrs employed in the arguments
just given must be replaced by an equivalent residual stress rrs
eq. If the principal components are
rs
rs
rrs
1 , r2 , and r3 , it can be formulated to

HW, 2

of the material. All thermally stable obstacles


that have an additional work-hardening effect
(Ref 17) shift the onset of residual stress relaxation and the entire rrs versus T curve to higher
temperatures.
Residual stress relaxation is also affected by
the superposition on a localized residual stress
rrs of an applied loading stress rls in the same
direction. Increasing rls values shift the onset of
residual stress relaxation to shorter times and/or
lower temperatures according to the relation

0.8
150
0.6
10

1.2

102

10
ta, min

103

104

(c)

1.6
1

Fig. 7

Influence of annealing time and temperature on


the (a) absolute values of macro residual stress,
(b) half-width, and (c) mean strain and micro residual stress
on the surface and their description by the Avrami approach for shot-peened AISI 4140 in a quenched and tempered condition

Fig. 8

2
log ta

Plot of log ln versus log ta for a shot-peened 12% Cr steel

60 / Effect of Materials and Processing

(Eq 11)

However, macro residual stresses are inhomogeneously distributed over the cross section, and
for the residual stress relaxation in a localized
area the following relationship holds true:
rrs
0
E

e t e p e e 0

(Eq 13)

Thus, the Norton law and Eq 13

rs

constant ep

r
E

e p C*(T) (rrs)n e e

rs
(rrs
0 r )
ep
E

r rs
E

(Eq 14)

(Eq 12)

Regarding real values of the residua1 stresses,


plastic strains of several tenths of a percent maximum are produced by complete residual stress
relaxation (rrs 0); that is, the deformation is
in the microcreep range.
Equations 10 and 11 form the basis for quantitative estimates of the residual stress relaxation.
Multiaxial and inhomogeneous residual stress
states are neglected or excluded. Although Eq 10
cannot be substituted directly in Eq 11 on account of the variable value of the stress (r
rrs), it can be seen qualitatively from a combi-

give a proportionality between the elastic and


plastic strain rates, and between the residual relaxation rate and the actual residual stress to the
power of the Norton exponent n, respectively.
The experimental data shown in Fig. 12 reveal
strain rates that present a strong dependency on
temperature and residual stress values (Ref 26).
Furthermore, the data reveal strain rates that are
typical for creep processes. This finding supports
the conclusion drawn above that diffusioncontrolled dislocation creep in the residual stress
field should be the rate-controlling process for
the relaxation of residual stresses in steels.

650
200

500

T( t )

200

Ta = 450 C

rs, MPa

rs, MPa

ta = 6000 min
60 min

400
6 min

600

Measured
Calculated

400

Tsalt = 450 C

300

Surface data

450

Transient relaxation

350

T, C

550

250

0.1
0.2
0.3
Distance from surface, mm

0.4

Measured and calculated macro residual stress


at Ta 450 C (840 F) and different annealing
times versus distance from the surface of shot-peened AISI
4140 in a quenched and tempered condition (450 C, or
840 F, for 2 h)

Fig. 9

rs+ ls

t2, 2

20

40

60

80

100

Quenched and tempered


650 C (1200 F)
450 C (840 F)
Hardened

T1

T2

Temperature, T

Fig. 11

Absolute values of macro residual stress after


short-time immersion in a salt bath at 450 C
(840 F) versus time. Comparison is made with curves calculated with the stress-transient method using the occurring
T(t) dependence and with modeling for isothermal annealing.

Fig. 10

Measuring property

DH, eV

B, min1

Macro residual stress

3.30

0.080

1.40 1018

Macro residual stress


Macro residual stress
Half-width
Mean strain (micro rs)
Macro residual stress

t2 >> t1

2
1

Creep yield strength versus temperature at different creep times and strain rates

t, s

Table 1 Material properties DH, m, and B of the Avrami approach determined for different
heat treatment conditions of the shot-peened steel AISI 4140
Normalized

T 1

100

105
106
107

Heat treatment

t1, 1

rs

200

Isothermal relaxation
0 min

800

In certain cases in practice, uniaxial deformation is often employed in addition to stressfree annealing to relieve residual stresses. In the
case of forming, for example, the residual
stresses can be reduced by a second forming
stage using a smaller reduction in the crosssectional area. This can be achieved by redrawing, restretching, rerolling, repressing, and
straightening (Ref 16, 17). However, these techniques can be used only on simply shaped components with a uniform cross section. In the case
of welded seams, a uniaxial load is applied to
reduce or redistribute macro residual stresses.
When a critica1 value of the applied loading
stress is exceeded, directed dislocation movement converts the elastic strain associated with
the macro residual stress into micro plastic
strain.
Several typical examples will serve to illustrate the relaxation of residua1 stresses due to
joining and shot peening by uniaxial deformation. In an evaluation of the effect on the strength
of macro residual stresses set up during welding,
it is important to know the stability of these
stresses on loading the weld. Figure 13 illustrates
the considerable reduction in macro residual
stresses accompanying tensile loading of joints

Creep yield strength, Rce

r
E

Residual Stress Relaxation by


Uniaxial Deformation

Strain rate p, s1

et constant ep ee ep

nation of the two expressions that increasing residual stress values lead to a more effective residual stress relaxation. As a result of the greater
driving force, shorter times and/or lower temperatures are necessary. This is in agreement
with the experimental results presented.
The influence of the magnitude of the residual
stresses can also be illustrated by describing the
residual stress relaxation with the Norton approach, which is known from high-temperature
creep. The total strain rate as the sum of elastic
and plastic strain rate must vanish according to

>>

cannot, however, be compared directly to a creep


test. It is much more like a stress relaxation experiment. In the latter case, the total strain remains constant while elastic strain is converted
into plastic strain. With a homogeneous stress
distribution over the cross section of a specimen,
the following expression would be valid:

3.38
3.29
2.48
2.64
2.33

0.080
0.122
0.116
0.096
0.110

1.90 1021
1.22 1021
1.09 1013
5.32 1012
1.20 1016

108

T = 450 C

109

400 C
350 C
300 C

1010
1011
60

100

200

400

250 C
600

1000

rs, MPa
Plastic strain rate versus mean residual stress of
shot-peened AISI 4140 in a quenched and tempered condition (450 C, or 840 F, for 2 h). Determined
from the data of Fig. 8(a)

Fig. 12

Stability of Residual Stresses / 61


produced by electron beam welding and tungsten
inert gas (TIG) welding in a maraging steel (Ref
37). In the electron beam welded specimen, it
begins at about 70% of the yield strength Re of
the unannealed weld. Residual stress relaxation
is not complete. In the case of the TIG weld,
noticeable residual stress relaxation first occurs
above Re. By the time the tensile strength Rm is
reached, residual stress relaxation is virtually
complete.
Shot peening residual stress states under tensile or compressive loading are unstable, and relaxation occurs when critical loading stresses are
exceeded. Corresponding results of surface residual stresses of the quenched and tempered
steel AISI 4140 are presented in Fig. 14 as a
function of loading stress and in Fig. 15(a) as a
function of total deformation (Ref 38). At tensile
loading, relaxation of the peening-induced compressive residual stresses in the surface of 540
MPa (80 ksi) sets in at 95% of the yield
strength of the unpeened material (Re(t)  Rp0.01(t)
 1180 MPa, or 170 ksi). In compression, relaxation starts at 45% of the yield strength of the
unpeened material (Re(c)  Rp0.01(c)  1300 MPa,
or 190 ksi). In the quenched and tempered condition of this material exists a typical strengthdifferential effect of Re(c) Re(t)  120 MPa (20
ksi) (Ref 38). The relaxation rate of macro resid-

200

Re

ual stresses directly after the onset of relaxation


occurs much more rapidly during tension than
during compression. After a total strain et
1% (rls  1250 MPa, or 180 ksi), the
surface residual stresses rrs are completely removed (see Fig. 15a). If et  1%, the sign of
rrs changes. In the case of tensile loading, the
residual stress relaxation, however, is incomplete. Thus, the relaxation behavior is different
and anisotropic during tensile and compressive
loading. Combined with the relaxation of macro
residual stresses, only a relatively small decrease
of the half-width values, which are a measure of
the micro residual stresses, is observed at the surface. Figure 15(b) illustrates this behavior as a
function of total strain.
The effect of different strengths on the relaxation behavior of shot peening residual stresses
at the surface of AISI 4140 due to uniaxial deformation is illustrated in Fig. 16. Results of Fig.
14 for specimens quenched and tempered at 450
C (840 F) are compared with results for normalized specimens and for specimens quenched
and tempered at 650 C (1200 F) (Ref 39). The
courses of rrs versus rls are similar. However,
the absolute values of the critical loading stresses
for the onset of residual stress relaxation in tension |rlscrit(t)|, and in compression, |rlscrit(c)|, increase with increasing yield strength of this steel
in different heat treatment conditions.
Figure 17 represents the influence of loading
stresses rls on the shot peening residual stresses

Rm

rrs at the surface of quenched and tempered AISI


4140 for different testing temperatures (Ref 40).
The rrs versus rls course at 25 C (75 F) is the
same as in Fig. 14. At higher temperatures (250
and 400 C, or 480 and 750 F), the initial macro
residual stresses are reduced by heating and
waiting for temperature compensation due to
thermal residual stress relaxation. The values of
the critical loading stresses for the onset of residual stress relaxation during tension and compression as well as the relaxation rate decrease
clearly with increasing temperature.
The relaxation of macro residual stresses by
uniaxial deformation begins at relatively small
loads or plastic strains. It may be complete or
partial, both the degree and rate of residual stress
relaxation depending on the type and state of the
material as well as on the nature of the applied
load.

Resistance to Residual Stress Relaxation


The deformation behavior and relaxation of
macro residual stresses in materials can be
roughly approximated in terms of a cylindrical
rod with longitudinal residual stressesnamely,
constant compressive residual stresses (rrs
s ) at
the surface and constant tensile residual stresses
(rrsc ) in the core (Fig. 18). The surface is characterized by the yield strength Re,s and the core
by Re,c. Plastic deformation commences in the
core at the critical tensile loading stress:
rlscrit(t) Re(t),c rrs
c

(Eq 15)

Plastic deformation under compressive load first


occurs in the surface region due to the superposition of loading and residual stresses at the critical compressive loading stress:

200

200

400

400

800

Re Rm

1200

0
rs, MPa

rs, MPa

ls, MPa
Welding residual stress versus tensile loading
stress of a maraging steel. (a) Transverse residual stresses in the welding seam center of an electron beam
welded joint. (b) Transverse residual stresses 3 mm (0.12
in.) from the welding seam center of a TIG welded joint

Fig. 13

200

|rlscrit(c)| |Re(c),s| |rrs


s|

400

Residual stress relaxation does not occur during


tensile loading for

600
6

rls  rlscrit(t) Re(t),c rrs


c
4

t, %

and during compressive loading for


|rls|  |rlscrit(c)| |Re(c),s| |rrs
s|

15

HW, min

rs, MPa

0
200

600
1600

15
30
45

400

800

800

1600

60
6

ls, MPa
Macro residual stress at the surface versus loading stress under tension and compression, respectively, of shot-peened AISI 4140 in a quenched and
tempered condition (450 C, or 840 F, for 2 h)

Fig. 14

t, %

Influence of total deformation on (a) the macro


residual stress and (b) the half-width change at
the surface of shot-peened AISI 4140 in a quenched and
tempered condition (450 C, or 840 F, for 2 h)

Fig. 15

(Eq 17)

30

200

(Eq 16)

(Eq 18)

The properties rlscrit(t) and rlscrit(c) thus have the


meaning of resistances against residual stress relaxation by uniaxial deformation.
The critical compressive loading stress,
rlscrit(c), can be determined experimentally from
corresponding loading and unloading tests. The
yield strength of the surface, Re(c),s, can be estimated from Eq 16 with measurements of the surface residual stresses and the knowledge of the
initial surface residual stress by this relationship:
ls
|Re(c),s| |rcrit(c)
| |rrs
s|

(Eq 19)

In the case of a multiaxial residual stress state,


modified yield strengths of the surface occur.
Consideration of an isotropic biaxial surface re-

62 / Effect of Materials and Processing

req |Re(c),s|
2
ls
rs
(rlscrit(c))2 (rrs
s ) rcrit(c) rs

(Eq 20)

where req is the equivalent loading stress. This


relationship leads to somewhat smaller yield
strengths, as with a uniaxial estimation (Ref 29,
41). The application of Eq 20 for the shot-peened
AISI 4140 in different conditions results in the
Re(c),s values summarized in Table 2. The corresponding critical loading stresses for initiation of
macro residual stress relaxation rlscrit(c), the yield
strengths of the unpeened conditions Re(c), the
ratio Re(c),s /Re(c), and the half-widths ratio of surface and core are also indicated. For the normalized condition the relation
|Re(c),s|  |Re(c)|

is valid. For the quenched and tempered conditions, however, the yield strengths show with
|Re(c),s|  |Re(c)|

a reverse behavior. The shot peening treatments


thus generate in comparison with the core a
work-hardened surface state in the normalized
condition and a work-softened surface state in
the quenched and tempered conditions. The final
observations seem to be in contradiction with the
work-hardening behavior obtained on the basis
of the measurements of the half-width values of
surface and core. The last column in Table 2

presents the ratio HWs /HWc with values between


1.05 and 1.58. Therefore, in all cases a peeninginduced microstructural work-hardening effect
appears. The smallest effect occurs at the 450 C
(840 F) quenched and tempered condition. The
other heat-treated states show a considerably
larger microstructural work-hardening effect.
These apparent contradictions become clear
with a more exact consideration of the conditions
of the local deformation of the surface layers
during shot peening and the subsequent uniaxial
compression test. Besides a biaxial plastic
stretch-forming of the surface area with maximum flow stresses and work hardening direct at
the surface, the peening-induced deformations
include additional Hertzian stresses with maximum effects below the surface (Ref 42). Now,
at the determination of the surface yield strength
Re(c),s under homogeneous compressive loading,
the direct surface layer is deformed exactly inverse to the deformation direction during shot
peening. In this case, the Bauschinger effect relieves the deformations in the inverse direction
and reduces the resistance of the work-hardened
surface layers due to back stresses and an anisotropic mobility of dislocations. It is known that
the Bauschinger effect appears especially distinctive in quenched and tempered steels (Ref
43). Therefore, it is not surprising that the shot
peening treatments of quenched and tempered
AISI 4140 lead to values of Re(c),s /Re(c)  1. The
surface yield strengths are controlled not only by
the peening-induced surface work-hardening,
which is characterized by the half-width ratio
HWs /HWc, but also by the Bauschinger effect.
Obviously, the condition Re(c),s /Re(c)  1 is valid
if the peening-induced work-hardening has a
stronger effect than the Bauschinger effect. If the

Bauschinger effect dominates, however, the ratio


of the yield strengths is Re(c),s /Re(c)  1.

Modeling of the Relaxation Behavior


A quantitative description of the relaxation
behavior of multiaxial macro residual stresses
with known distribution over the specimen cross
section under uniaxial deformation is possible
with the aid of finite-element modeling (Ref 28,
44). Isoparametric rectangular elements with
plane strain behavior were applied for this purpose. The corresponding depth courses of the
macro residual stresses as well as the initial parts
of the stress-strain curves under tensile and compressive loading must be known and must be input with certain assumptions in a discrete form.
Figure 19 shows corresponding results of tests
at 25 and 400 C (75 and 750 F) for AISI 4140
in a quenched and tempered condition. There is
a good agreement between experimental results
and finite-element modeling. An important finding is that besides the pure mechanical and thermal relaxation of macro residual stresses during
heating and waiting for temperature compensation, no additional relaxation effects occur under
200
0
rs, MPa

sidual stress state that is generated due to vertical


shot peening results on the basis of the von
Mises hypothesis for the onset of residual stress
relaxation at compressive loading to:

T = 25 C
T = 250 C
T = 400 C

200
400
600
1600

800

800

1600

ls, MPa
,
200

,
,

Macro residual stress at the surface versus loading stress under tension and compression, respectively, at different deformation temperatures for shotpeened AISI 4140 in a quenched and tempered condition
(450 C, or 840 F, for 2 h)

Fig. 17

Quenched and
tempered 450 C
Quenched and
tempered 650 C
Normalized

rs, MPa

Surface

Core
rs
rs

200

Surface
Cross
section
area

400

srs
600
1600

1200

800

400

400

800

1200

1600

ls, MPa
Macro residual stress at the surface versus loading stress under tension and compression, respectively, of shotpeened AISI 4140 in different heat treatment conditions: normalized; quenched and tempered at 450 C (840
F), for 2 h (qt 450); and quenched and tempered at 650 C (1200 F) for 2 h (qt 650)

Fig. 16

Simplified distribution of longitudinal macro


residual stress for residual stress relaxation due
to uniaxial loading

Fig. 18

Stability of Residual Stresses / 63

The fatigue strength of steels can be greatly


influenced by macro and micro residual stresses.
Since in certain cases they can be considered as
locally variable mean stresses, they can lead to
substantia1 increases in fatigue strength (cf. Ref
16, 17). The interaction of macro residual
stresses with the cyclic deformation mechanism
during fatigue in the microcrack-free stage is of
particular importance. Coupled with this is the
stability or the cyclic relaxation of residual
stresses. Residual stress relaxation occurs when
a critical value of the applied loading stress amplitude is exceeded and a cyclic directed dislocation movement converts the elastic strain associated with the macro residual stress into
micro plastic strain.
Numerous investigators have shown that the
effect of macro residual stresses decreases with
increasing stress amplitude and growing number
of cycles as a result of residua1 stress relaxation
(cf. Ref 16, 17). This is illustrated in Fig. 20 for
quenched and tempered AISI 4140 in a shotpeened condition that has undergone a push-pull
fatigue test (Ref 28, 46). Figure 20(a) presents
cyclic deformation curves with a plot of the plastic strain amplitudes, ea,p, versus the number of
cycles, N, during stress-controlled cyclic loading
for different stress amplitudes, ra. Initially, the
specimens show macroscopically a quasi-elastic
cyclic deformation behavior at 400, 500, and 600
MPa (60, 70, and 90 ksi). At loading with ra
700 MPa (100 ksi), small plastic strain amplitudes occur from the beginning of the test. After
a distinct number of cycles for incubation, which
decreases with increasing stress amplitude, cyclic work-softening occurs at all amplitudes.
This augmentation of plastic strain amplitude
continues until specimen failure. The maximum
plastic strain amplitude observed just before failure increases with increasing stress amplitude.
The alterations of the shot peening-induced
macro residual stresses at the surface rrs with the
number of cycles are shown in Fig. 20(b). The
residual stress values are strongly reduced in the
first cycle. Afterward, a linear dependence of the

Micro residual stresses behave in a complex


fashion during the relaxation of macro residual
stresses by uniaxial deformation. So far, only
isolated experimental data based on x-ray profile
analysis are available.
When a material is formed or machined, both
directed and inhomogeneous micro residual
stresses are set up. Those of the first type are the
result of back stresses due to dislocation pileups
at grain or phase boundaries and elastic strained
second phases of heterogeneous materials. If
these dislocations move in the reverse direction
during localized plastic deformation, the back
stresses and hence the directed micro residual
stresses are initially reduced. Further deformation causes renewed buildup of back stresses in
the opposing direction associated with dislocation multiplication and hardening, which must
once again increase the micro residual stresses.
Dislocation arrangements in a random distribution or in tangles or cells give rise to inhomogeneous micro residual stresses. Microplastic
deformation can lead to the rearrangement of
dislocations into arrangements of lower energy
and thus bring about micro residual stress relaxation. If new dislocations are produced, a renewed buildup of micro residual stresses is superimposed on the relaxation process.
A reduction in the micro residual stresses in
hardened steels is observed both on deformation
and on machining (Ref 16, 17, 32, 45). Two superimposed effects can operate here. After hardening, a very high density of dislocations is presenteither randomly distributed or in tangles.
Microplastic deformation brings about a rearrangement of these dislocations into configurations with lower distortion energy and therefore
a reduction in the micro residual stresses. On the
other hand, solute carbon atoms may be induced
to jump into the energetically more favorable octahedral sites in the martensite lattice under the
influence of the stress field of the moving dislocations. This causes a reduction in the tetragonality and hence the lattice distortion due to
solute carbon atoms (Ref 45).

rrs A(ra) m(ra) log N

(Eq 21)

200
T = 25 C

0
rs, MPa

Behavior of Micro Residual Stresses

residual stresses on the logarithm of N occurs


according to a logarithmic creep law:

200
400
600

Calculated
Measured
T = 400 C

200
0
rs, MPa

Residual Stress Relaxation by Cyclic


Deformation

200
400

Calculated
Measured

600
1600

800

800

1600

ls, MPa
Measured and calculated macro residual stress
at the surface versus loading stress under tension and compression, respectively, at different deformation temperatures for shot-peened AISI 4140 in a quenched
and tempered condition (450 C, or 840 F, for 2 h)

Fig. 19

a = 700 MPa

a,p in 103

quasi-static uniaxial loading at elevated temperatures.

2
1

(a)

600 500
400
0
0
700

rlscrit(c)

Temperature

Re(c)

F

MPa

ksi

MPa

ksi

MPa

ksi

Re(c),s /Re(c)

25
250
400

75
480
750

275
300
340

40
45
49

345
300
275

50
45
40

480
496
516

70
72
75

1.39
1.65
1.87

1.58
1.33
1.28

Quenched and tempered


650 C (1200 F)
20
450 C (840 F)
25
250
400

70
75
480
750

340
600
400
360

49
85
60
52

805
1300
950
830

115
190
140
120

685
953
758
606

100
138
110
88

0.85
0.73
0.79
0.73

1.56
1.05
1.05
1.05

Normalized

500
a = 400 MPa

400

Re(c),s

C

Heat treatment

200

Table 2 Critical compressive loading stress rlscrit(c), yield strength Re(c) of unpeened conditions, surface yield strength Re(c),s, yield strength ratio Re(c),s /Re(c), and ratio of half-width values
HWs /HWc of shot-peened AISI 4140 in different heat treatment conditions

rs, MPa

600

HWs /HWc

600
10

(b)

102
104
Number of cycles, N

106

Plastic strain amplitude and longitudinal macro


residual stress versus number of cycles during
strain-controlled push-pull tests with different stress amplitudes of shot-peened AISI 4140 in a quenched and tempered condition (600 C, or 1110 F, for 2 h)

Fig. 20

64 / Effect of Materials and Processing


tionally peened conditions and warm-peened
conditionsthe changes in the macro residual
stresses, rrs
l,s, measured at the side of the specimens primarily loaded in compression as a function of the logarithm of the number of cycles.
For both peening variants, the residual stresses
at the surface relax more rapidly with increasing
stress amplitude during the first cycle as well as
during further cycling. The linear correlation between residual stresses and the logarithm of the
number of cycles N according to Eq 20 can be
recognized for wide intervals of N 1. In the
conventionally peened samples, however, the

900
800

200

700
600

300

l,s

500
400

400

300
500

600

700

10

102

103
104
Number of cycles, N

105

106

107

(a)
0
a,s, MPa
100

1000

Tpeen = 290 C

900
200

800
700

rs , MPa

600
500

400

400

500

400

600

Top longitudinal
Bottom longitudinal
Top transverse
Bottom transverse

200

0
10

1000

100

l,s

|rs|, MPa

600

a,s, MPa

Tpeen = 20 C

300
800

macro residual stresses relax faster than in the


warm-peened condition. Pronounced differences
can be seenespecially in the first cycle, where
the residual stresses in the warm-peened samples
show a higher stability than those in conventionally peened samples. For N 1, the relaxation
rate of the variant peened at room temperature is
also higher compared to the variant peened at
290 C (555 F).
The resulting surface values of the half-widths
HWs, related to their initial values HWs,0, are
given in Fig. 23 as a function of the logarithm
of the number of cycles. The half-widths of the

rs , MPa

due to the so-called cyclic creep, where the properties A and m can be determined for each stress
amplitude from the experimental data in the linear sections of the curves in Fig. 20(b). In the
case of the stress amplitude ra 400 MPa (60
ksi), after the first cycle no further residual stress
relaxation is observed with increasing number of
cycles N. Obviously, this ra value is smaller than
the materials resistance against cyclic residual
stress relaxation ra,crit at the surface. For the two
highest stress amplitudes, ra 600 and 700
MPa (90 and 100 ksi), at the end of fatigue life
greater residual stress relaxation rates are observed than predicted by the rrs /log N law. The
beginning of this phase can be correlated with
the onset of cyclic work softening in the ea,p versus N curves in Fig. 20(a). Hence, increasing
plastic strain amplitudes at constant stress amplitude may cause increasing relaxation rates of
residual stresses.
For shot-peened AISI 4140 in a quenched and
tempered state and subjected to alternating bending at the stress amplitude ra,s 700 MPa (100
ksi), Fig. 21 presents the course of the absolute
values of the macro residual stresses at the surface of flat specimens versus the number of cycles in the longitudinal and transverse directions
(Ref 40). While the longitudinal residual stresses
at the top side (which was loaded in compressive
direction in the first half-cycle) relax from 600
MPa (90 ksi) to about 480 MPa (70 ksi),
the longitudinal residual stress at the bottom side
shows no considerable relaxation. In the second
half-cycle, this trend is reversed and the residual
stresses measured become similar. In the course
of further cyclic loading up to 105 cycles, the
residual stresses decrease linearly with the logarithm of the number of cycles according to Eq
21. After crack initiation at 105  Ni  4 105,
further reductions in the magnitudes of the residual stresses are measured. The residual
stresses in the transverse direction always relax
more slowly than in the longitudinal direction.
For initial stress amplitudes at the surface between 300 and 1000 MPa (45 and 145 ksi), Fig.
22 summarizes for the same steelin conven-

102
104
Number of cycles, N

700

106

10

102

103
104
Number of cycles, N

105

106

107

(b)
Absolute values of macro residual stress at the
surface versus number of cycles during alternating bending tests of shot-peened AISI 4140 in a
quenched and tempered condition (450 C, or 840 F, for
2 h) for a surface stress amplitude of 700 MPa (100 ksi)

Fig. 21

Macro residual stress at the surface in longitudinal direction versus number of cycles during alternating
bending tests of quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h) at different surface stress
amplitudes. (a) Conventionally shot-peened condition (Tpeen 20 C, or 70 F). (b) Warm-peened condition (Tpeen 290
C or 555 F)

Fig. 22

Stability of Residual Stresses / 65


conventionally peened samples relax for ra,s
700 MPa (100 ksi). With increasing stress amplitude, an increasing relaxation is observed.
The reduction of the half-widths and, hence, of
the micro residual stresses is much more pronounced for the samples peened at room temperature compared to the other peening variant.
For the warm-peened states, the measured values after cycling are often a bit higher than the
initial values. Only for ra,s 900 MPa (130
ksi) and N 104 can a significant decrease be
seen.
For the same samples peened at 20 or 290 C

(70 or 555 F), the stability of the residual


stresses was investigated in detail (Ref 47, 48).
The depth distribution of the residual stresses
and the half-widths in the region close to the
surface was determined at different numbers of
cycles at a surface stress amplitude 1000 MPa
(145 ksi). The curves for the conventionally
peened samples in Fig. 24(a) show that the compressive residual stresses in the region close to
the surface are strongly reduced with increasing
number of cycles. During the first cycle, the absolute value of the residual stress of the surface
decreases from about 600 MPa (90 ksi) to 350

1.0

0.9

a,s, MPa

HWs/HWs,0

300
400
0.8

500
600
700

0.7

800
900
1000

0.6
1

10

102

103

104

105

106

107

Number of cycles, N

(a)

1.0

MPa (50 ksi). After N 1000, the |rrs| value is


150 MPa (20 ksi). The depth where the residual
stresses change their sign is not significantly affected. As shown in Fig. 24(a), a shot peening
treatment at 290 C (555 F) considerably delays
the relaxation of residual stresses. During the
first cycle, the initial absolute value of the residual stress at the surface, |ra,s| 660 MPa (95
ksi), is reduced only by about 140 MPa (20 ksi).
After N 1000, |ra,s| 380 MPa (55 ksi) is
measured. The affiliated depth distributions of
the half-widths of the conventionally peened
condition in Fig. 25(a) verify that with increasing number of cycles, reductions of the halfwidths and consequently of the micro residual
stresses occur due to larger rearrangements of
mobile dislocations in the surface layers. In opposition to this finding, under the same loading
conditions the warm-peened condition shows
practically no relaxation of the micro residual
stresses. Obviously, the dislocation arrangements in the warm-peened condition under cyclic loading are considerably more stable than in
the conventionally peened condition. This is due
to dynamic and static strain aging effects that
occur during and after warm peening (Ref 47,
48). Dynamic strain aging results in a more diffuse and stable dislocation structure caused by
pinning of dislocations by solute carbon atoms
and formation of extremely fine carbides.
The effect of stabilized residual stresses on fatigue behavior is illustrated in Fig. 26, with S-N
curves at alternating bending of the conventionally peened and two warm-peened conditions in
comparison with the ground condition. The S-N
curves for a failure probability of P 50% were
determined in each case from approximately 30
tests after the arcsin P procedure (Ref 49). Shot
peening at room temperature increases the fatigue strength Rfb for about 90 MPa (13 ksi) from
440 MPa (65 ksi) to 530 MPa (80 ksi) compared
to the ground condition. Compared to that, peening at 200 and 290 C (390 and 555 F) increases
the fatigue strength to 590 and 640 MPa (85 and
95 ksi), respectively.

a,s, MPa
0.9

Resistance to Residual Stress Relaxation

400

The cyclic relaxation of macro residual


stresses can be divided into four phases (Ref 16,
17, 50):

HWs/HWs,0

500
600
0.8

700

Quasi-static relaxation during the first half of

800

the first cycle, which is caused by quasi-static


deformation processes
Quasi-static relaxation during the second half
of the first cycle, which is influenced by the
load reversal
Cyclic relaxation at 1 N Ni (Ni number of cycles to crack initiation) due to cyclic
deformation processes, which usually leads to
a linear reduction of the residual stresses as a
function of the logarithm of the number of
cycles according to Eq 21 due to cyclic creep
Cyclic relaxation at Ni N Nf (Nf number of cycles to failure) for sufficiently high
stress amplitudes at the surface, which is

900
0.7

1000

0.6
1
(b)

10

102

103

104

105

106

107

Number of cycles, N

Ratio of half-width at the surface versus number of cycles during alternating bending tests of quenched and
tempered AISI 4140 (450 C, or 840 F, for 2 h) at different surface stress amplitudes. (a) Conventionally shotpeened condition (Tpeen 20 C, or 70 F). (b) Warm-peened condition (Tpeen 290 C, or 555 F)

Fig. 23

66 / Effect of Materials and Processing


enforced in the plastic zones at the tips of
initiated surface cracks.
Assuming a predominantly uniaxial macro residual stress state, it is a fundamental principle
that relaxation of residual stresses at the surface
will commence within the first cycle, if the stress
amplitude ra reaches the critical loading stress
rs,crit. It should be noted that in the case of pushpull tests the critical loading stress rs,crit is identical with the property |rlscrit(c)| (see the section
Residual Stress Relaxation by Uniaxial Deformation in this article). If ra  rs,crit, the macro
residual stresses should remain stable in the first
cycle. For N  1, relaxation of macro residual
stresses can begin gradually if the critical loading stress amplitude ra,crit is exceeded (Ref 16,
17, 40, 41, 47, 48, 50):
ra  ra,crit

Thus, the property ra,crit has the meaning of a


resistance against relaxation of macro residual
stresses. Equation 16 is equally valid for the
critical cycling loading stress ra,crit if Re(c),s is
replaced by the corresponding cyclic yield
strength for surface R cycl
e,s . That means for a predominantly uniaxial residual stress state, relax-

200

ation of macro residual stresses starts together


with cyclic plastic deformation in the direct surface if this condition is fulfilled:
rs
a ra,crit Rcycl
e,s m |rs |

(Eq 22)

The property m has the meaning of a residual


stress sensitivity factor. In this consideration, it
is assumed that m 1. The macro residual
stresses are stable if the condition is fulfilled:
rs
ra  ra,crit Rcycl
e,s |rs |

(Eq 23)

For cyclic bending tests under the conditions illustrated in Fig. 27, the critical loading stress
rs,crit can be found by plotting the residual stress
values after the first cycle from Fig. 22 versus
the loading stress on the surface. This is given
in Fig. 28 for conventionally peened and warmpeened samples. The data points at N 1 show
that the critical loading stress at which quasistatic residual stress relaxation begins is much
higher for the warm-peened condition (rs,crit
500 MPa, or 70 ksi) than for the conventionally
peened variant (rs,crit 310 MPa, or 45 ksi) due
to the above-mentioned strain aging effects. The
residual stress relaxation during cyclic loading

Tpeen = 20 C

rs, MPa

200

N=0

400

N=1
N = 10
N = 102
N = 103

600

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

Distance from surface, mm


(a)
200

Tpeen = 290 C
a,s = 1000 MPa

rs, MPa

200

N=1
N = 10

400

N = 102
600

N = 103

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

Distance from surface, mm


(b)
Macro residual stress in longitudinal direction versus distance from surface for quenched and tempered AISI
4140 (450 C, or 840 F, for 2 h) after different numbers of cycles at the surface stress amplitude of 1000 MPa
(145 ksi). (a) Tpeen 20 C (70 F). (b) Tpeen 290 C (555 F)

Fig. 24

Knowledge of the critical loading stress rs,crit


and the initial residual stress value rrs
s at the surface enables the estimation of the compressive
yield strength of the peened surface region Re(c),s
with Eq 20 if the property |rlscrit(c)| is substituted
by rs,crit according to:
(Eq 24)

Applying Eq 24 for the shot-peened variants investigated, clear differences of Re(c),s can be recognized (see Table 3) (Ref 48). For both variants, however, the resulting compressive yield
strength at the surface is smaller than that of the
core region, which is qualitatively in agreement
with the results in Table 2. This is also due to
the Bauschinger effect. However, the
Bauschinger-induced work softening is much
smaller for the warm-peened samples than for
the samples peened at room temperature, due to
the dynamic and static strain aging effects that
occur during and after warm peening.
By using the critical loading amplitude ra,crit
as well as the residual stresses at the surface
rrss,N1 after N 1, and the assumption that the
residual stress state at the surface is still almost
axisymmetric, the cyclic yield strength at the
surface Rcycl
e,s can be calculated similarly to Eq 24
by:
2
rs
2
rs
Rcycl
e,s (ra,crit) (rs,N1) rs,N1 ra,crit
(Eq 25)

N=0

800

Evaluation of the Relaxation Behavior

|Re(c),s|
2
rs
(rs,crit)2 (rrs
s ) rs rs,crit

a,s = 1000 MPa

800

due to cyclic creep is characterized by a linear


reduction of residual stresses with increasing
logarithm of the number of cycles for 1 N
104 according to Eq 21. Therefore, the residual
stress values at the surface after 104 cycles from
Fig. 22 can be taken as a measure of the degree
of cyclic residual stress relaxation. These values
are shown also in Fig. 28 for both shot-peened
variants versus the absolute value of the applied
loading stress amplitude at the surface ra,s. The
linear fit of these points intersects the line fitted
through the residual stresses measured at N 1
at the critical loading amplitude ra,crit, which is
a measure for the onset of cyclic residual stress
relaxation. This procedure leads to ra,crit 514
MPa (74 ksi) for the conventionally peened condition and ra,crit 714 MPa (103 ksi) for the
warm-peened condition (see Fig. 28).

These values are listed in Table 4 (Ref 48). It


can be seen that Rcycl
e,s for the conventionally
peened samples is much smaller than for the
warm-peened state. Moreover, the ratio of the
calculated cyclic yield strength at the surface and
that for the core region Rcycl
1090 MPa, or
e
160 ksi) found in push-pull tests (Ref 28) indicates that the cyclic work softening that is typical
for quenched and tempered steels does not apcycl
pear in the warm-peened variant Rcycl

e,s /Re
1.07). The increase of Rcycl
at
the
warm-peened
e,s

Stability of Residual Stresses / 67

Behavior of Micro Residual Stresses


Residual stress relaxation when N  l can be
attributed to cyclic plastic deformation. In this
case, the dislocation arrangements present in the
as-received materials are rearranged to configu-

Summary
The stability of the mechanically treated surface state of components during the application
of thermal and/or mechanical energy is desirable

800

P = 50%
700

Tpeen = 290 C
Tpeen = 200 C

600

Tpeen = 20 C
500
Ground
400
104

105

106
Nf

640
590
530

Rfb, MPa

rations characteristic of the fatigued condition at


the given stress amplitude. Especially for steels
with high dislocation densities (cold-worked as
well as hardened or quenched and tempered conditions), dislocation rearrangements are frequently associated with work-softening processes in the microcrack-free stage of fatigue.
Rearrangement of dislocations from configurations typical of the machined or heat-treated
condition to that of the fatigued state is fundamental to the relaxation of macro residual
stresses. Residual elastic strains associated with
the macro residual stresses are once again converted into microplastic strains. If fatigue softening processes are involved, relaxation of micro
residual stresses can also be expected. In the case
of machined and heat-treated materials, this
manifests itself in decreasing hardness and
sharper x-ray interference lines (Ref 16, 17). Differences in the residual stress relaxation behavior
of cold-worked, machined, and heat-treated materials under cyclic load can be attributed to different dislocation arrangements and densities.
The closer the similarity to the dislocation configuration characteristic of fatigue, the less the
extent of dislocation rearrangement and hence of
residual stress relaxation.

a,s, MPa

samples is assumed also to be the result of a very


strong pinning of dislocations by clouds of carbon atoms and extremely fine carbides that exist
due to strain aging effects. This pinning is so
strong that even at highest loading amplitudes,
ra,s, the pinned dislocations cannot move. Therefore, dislocations that were newly generated during cyclic loading cause the macro residual stress
relaxation. The increase of the half-widths for
these variants during alternating bending tests
(see Fig. 23) is a further hint for this assumption.
With these relationships, the effects of the different peening treatments on fatigue strength can
be interpreted. Compared to the conventionally
peened condition, the significantly higher fatigue
strength of the warm-peened variants cannot be
explained only by the slightly higher compressive residual stresses. Moreover, their higher stability seems to be the main reason for this behavior.

440

107

108

S-N curves (failure probability, P 50%) for


quenched and tempered AISI 4140 (450 C, or
840 F, for 2 h). Samples shot-peened at different temperatures are compared with the ground condition.

Fig. 26

Cross section
area

Surface

4.0

Tpeen = 20 C
a,s = 1000 MPa

N=0
N=1
N = 10

3.5

srs

HW, 2

N = 102

Core

N = 103

rs

rs
c

3.0

2.5
Surface
2.0

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

Distance from surface, mm

Simplified distribution of longitudinal macro


residual stress for residual stress relaxation due

Fig. 27
to bending

(a)
4.0

Tpeen = 290 C
a,s = 1000 MPa

100

N = 1 N = 104
Tpeen = 20 C

N = 102

200

Tpeen = 290 C

N = 103

300

l,s

3.0

rs , MPa

N=1
N = 10

3.5
HW, 2

N=0

2.5

400
500

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

Distance from surface, mm


(b)
Half-width versus distance from surface of a quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h)
after different numbers of cycles at a surface stress amplitude of 1000 MPa (145 ksi). (a) Tpeen 20 C (70 F).
(b) Tpeen 290 C (555 F)

Fig. 25

s,crit

600
700
200

2.0

a,crit
s,crit

a,crit
400

600
800
|s| or a,s, MPa

1000

Macro residual stress at the surface in longitudinal direction versus absolute value of surface
stress or surface stress amplitude of quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h) after alternating
bending to N 1 and N 104 in conventionally shotpeened and warm-peened conditions

Fig. 28

68 / Effect of Materials and Processing


Table 3 Quasi-static surface yield strength in compression Re(c),s of quenched and tempered
AISI 4140 in conventionally peened and warm-peened conditions
Shot peening
temperature
C

20
290

rrs
s

rs,crit

Re(c),s

F

MPa

ksi

MPa

ksi

MPa

ksi

Re(c),s /Re(c)

70
555

310
500

45
75

600
660

90
95

801
1008

116
146

0.60
0.78

Table 4 Cyclic surface yield strength Rcycl


e,s of quenched and tempered AISI 4140 in
conventionally peened and warm-peened conditions
Shot peening
temperature
C

20
290

rs,rs N1

ra,crit

cycl
Re,s

F

MPa

ksi

MPa

ksi

MPa

ksi

cycl
Rcycl
e,s /Re

70
555

514
714

74
103

520
620

75
90

895
1156

130
168

0.82
1.07

in order to preserve improvements in mechanical


properties. In this connection, residual stress relaxation behavior is the most important aspect to
be considered. Macro and micro residual stresses
can be reduced or completely relaxed either by
heat treatment or under unidirectional or cyclic
mechanical loading.
Elastic residual strains related to macro residual stresses can be converted into microplastic
strains by suitable deformation processesfor
example, by dislocation slip or creep. The onset
of residual stress relaxation is determined by the
critical applied loading stress, rcrit, with the
meaning of a resistance against relaxation of residual stresses, which is accompanied by the beginning of plastic deformation. In the case of
thermal treatment (rcrit 0), unidirectional deformation (rcrit rlscrit(c) and rs,crit, respectively),
as well as cyclic loading (rcrit ra,crit), residual
stress relaxation occurs when under the assumption of a predominantly uniaxial stress state this
relationship is valid:
|rls| |rcrit| |Ri| |rrs|

(Eq 26)

The characteristic materials resistances are Ri


Rce for thermal treatment, Ri Re(c),s and Re(t),s,
respectively, for unidirectional loading, and Ri
Rcycl
e,s for cyclic loading. Stability of the residual stress state exists, and residual stress relaxation does not occur, if this condition is fulfilled:
|rls|  |rcrit| |Ri| |rrs|

sufficient stable obstacles to dislocation movement, which increases rcrit at the loads given. For
T  0.4 Tm (Tm melting temperature in K)
these could, for example, be grain and phase
boundaries, finely dispersed incoherent particles,
coarse secondary phases, and, as long as no diffusion can occur, solute alloying atoms and certain dislocation arrangements. The onset and extent of residual stress relaxation is influenced in
a complex manner by the combined effects of
heat treatment, unidirectional and cyclic stress,
and multiaxial applied and/or residual stress
states. The kinetic of residual stress relaxation is
essentially determined by the difference |rls|
|rrs| |Ri|. The greater the applied loading stress
isthat is, the higher the temperature, the
longer the time, and the greater the magnitude
and amplitude of the applied stress, plastic deformation, or number of cyclesthe greater the
increase in the rate of residual stress relaxation.
The rate often increases as the absolute value of
the residual stress increases. The fewer the stable
obstacles to dislocation slip, the faster the rate.

(Eq 27)

The applied loading stress r may be zero,


have a value that does not vary with time (e.g.,
thermal residual stress relaxation), have a steadily increasing value (e.g., unidirectional deformation), or have a value varying periodically
with time (e.g., cyclic deformation). Since, according to Eq 26, the onset of residual stress relaxation depends on the resistance of the material
rcrit as well as on the applied loading stress r,
all the parameters characterizing the state of the
material as well as temperature and time or frequency are of importance. The onset of residual
stress relaxation is delayed by the presence of

REFERENCES
1. E. Macherauch and V. Hauk, Ed., Proc. Int.
Conf. Residual Stresses 1, DGM-Informationsgesellschaft, Oberursel,1987
2. B. Beck, S. Denis, and A. Simon, Ed., Proc.
Int. Conf. Residual Stresses 2, Elsevier Applied Science, London, 1989
3. H. Fujiwara, T. Abe, and T. Tanaka, Ed.,
Proc. Int. Conf. Residual Stresses 3, Elsevier Applied Science, London, 1992
4. M. James, Ed., Proc. Int. Conf. Residual
Stresses 4, Society for Experimental Mechanics, 1994
5. T. Ericsson, M. Oden, and A. Andersson,
Ed., Proc. Int. Conf. Residual Stresses 5, Institute of Technology, Linkopings Universitet, 1998
6. A. Niku-Lari, Ed., Proc. Int. Conf. Shot
Peening 1, Pergamon Press, Oxford, 1982
7. H.O. Fuchs, Ed., Proc. Int. Conf. Shot Peen-

ing 2, American Shot Peening Society, 1984


8. H. Wohlfahrt, R. Kopp, and O. Vohringer,
Ed., Proc. Int. Conf. Shot Peening 3, DGMInformationsgesellschaft, Oberursel, 1987
9. K. Iida, Ed., Proc. Int. Conf. Shot Peening
4, Japan Society for Precision Engineering,
Tokyo, 1990
10. D. Kirk, Ed., Proc. Int. Conf. Shot Peening
5, Oxford, 1993
11. J. Champaigne, Ed., Proc. Int. Conf. Shot
Peening 6, San Francisco, 1996
12. A. Nakonieczny, Ed., Proc. Int. Conf. Shot
Peening 7, Institute of Precision Mechanics,
Warsaw, 1999
13. E. Kula and V. Weiss, Ed., Residual Stress
and Stress Relaxation, Plenum Publishing
Corp., 1982
14. A. Niku-Lari, Ed., Advances in Surface
Treatments, Vol 4, International Guidebook
on Residual Stresses, Pergamon Press, Oxford, 1987
15. V. Hauk, Ed., Structural and Residual
Stress Analysis by Nondestructive Methods:
Evaluation, Application, Assessment, Elsevier Science B.V., 1997
16. O. Vohringer, in Residual Stresses: GenerationMeasurementValuation, Vol 1, E.
Macherauch and V. Hauk, Ed., DGMInformationsgesellschaft, Oberursel, 1983,
p 4983
17. O. Vohringer, in Advances in Surface Treatments, Vol 4, International Guidebook on
Residual Stresses, A. Niku-Lari, Ed., Pergamon Press, Oxford, 1987, p 367396
18. M.R. James, in Advances in Surface Treatments, Vol 4, International Guidebook on
Residual Stresses, A. Niku-Lari, Ed., Pergamon Press, Oxford, 1987, p 349365
19. E. Macherauch and P. Muller, Z. Angew.
Phys., Vol 7, 1961, p 305312
20. B.E. Warren, Prog. Met. Phys., Vol 8, 1961,
p 147202
21. B.E. Delhez, Th.H. Keijser, and E.J. Mittemeijer, Fresenius Z. Anal. Chem., Vol 312,
1982, p 116
22. O. Vohringer and E. Macherauch, Hart.Tech. Mitt., Vol 32, 1977, p 153168
23. B. Hoffmann, O. Vohringer, and E. Macherauch, Mater. Sci. Eng. A, Vol 234236,
1997, p 707710
24. B. Hoffmann and O. Vohringer, in Residual
Stresses and Distortion due to Thermal Effect, D. Aurich, K.-H. Kloos, G. Lange, and
E. Macherauch, Ed., DFG-Research-Report,
Wiley-VCH Verlag, Weinheim, 1999, p
160174
25. J. Hoffmann, B. Scholtes, O. Vohringer, and
E. Macherauch, in Proc. Int. Conf. Shot
Peening 3, H. Wohlfahrt, R. Kopp, and O.
Vohringer,
Ed.,
DGM-Informationsgesellschaft, Oberursel, 1987, p 239246
26. V. Schulze, F. Burgahn, O. Vohringer, and
E. Macherauch, Materialwiss. Werkstofftech., Vol 24, 1993, p 258267
27. V. Schulze, O. Vohringer, and E. Macher-

Stability of Residual Stresses / 69

28.

29.
30.
31.

32.

33.
34.

auch, in Proc. Int. Conf. Shot Peening 5, D.


Kirk, Ed., Oxford, 1993, p 264273
H. Holzapfel, V. Schulze, and O. Vohringer,
in Shot Peening: Compilation of Conferences, J. Flavenot and B. Scholtes, Ed., Centre Technique des Industries Mecaniques
(CETIM), 1995, p 1530
T. Hirsch, O. Vohringer, and E. Macherauch, Hart.-Tech. Mitt., Vol 38, 1983, p
229232
J. Hoffmann, Dr.-Ing. thesis, University
Karlsruhe (TH), 1985
J. Hoffmann, B. Scholtes, O. Vohringer, and
E. Macherauch, in Proc. Int. Conf. Residual
Stresses 1, E. Macherauch and V. Hauk,
Ed., DGM-Informationsgesellschaft, Oberursel, 1987, p 695702
O. Vohringer, in Proc. Int. Conf. Shot Peening 3, H. Wohlfahrt, R. Kopp, and O. Vohringer, Ed., DGM-Informationsgesellschaft,
Oberursel, 1987, p 185204
M.W. Fine, Introduction to Phase Transformations in Condensed Systems, Macmillan,
1965
M. Roth, in Proc. Int. Conf. Residual
Stresses 1, E. Macherauch and V. Hauk,

35.
36.
37.
38.

39.

40.

41.

Ed., DGM-Informationsgesellschaft, Oberursel, 1987, p 435440


M. Roth, Z. Werkstofftech., Vol 18, 1987, p
225228
E. Hanke, Fortsch.-Ber. VDI-Z., Vol 5 (No.
9), 1969
H. Hickel, H. Wohlfahrt, and E. Macherauch, DVS-Bericht 26, Strahltech. VI, 1973,
p 5356
H. Hanagarth, O. Vohringer, and E. Macherauch, in Proc. Int. Conf. Shot Peening 4,
K. Iida, Ed., Japan Society for Precision Engineering, Tokyo, 1990, p 327336
B. Eigenmann, V. Schulze, and O. Vohringer, in Proc. Int. Conf. Residual Stresses 4,
M. James, Ed., Society for Experimental
Mechanics, 1994, p 579588
H. Holzapfel, V. Schulze, O. Vohringer, and
E. Macherauch, in Proc. Int. Conf. Residual
Stresses 5, T. Ericsson, M. Oden, and A.
Andersson, Ed., Institute of Technology,
Linkopings Universitet, 1998, p 115120
H. Holzapfel, V. Schulze, O. Vohringer, and
E. Macherauch, in Proc. Int. Conf. Shot
Peening 6, J. Champaigne, Ed., San Francisco, 1996, p 385396

42. H. Wohlfahrt, in Residual Stress and Stress


Relaxation, E. Kula and V. Weiss, Ed., Plenum Publishing Corp., 1982, p 19151921
43. B. Scholtes and O. Vohringer, Hart.-Tech.
Mitt., Vol 41, 1986, p 347354
44. H. Holzapfel, Dr.-Ing. thesis, University
Karlsruhe (TH), 1994
45. B. Hoffmann, O. Vohringer, and E. Macherauch, Mater. Sci. Eng. A, Vol 319321,
2001, p 299303
46. V. Schulze, K.-H. Lang, O. Vohringer, and
E. Macherauch, in Proc. Int. Conf. Shot
Peening 6, J. Champaigne, Ed., San Francisco, 1996, p 413423
47. A. Wick, V. Schulze, and O. Vohringer, in
Proc. Int. Conf. Shot Peening 7, A. Nakonieczny, Ed., Institute of Precision Mechanics, Warsaw, 1999, p 102109
48. A. Wick, V. Schulze, and O. Vohringer,
Mater. Sci. Eng. A, Vol 293, 2000, p 191
197
49. D. Dengel, Z. Werkstofftech., Vol 8, 1975,
p 253261
50. H. Holzapfel, V. Schulze, O. Vohringer, and
E. Macherauch, Mater. Sci. Eng. A, Vol 248,
1998, p 918

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p70-86
DOI: 10.1361/hrsd2002p070

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Effect of Residual Stress on Hydrogen


Embrittlement and Stress Corrosion
Cracking
A.I. Kovalev, D.L. Wainstein, V.P. Mishina, Surface Phenomena Researches Group, Moscow
V.V. Zabilsky, Physical Technical Institute, Russia

METALLURGISTS AND METALS SCIENTISTS traditionally consider hydrogen an undesirable impurity in metals and alloys, associating its presence with the origination of
floccules, porosity, and corrosion. Due to specific behavior of the element in steel, it makes
worse technological and exploitation properties
of the material. Hydrogen embrittlement (HE),
as well as other types of brittle fracture, of metals
result from nucleation and development of microcracks caused by internal stresses. Peculiarities of HE are involved with steel physical properties and mechanisms of its interaction with
hydrogen. Considerable hydrogen solubility difference depending on external factors, a high diffusion mobility, and the capability to form molecules on interfaces that induces a high internal
hydrostatic pressure all promote various types of
brittleness.

Hydrogen in Steel
Sources of Hydrogen Penetration in Steel.
Water-containing raw charge materials and fur-

nace gases involved in the open-hearth (Martins) process are the main sources of hydrogen
penetration in steel (Ref 1). Industrial cast usually contains H2 of not less than 1.5 to 2.0 cm3 /
100 g. Use of vacuum evaporation in a ladle or
at casting is the only way to decrease the content
of hydrogen in steel down to a concentration not
affecting the mechanical properties, of the steel.
Penetration of hydrogen into a carbonic and
alloyed steel was observed in a hydrogen atmosphere at a high pressure (Ref 2, 3). Metal welding is one of the most dangerous technological
processes inducing penetration of hydrogen in
steel. During the process, hydrogen, contained in
electrode coatings as organic compounds and
water molecule, is dissociating, then being in the
atomic state it dissolves in a molten metal and
partly diffuses into neighboring zones of the basic material. Various electrochemical processes,
etching in acids, and interaction with liquid
agents containing hydro-sulfide also promote to
steel saturation with hydrogen. Therefore, hydrogen is an inevitable impurity in steel pro-

T, C

Solubility of hydrogen, cm3/100 g Fe

102

900 800 700

600

500

K K0 p1 / 2 exp(D

H / RT)

lg H 1376/T 0.665

400

(Eq 1)

where K is constant of solubility, K0 is constant


of solubility in standard conditions, R is gas constant, T is temperature, and DH is endothermal
heat of solution. Those of hydrogen in pure iron,
calculated by W. Geller and T.H. Sun (Ref 4)
and then experimentally refined (Ref 57), account for 27230 J/ 0.5 mol H2 (6500 cal/0.5 mol
H2). Solubility of hydrogen in iron at various
temperatures and atmospheric pressures (Ref 5
8) is shown in Fig. 1.
From recent data, solubility of hydrogen in a
metal also depends on its purity, microstructure,
grain size, and a character of distribution of dislocations (Ref 915). The solubility curve equation for -Fe at H pressure of 0.1 MPa (0.015
ksi) is:
(Eq 2)

where H is hydrogen concentration, at.%. Temperature dependence of hydrogen solubility in cFe can be evaluated from (Ref 3):
lg H 1411/T 0.468

(Eq 3)

That in liquid iron at a pressure of approximately 0.1 MPa (0.015 ksi)


lg H 1820/T 0.112

101
0.4

Fig. 1

1700 1400 1100

duced during most modern metallurgical methods.


Hydrogen Solubility and Diffusion in Steel.
Hydrogen solubility in steel depends on temperature and pressure to suit the following equation:

0.6

0.8

1.0
1.2
1000/T, K1

1.4

1.6

1.8

Hydrogen solubility in iron in dependence on temperature at atmospheric pressure. Data source: Ref 25

(Eq 4)

A large quantity of H2, measurable with tens


cm3 per 100 g of metal, can be brought into a
steel from a gas phase only at high pressures and
temperatures (Ref 16). At standard conditions, hydrogen gradually escapes from steel, its solubility
in steel then being negligible. At temperatures
higher than 200 C (392 F) a real solid solution

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 71


of hydrogen in iron becomes unstable (Ref 17).
Thermodynamic data of H diffusion in iron at
high temperatures are well reproducible in various works (Ref 1820) that cannot be told about
the data obtained below 200 C (392 F). Diffusion mobility of hydrogen in iron at low temperatures varies in a wide range and depends on
imperfection of the crystal lattice and the content
of impurities. For example, Ref 2124 there
were obtained values of the diffusion coefficient
at 25 C (77 F) varying from 109 up to
105cm2 /s. When studying diffusion in pure
iron by intensity of hydrogen emission out of
massive samples, it was found that its diffusion
mobility is sharply depressed at lower than 200
C (392 F) (Fig. 2). Above 200 C (392 F) the
diffusion coefficient is the following:
D 1 4 103 exp (3200/ kT) cm2 /s

(Eq 5)

Below 200 C (392 F) that is:


D 0.12 exp (7820/ kT) cm2 /s

(Eq 6)

Just as impurities forming interstitial solid solutions, hydrogen interacts with various crystal
defects, creating nonuniform solid solutions.
Presence of microdeformation near the crystal
imperfections increases hydrogen solubility
there. Therefore, the solubility of hydrogen in
steel rises with an increase in density of defects.
When diffusing inside the crystal, hydrogen atoms are retained by the lattice imperfections, and
its effective diffusion coefficient essentially decreases. This results in sharply diminishing diffusion mobility of hydrogen in iron at low temperatures. Defects retaining hydrogen are known
as traps. Its concentration controls an excessive
hydrogen solubility, compared with the lattice
one (i.e., in an ideal lattice). Moreover, the number of places capturing hydrogen is inversely
proportional to the value of the effective diffusion coefficient.
The role of traps can be played by edge dislocations, boundaries of mosaic blocks or grains,
single pores uniformly distributed in crystals, inclusions of a second phase, and atoms of alloy-

ing elements. The energy of hydrogen interaction with different structural components varies
in a narrow interval (0.10.4 eV per atom). In
particular, the one with dislocations in iron was
evaluated as about 1.6 1020 J (0.1 eV per
atom) (Ref 25).
The Role of Stresses. Hydrogen atoms interact
with any anisotropic strain field, similar to its
interaction with dislocations. If concentration of
hydrogen atoms is insignificant, then distribution
of the dissolved ones in a strain field of a crystal
lattice can be found using:
C C0 exp (U/kT)

(Eq 7)

where C is a concentration of hydrogen in a certain point of the field, C0 is an average concentration, k is the Boltzmann constant, T is the kelvin temperature, and U is the interaction energy
of hydrogen atoms with the strain field in the
place of a defined concentration, C.
Accounting for only hydrostatic components
of the field, the interaction energy, U, makes:
U p dV

(Eq 8)

where p is an average hydrostatic pressure, that


is, p 13(rxx ryy rzz), and dV is the lattice
volume alteration under the influence of hydrogen.
From these equations it follows that at a constant temperature the following condition is
valid:
ln (C/Co) A 13(rxx ryy rzz)

(Eq 9)

where A kT/dV.
According to (Eq 9), hydrogen in the crystal
lattice is redistributed under stress influence.
These deductions were confirmed with experiments (Ref 26, 27). It is settled that the effective
diffusion coefficient of hydrogen essentially decreases, and its solubility increases under loading the metal in a macro-elastic interval (below
yield strength). Thus, stresses promote nucleation of new reversible hydrogen traps, disappearing after elimination of an external stress.
Similar sites arise near the crack tip, in a several

micron size zone of maximal three-axis stresses,


where hydrogen dissolved in the metal lattice
diffuses and its concentration in steel at equilibrium conditions attains maximal values as:
C rmax C0 exp (1.9 103 ry)

Therefore, in steel having the yield strength ry


of approximately 1725 MPa (250 ksi) the equilibrium concentration of hydrogen in the zone of
maximal tensile stresses exceeds its content in
volume by 26 times (Ref 28) (Fig. 3).
At absence of a crack or any other concentrator, the stresses are localized in front of a dislocation pile-up, where considerable enrichment
with hydrogen also can be observed. Both factors, concentration of tensile stresses and dissolved hydrogen, initiate nucleation of cracks
and promote its propagation.
Hydrogen Effect on Structure and Phase
Transformations in Steel. The basic consequence of hydrogen presence in iron and steels
at solidification is the development of porosity,
due to a gas-eutectic transformation (Ref 29)
similar to the well-investigated eutectic one. Total amount of pores is proportional to those of
dissolved hydrogen and is inversely proportional
to the pressure of crystallization. Hydrogen in
steel stabilizes austenite and decreases the critical point A3, the temperature of martensite transformation, and the critical rate of cooling for
quenching (Ref 29). This is the cause of martensite formation even in a carbon-free iron at standard rates of cooling. Unlike a carbonic martensite, at low temperatures of tempering there
arises a hydrogen-free martensite and a gaseous
hydrogen escaping from a sample or accumulating in micropores, creating high pressure and microcracks like floccules (Ref 29).
States of Hydrogen in Steel. Hydrogen absorbed by metal can be present in various states:
An interstitial solid solution
A segregating one on imperfections of the

crystalline structure
Adsorbing one on the surface of micropores

and particles of second phases

T, C
107

800

400

200

100

25

D, cm2/s

107

107

107
107
0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6
1000/T, K1

Fig. 2

Diffusions of hydrogen in pure annealing iron.


Source: Ref 1825

H
H2

HH

H2
H2

H
HH

HH

H
HH
H

H
H2

H2
H2

H
H

Fig. 3

(Eq 10)

Influence of hydrogen on crack propagation. Source Ref 28

H
H
H

72 / Effect of Materials and Processing


In a molecular form, accumulating in micro-

pores
Forming hydrides
Interacting with alloying elements and second
phases
When hydrogen solutes in transition metals,
its s-electrons transfer to a vacant d-shell of a
metal with the lower energy level. The free proton of hydrogen is partly screened by electrons,
which as electronic gas can be involved in its
moving in the metal lattice (Ref 28). Hydrogen
reduces a shear modulus of -iron; at 100 K temperature, this is approximately an 8% decrease
per 1% of hydrogen atoms (Ref 30). At a room
temperature and standard atmosphere pressure
the solubility of hydrogen in steel is insignificant; its supersaturated solid solution is unstable
and decomposes over time. One part of hydrogen
dissolves into the atmosphere and another is
caught by crystalline lattice imperfections. It is
necessary to distinguish two types of defects active in relation to hydrogenpores (or collectors) and traps. Collectors represent volumes
where hydrogen molization is observed and a
considerable hydrostatic pressure develops.
Traps represent defects of smaller sizes than collectors, such as vacancies, substitution and interstitial atoms, edge dislocations, its pile-ups
and thresholds, high-angle boundaries, and fields
of a volume tension (i.e., near the crack tip).
There is a principle difference between traps and
collectors. Transition of hydrogen from a crystal
lattice into traps is reversible, while transition
into collectors is accompanied by formation of
H2 molecule and is nonreversible. The considered features depend on a steel structure.
Ferrite-Pearlite Steels. Nonmetallic inclusions, such as manganese sulfides, are traps of
hydrogen in these soft steels. The maximum solubility of hydrogen in such steel depends not
only on the content but also on the shape of manganese-sulfide particles. The rolled-out prolonged ones increase hydrogen solubility and decrease its effective diffusion coefficient in

direction perpendicular to the planes enriched by


the inclusions.
Globularization of the sulfides decreases hydrogen solubility and increases its effective diffusion coefficient. Sulfide-matrix boundaries are
the sites of a preferential accumulation of dissolved hydrogen, essentially weakening the cohesive forces and promoting disruption of the
interface continuity.
Martensitic Steels. High interior residual
stresses inducing an elastic bend of the martensite crystals and localizing in places of its junction with boundaries of primary austenitic grains
are characteristic for these steels having the
structure of an untempered martensite. In addition, martensitic steels feature a high density of
crystalline structure defects, which are boundaries of grains and twins, single dislocations. All
these elements of structure affect solubility and
diffusion of hydrogen in steel. Dispersed cementite particles observed after tempering also
act as traps for hydrogen. In maraging steels such
a role is executed by intermetallic-matrix interfaces and elastic deformation fields around the
inclusions.
Types of Hydrogen Embrittlement. In general, hydrogen affects metals in two ways:
Physical action at low temperatures when

chemical reactions are not observed

Physical-chemical effect when a chemical in-

teraction of hydrogen with various phases and


components of an alloy on the surface and in
the bulk is observed.
The attempt to enumerate all the phenomena
tied with harmful influence of hydrogen in metals was made in 1972 (Ref 31). Reference 28
specifies the following results.
Hydrogen corrosion: Develops in carbonic

Fig. 4

Microfloccule in ferrite-pearlite steel. Courtesy of


A.I. Kovalev

and low-carbon steels at long-term exposure


in a high-pressure hydrogen at high temperatures. The cause of this brittleness is hydrogen-carbon interaction accompanied by steel
surface decarbonization and methane formation (Ref 32).
Primary gas porosity: Develops due to hydrogen precipitation in a molecular form in a
smelt or in the front of crystallization (Ref 33)
Secondary porosity: Develops at decomposition of a supersaturated solid solution with
formation of small submicroscopic pores
filled by hydrogen (Ref 33, 34)
Decreasing plasticity under tension: Observed at low-rate deformation of steel containing hydrogen (Ref 33, 34)
Delayed fracture: Steel cracking under constant loading below the yield limit
Floccules or disruption cracks: Defects
mainly found in large forged pieces. Floccules are the result of gas hydrogen precipitation in pores and micro flaws during steel
crystallization and cooling of the slab or the
forged piece (Ref 35, 36).
Blisters or bubbles: Develop at electrolysis,
metal exposing in a water vapor or in corrosion agents containing H2S (Ref 33, 35)

Corrosion cracking: For the most part is tied

with precipitation of an atomic hydrogen during corrosion reactions, its adsorption on the
surface and dissolution in metal with development of hydrogen embrittlement
Main Aspects of the Hydrogen Embrittlement Theory. Definite kinds of HE involve two
basic aspectsthe mechanism of hydrogen
transfer to a fracture place, and 2-H influence on
the steel fracture toughness.
Usually hydrogen is transported by moving
dislocations or diffuses (itself) through the crystal lattice under inhomogeneous fields of elastic
stresses (Ref 3740). Being in steel, hydrogen
decreases the fracture work by influencing mechanisms of plastic deformation and reducing the
energy of disruption of interatomic bonds near
the crack tip.
In ferrite-pearlite steels, HE is accompanied
by nonreversible structural damages in the form
of micropores containing molecular hydrogen
under high pressure. Those conditions, however,
are not enough for steel fracture that occurs due
to hydrogen migration together with moving dislocations into the zone of deformation near the
tip of a developing crack. Fracture of soft steels
is always accompanied by a considerable plastic
deformation essentially affected by hydrogen
(Ref 37) facilitating the beginning of plastic deformation of ferrite due to adsorption effects.
However, hydrogen increases the hardening coefficient at higher deformation degrees (Ref 41).
As a result, it localizes the plastic deformation
in the zone of fracture, essentially reducing the
plasticity of steel and diminishing the total fracture energy.
In steels with martensitic structure, hydrogen
migrates through the field of elastic deformations into the zone of three-axial stresses before
the front of a crack (Ref 42, 43). When the content of external and dissolved hydrogen reaches
a critical value, the crack immediately propagates by the mechanism of disruption of interatomic bonds (Ref 44). In this case, the basic
mechanism of HE is tied with breaking weakened bonds in grain boundaries and decreasing
the plastic deformation energy in local sites of
ductile fracture.
Forms of HE of steels are numerous, but only
in some cases (floccules, delayed fracture, and
stress corrosion) the role of macro and micro
stresses influencing hydrogen behavior is clearly
defined.

Floccules
Floccules are interior cracks (flaws) observed
in steel: in the central zone of deformed preforms and, as an exception, in some sites of
molded wares. In polished (not etched) cross
macro-grinds these defects look like hairline
cracks (Fig. 4) opened along a plastic flow direction of forged pieces and rolled products.
Only when considerable axial stresses are present in a pre-form, the orientation of floccules al-

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 73


ters and can be perpendicular to the axis. By
macroscopic investigation of fracture surfaces of
steel affected by floccules, the light brilliant
spots of an oval shape are observed. Fractography research of those sites, using an electronic
microscope, allows the detection of the micro
quasi-chips involving a slightly expressed riverpattern picture, pores, and plastically opened
hairline cracks (Ref 44). These defects can be
found in all steels, but mostly they are observed
in chromic, chrome-nickel, chrome-nickel-tungsten, and chrome-manganese steels. Floccules
originate in forged pieces or rolled stock of large
section some distance apart on the surface under
accelerated cooling at temperatures below 150 to
200 C (302 to 392 F), sometimes at 300 C
(572 F), but more often at room temperature.
One of the main reasons of floccules formation
is a dissolved hydrogen that does not have
enough time to escape large forged pieces at
temperatures below 200 C (392 F) and diffuses
into neighboring pores, creating molecules and
considerable gas pressures (Ref 45, 46). However, hydrogen present in metal is a necessary,
but not the only, condition for origination of the
defects under consideration. Internal tensile
stresses developing under plastic deformation,
non-uniform cross section cooling of steel wares,
and structural transformations stimulate formation of floccules (Ref 4749). Decrease of rupture resistance in sites enriched by nonmetallic
inclusions or in liquation zones promotes this
kind of brittleness. The most favorable conditions occur when the flocculation temperatures
coincide with an interval of development of martensite transformation. Dissolving in austenite in
a higher quantity than in martensite, hydrogen
during martensite transformation moves from a
solid solution into pores. Resultant high hydrostatic pressures summarize with stresses caused
by volume alteration under c phase transformation. For this reason, a quenched-in-air
chrome-nickel steel is very susceptible to flocculation. This is the result of total action of the
molecular hydrogen pressure in pores and the
field of tensile stresses existing in the metal (Ref
50). Floccules are developing during the time,
and the process quite often does not terminate
when cooling finishing and continues for many
hours at room temperature.
At high contents of hydrogen and temperatures approximately 400 to 600 C (752 to 1112
F), formation of H2 molecules in pores results
in intergranular brittle failure of steel, without
participation of external (relative to the pores)
stresses. For fracture at lower hydrogen concentrations and temperatures below 200 C (392 F),
the combined effect of tensile stresses and gas
pressure in pores is necessary. In that case, a
brittle transgranular failure develops.
Processes of formation of floccules are nonreversible. An extremely slow cooling after deforming a preform made of steel susceptible to
flocculation prevents its development. The same
effect was observed at a prolonged isothermal
annealing at heightened or high temperatures.
However, heating in an atmosphere containing

even 12 to 15% hydrogen can again make steel


susceptible to flocculation. That tendency of
steel can develop at welding or cementation
when wares are quenched directly after treatment
heating. Floccules are not observed in a number
of superalloyed steels such as ferritic, semiferritic, and austenitic. High-speed steels are not
susceptible to flocculation, in spite of the fact
that very high internal stresses at accelerated
cooling develop there. Influence of alloying elements on steel susceptibility to flocculation is
tied with action on austenite transformation under cooling, anisotropy of properties, and a quantity of generated hard inclusions. At Ac3 temperature decreasing and Ac3-Ar3 hysteresis
increasing the diffusion of hydrogen and its escaping off steel pre-forms becomes slower. This
explains a high susceptibility to flocculation of
a chrome-nickel steel after its alloying with molybdenum or tungsten. Manganese and silicon,
promoting anisotropy of mechanical characteristics of a rolled stock, favor its susceptibility to
flocculation. Increasing concentration of carbon
in a hydrogen-containing steel promotes formation of methane and floccules.
Preventing Flocculation in Steel. The main
condition of preventing floccules in steel is decreasing its supersaturation with hydrogen and
elimination of H2 molecule formation at temperatures below 200 to 300 C. Also, it is necessary to minimize the anisotropy of mechanical
properties of deformed pre-forms. Melting and
casting of steel in a vacuum is the basic and the
most effective way to eliminate of floccules.
Transportation to forging and rolling of ingots
in a hot state, being in common practice, is undesirable for a steel susceptible to flocculation,
because a high quantity of hydrogen remains.
Special heat treatment schemes preventing flocculation are available for large pre-forms made
of steels of various classes.

Delayed Fracture in Steels


Delayed fracture, or the real hydrogen embrittlement, is an evolving-in-time process of
cracking (fracture) materials exposed to hydrogen and static stresses. This type of hydrogen
embrittlement is the greatest practical problem,
especially for high-strength steels, as far as the
initiation and the growth of crack with hydrogen
present are observed at stresses much less than
the yield strength of steel. The crack growth up
to the critical value with hydrogen present begins
accordingly at the threshold value of the stress
intensity coefficient, K1th, which has a much
lower value than the fracture toughness in the air
or inert medium, K1C. Thus, the danger of the
delayed fracture process exists in that the fracture of constructions occurs under loads that are
entirely undangerous under normal operating
conditions (i.e., in the absence of hydrogen). Another important point is that the process of delayed fracture develops most intensively at temperatures close to room temperature, at which the
overwhelming majority of materials and products are used.

The main factors determining the kinetic features of delayed fracture are the chemical composition and the structure of steel, a hydrogen
content, the presence of stress concentrators, the
test temperature, the amount and the distribution
of impurities and nonmetallic inclusions, and the
presence of inherent stresses (Ref 29). In this
case the delayed fracture can develop both under
the effect of external hydrogeneous media and
inherent hydrogen having entered the metal as
the result of melting, welding, etching, anticorrosive electroplating, and other production
operations. An example is in the formation of
so-called cold cracks in the heat-affected zone
while welding alloyed steels (Ref 51). The increased hydrogen content in the near-weld zone
is the result of hydrogen diffusion from the melt
where it enters due to dissociation of moisture
in electrode coating (welding flux) (Ref 52, 53).
The effect of the steel-strength level on the
resistance to delayed fracture has been supported
by many studies. It has been shown that at the
low-strength level (up to 750 MPa, or 110 ksi)
the influence of hydrogenation is comparatively
moderate. At strength more than 1000 MPa (145
ksi), the resistance to delayed fracture reduces so
that the value of the threshold stress, rth, decreases to 200 MPa (29 ksi) (Ref 53). It has
been found also that as to the high-strength
steels, especially in the presence of stress concentrators, delayed fracture can develop even
with the low-hydrogen content not exceeding approximately 10 4 at.% (1.1 cm3 per 100 g of
metal) (Ref 1). Numerous examples of delayed
fracture of products fabricated from highstrength alloyed steels (shafts made of Fe-0.45C2Ni-Cr-Mo-V steel, axles made of Fe-0.2C-2Cr4Ni steel in the carburized state, etc.) are
presented in a monograph (Ref 54). The presence
of internal hydrogen in association with high residual stresses with a value reaching 1000 MPa,
(145 ksi) is the most commonly occurring cause
for delayed fracture in the case under consideration.
A great deal of publications in which various
methods of hydrogenation are presented are devoted to the study of delayed fracture of steels.
A main body of experimental data has been obtained with the use of the cathode polarization
method. Therefore, in analysis of the influence
of chemical composition and structure, it is necessary to take into consideration that, because of
electrolytic hydrogenation of large-size samples,
the occurrence of nonuniform distribution of hydrogen over the section can give rise to the largescale effect (Ref 55). Hydrogenation from the
gas phase under conditions of elevated temperatures (above 473 K) may cause irreversible damages in the form of bulgings and pores appearing
because of high pressure during the formation of
inner molecular hydrogen or methane. Additionally, in this case the probable surface decarburization can also have an effect on the initiation
and growth kinetics of cracks (Ref 56). From a
practical standpoint, on the contrary, these im-

74 / Effect of Materials and Processing


portant experiments make it possible to simulate
service conditions for the purpose of developing
adequate measures to eliminate delayed fracture.
Main Criteria of Delayed Fracture of
Steels. In general, steel delayed fracture takes

1
1
Stress,

th

Time for failure,


Curves of long-term strength of hydrogenated
steel (scheme). 1, time before failure s; 2, incubation period si, r1, upper critical stress; rth, lower criticastress. Source: Ref 35, 56, 57

Fig. 5

Stress, , MPa

900

700

3
4

500

300
1

100
0

1
20

10

30
40
50
Time to failure, , h

60

100

Dependences of long-term strength of iron (1,


1), steel Fe-0.2C (2, 2), steel Fe-0.45C (3, 3)
and steel Fe-0.8C (4, 4). 14, unhydrogenated state (testings in the air); 14 testings in the process of hydrogenation. Source: Ref 59

Velocity of crack propagation, dl/dt

Fig. 6

IV

III

II

K1c

K1t

Stress intensity factor, K1

Fig. 7

Schematic illustration of kinetic diagram of


cracking of hydrogenated steel. Source: Ref 57

three stagescrack initiation; its slow growth


up to the critical value; and the stage of the unstable, accelerated crack growth. It is common
practice to characterize the sensitivity of steels to
delayed fracture depending on their chemical
composition and structure by so-called long-term
strength dependencies (Fig. 5), which usually are
represented in coordinates: applied stress r, logarithm s, where s is the time up to fracture. Dependencies mentioned are typical for both smooth
samples and samples with the stress concentrator.
The preliminary hydrogenation in constructing
dependencies, s f (r), can be performed by any
of the known techniquescathode polarization,
for example. These dependencies take the analogous character on tests performed just in hydrogeneous media (e.g., on tests in gaseous hydrogen
medium, hydrogen sulfide medium, or in the process of the electrolytic hydrogenation) (Ref 35,
56, 57).
In an analysis of dependencies s f (r), the
following characteristics are used as the main parameters:
The upper critical stress, r1, which corre-

sponds to the value of the tensile strength of


hydrogenated samples
The lower critical (threshold) stress, rth (i.e.,
the stress lower which the delayed fracture
process does not develop)
The time before the crack initiation, si (an incubation period)
The time up to final failure, s, which characterizes the prolonged strength of steel (Ref
35). Thus, a dangerous area of stresses
where the delayed fracture process develops
is an interval between the upper and lower
critical stresses. In this case the value of the
lower critical (threshold) stress, rth, represents a crucial important characteristic,
namely, the prolonged strength limit of steel
under the specific conditions of hydrogen embrittlement.
It is apparent that as the hydrogen content in
steel increases, all of the parameters mentioned
are changed. This especially affects the value of
the threshold stress, rth; the duration of the incubation period, si; and the time up to fracture,
s. From the data given in (Ref 58), for example,
it follows that as the hydrogen content in highstrength steel (1600 MPa, or 232 ksi) increases,
the duration of the incubation period successively decreases from several tens of hours to
several seconds. The time up to fracture in the
limiting case does not exceed 0.1 h. The duration
of the incubation period, si, is influenced by the
preholding under load including holding under
stresses r  rth. Reduction of si, observed in
this case is due to the formation of sites with the
increased hydrogen concentration, which then
serve as points of the crack initiation (Ref 35).
Curves of the long-term strength of iron, and
steels Fe-0.2C, Fe-0.45C, and Fe-0.8C are given
in Fig. 6 (Ref 59). Samples have been polished
mechanically and annealed in vacuum. Delayed
fracture tests have been carried out while holding
the samples under load in the conditions of con-

tinuous cathode saturation with hydrogen in the


26% aqueous solution H2SO4 (the current density i 10 mA/sm2). Smooth thin plates with
the cross section of 0.8 3 mm (0.03 0.12
in.) that ensures the through saturation of samples with hydrogen have been used as samples.
From the results obtained it is evident that the
higher the hydrogen content in steel is, the more
tangible the decrease of the long-term strength,
and it does not practically take place in the case
of armor-iron and low-carbon steel Fe-0.2C. The
value of the threshold stress for steel Fe-0.45C
is of 150 to 170 MPa (20 to 25 ksi) that is several
times lower than the yield strength of this steel.
The lifetime of samples with the stress concentrator under load is mainly determined by the
time before the crack initiation (i.e., the duration
of the incubation period) (Ref 55). In the study
of crack growth it is worthwhile to construct socalled kinetic diagrams of cracking (i.e., dependencies of the crack growth rate, v, on the stress
intensity factor, K1. To construct kinetic diagrams of cracking, fatigue precracked samples
are used. The character of kinetic diagrams of
cracking is shown schematically in Fig. 7. Some
characteristic sections can be distinguished in
the diagram. At section I, which corresponds to
the threshold value of the stress-intensity factor,
K1th, the crack does not practically grow or
grows very slowly. Therefore, it is assumed that
the value of K1th characterizes the condition of
crack nonpropagation in metal. In section II, the
crack growth rate increases rapidly, whereupon
in region III, the crack growth rate over a wide
range, K1, is retained practically unchanged. Unlike the stage of accelerated, unstable fracture
(section IV), the stage of the slow, stable crack
growth can embrace the greatest part of longevity of a product. It is apparent that the delayed
fracture development at this stage is controlled
by the rate of transferring hydrogen to the region
of the crack top. At section IV, the crack growth
rate increases rapidly again, and at K1 K1C
(plane-strain fracture toughness) the final fracture occurs. The characteristic most sensitive to
the hydrogen content is the value of the threshold
coefficient of stress intensity, K1th. According to
data given in Ref 60, an increase of diffusely
moving hydrogen content in high-strength steel
Fe-15Cr-5Ni-2Cu-Ti from 0.14 104 to 2
104 at.% causes the value of K1th to reduce by
77%.
Besides the main characteristics just discussed, a number of other characteristics also
have been suggested for defining the sensitivity
to delayed fracture. For example, dependencies
of the initial coefficient of stress intensity, K10,
the time up to fracture, s, type are suggested to
be used (Ref 60). In this case long-term strength
curves can be described by the logarithmic relations of the K10 a ln s b type bounded
by the value K1C at the top and the threshold
value of the stress intensity coefficient K1th at the
bottom. In work (Ref 61) it was suggested to
construct the long-term strength curves in coordinates r 1/Zs, but not in conventional coordinates r lg s. In this case dependencies of

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 75


the long-term strength s f (r) are transformed
into straight lines convenient for further treatment according to the equation r rth
c(1/ s), where c is a constant. And finally, the
material sensitivity to delayed fracture may be
estimated by its sensitivity to the deformation
rate, namely, by the degree of their plasticity decrease under static loading at the rate of e
104 106 sec1. Such tests, for example,
are in common practice among Japanese researchers in making an estimate of the sensitivity
to the delayed fracture of high-strength maraging
steels (Ref 6264).
Mechanism of Crack Initiation and Its
Growth. It is the opinion of the authors who
studied a phenomenon of delayed fracture, that
in a qualitative sense the mechanism of the crack
initiation when the inner hydrogen embrittlement is as follows. For a crack to be initiated,
the critical concentration of hydrogen C* must
be created in local sites. The hydrogen redistribution in the solid solution is initiated by the
stress field produced by outer load. With the uniform distribution of tensile stresses, the value of
the chemical potential of hydrogen, l0, is the
same in the whole iron-hydrogen system; therefore, the hydrogen redistribution does not take
place. With the nonuniform distribution of
stresses, for example, because of the presence of
a concentrator, the pattern is essentially changed.
At the top of the concentrator (in the region of
maximum tensile stresses), the hydrogen chemical potential, l1, becomes less than that one in
the rest of the regions of the system, l2. The
appearance of the chemical potential gradient (l1
 l2) brings about the diffusion of hydrogen into
the tensile stress region that with time gives rise
to the critical concentration of hydrogen C* and
the crack initiation. The critical concentration of
hydrogen C* is directly proportional to the
square root of the time of diffusion (C* C0
qt, where q is the coefficient of proportionality)
and must be so much higher than the level of
applied stresses is lower (Ref 61).
It is apparent that the increase of the hydrogen
content causes a decrease of the duration of the
incubation period, si, which is necessary for creating the critical concentration of hydrogen.
Lowering of the temperature gives the reverse
result (i.e., an increase of si caused by a decrease
of the hydrogen diffusion rate). Since to create
the critical concentration of hydrogen it is essential to localize tensile stresses, the crack initiation depends on the type of samples used. In
samples with no concentrator, the crack initiation occurs, as a rule, in the axial part of a sample
(i.e.. in the zone of maximum three-axis stresses
where the hydrogen concentration is also maximum). In samples with the sharp stress concentrator, the crack initiation takes place adjacent to
its top; in the case of the less sharp and deep
notches, the failure begins at a certain distance
from the notch top (i.e., under the surface) (Ref
35). At the moderate temperatures, hydrogen
gathers mainly on phase boundaries, microdiscontinuities, grain boundaries, piling-up dislocations, and other defects. Therefore, sites of or-

igin of fracture can be both the interphase


boundaries between ferrite and cementite and
nonmetallic inclusions as wellfor example,
sulfides, which are the most efficient traps for
hydrogen (Ref 56).
The crack growth kinetics upon the internal
hydrogen embrittlement also is controlled by the
hydrogen diffusion rate into the region of maximum tensile stresses ahead of the crack top. It
is suggested that this region is on the boundary
of the plastically deformed zone or at the much
less distance equal to the double opening of the
crack, d. As soon as the hydrogen concentration
becomes critical in the weakest element of a
structure, fracture of this element occurs, following which the crack formed merges into the main
one (Ref 65). The crack propagation is under
progress due to the multiple recurrence of this
process. Therefore, the crack growth can be considered as the process consisting of a sequence
of incubation periods, followed by the restricted
advance of the crack. The process of the crack
stepwise growth is displayed evidently under the
testing conditions at lowered temperatures when
the process of hydrogen diffusion is retarded
(Ref 35). The occurrence of the incubation period of the crack start and the stepwise advance
of separate sections of the crack front is evident
from the results of the registration of the acoustic
emission, which have been obtained from testing
of low-alloy steel with 0.4% C content (Ref 66).
In addition, the registration of the acoustic emission is indicative of the realization of more complicated mechanisms of the crack growth. For
example, when performing work (Ref 67) for
testing the high-strength maraging steel, a
growth of a great quantity of microcracks, 30 to
50 mkm in size, which appeared at 1 to 3 mks
intervals, was revealed.
Contrary to the internal hydrogen embrittlement, in the external hydrogen embrittlement the
crack growth kinetics is controlled mostly by the
following surface phenomena: adsorption, dissociation, and chemosorption of hydrogen on the
metal surface. The following experimentally observed effects serve as a guide for this statement:
The crack growth in the gaseous hydrogen

medium begins practically instantaneously


after applying an external load with no
marked incubation period.
The activation energy of the process at the
stage of the stable crack growth (region III,
Fig. 7) differs noticeably from the activation
energy of the hydrogen diffusion in a solid
solution and is comparable to the activation
energy of surface processes.
When gaseous hydrogen is doped with impurities, which activate (H2S) or suppress
(SO2), the process of the surface interaction
of metal with hydrogen, the crack growth rate
is also increased or reduced, respectively (Ref
57). The location of the failure zone ahead of
the growing crack front depends also on the
type of hydrogen source. In case of the external hydrogen embrittlement, the failure
zone, as a rule, does not reach the point of the

maximum tensile stresses (i.e., it is located


much nearer to the crack top than it is at the
internal hydrogen embrittlement) (Ref 65).
The extent of the embrittlement in the medium
of gaseous hydrogen depends both on the pressure and the chemical composition of a medium,
in part, of the steam content and oxygen, as well.
In the case of high-strength steels a dramatic effect of the hydrogen pressure on the crack
growth rate has been revealed at pressures not
exceeding 1.3 103 MPa (Ref 68). The presence of steams in the hydrogen atmosphere has
a quite profound initiating effect on the delayed
fracture development. Oxygen has the opposite
effect of efficacious retarding of the crack
growth up to the critical value. On data given in
Ref 69, for example, a mere 0.6% of O2 is sufficient for the complete stop of the crack development in the medium with humidified hydrogen
(steel with the yield strength of 1600 MPa, or
232 ksi). The crack stopped by oxygen begins to
grow again only after the complete removal of
oxygen from the medium of tests. It is suggested
that oxygen forms oxides that prevent the hydrogen from penetration into the crack top, and after
the removal of oxygen the reverse process occurs
(i.e., oxides are restored by hydrogen) (Ref 70).
The temperature factor should be set off from
other factors that affect the character of delayed
fracture. According to the data in hand, the maximal sensitivity of steels to delayed fracture takes
place at temperatures close to, room temperature
(Ref 29). But in a number of cases this regularity
is not obeyed. In Ref 71, the temperature dependence of the delayed fracture of the high-strength
steel Fe-15Cr-5Ni-2Cu-Ti containing 5 104
at.% of hydrogen was studied. The threshold
value of the stress intensity coefficient for this
steel is K1th 46 MPa m1 / 2. At 20 C (68 F)
the crack growth occurs at the rate of 2 107
m/s. At 100 C (212 F) the crack growth rate
up to the critical value increases by more than
300 times and reaches 6.6 105 m/s. In this
case the time up to fracture (s) reduces from 500
to 600 min at 20 C (68 F) to s 5 to 6 min
100 C (212 F). At the temperature of 180 C
(356 F) the development of delayed fracture
stops completely. The calculation of the activation energy of the crack growth Q has shown
good agreement between the obtained value
(64.8 kJ/mol) and the activation energy of the
hydrogen embrittlement (60.9 kJ/mol).
Chemical Composition and Structure Influence. The resistance of steels to delayed fracture depends on the carbon content, alloying, and
impurity elements. From the data given in Fig.
6, it is evident that a buildup of the carbon content is attended by the significant reduction in
the resistance to delayed fracture. This effect derives from the fact that with increased carbon
content the strength properties of steel are improved, the level of residual internal microstresses is raised, and the amount and volume of
the carbide phase are increased. The interphase
surface area (i.e., the number of places, or traps,
for the accumulation of hydrogen and the crack
initiation) is correspondingly increased (Ref 29).

76 / Effect of Materials and Processing


The influence of alloying elements is also
quite profound. The effect of noncarbide-formative alloying elements (Mn, Al, Si, Co, Ni) is
mainly connected with their influence on the
properties of matrixthat is, with the change of
its resistance to the viscous, brittle, and grainboundary fracture. These elements (but Ni) decrease the long-term strength of steels subjected
to quenching followed by high tempering (Table
1). Carbide-formative elements (Cr, Mo, Ti, Nb,
V) have an influence mainly through the change
of the amount of the carbide and carbide-nitride
phase, its distribution, and morphology. The
positive influence of molybdenum, niobium, and
vanadium, which bring about to an abrupt increase of the time up to fracture, s, is most pronounced (Table 1). In the case of noncarbideformative elements there is, as a rule, a direct
relationship between the resistance to the hydrogen embrittlement and standard characteristics
of brittle fracturefor example, the critical temperature of brittleness, T50. In the case of carbide-formative elements such an unambiguous
relationship usually is not observed (Ref 72).
As to impurity elements, phosphor and sulfur
have the most negative influence (Table 1). The
negative influence of sulfur is connected with the
formation of manganese sulfides MnS, which,
just as indicated, are the most effective traps for
hydrogen. Manganese sulfides assist in initiating
cracks at more low critical concentration of hydrogen C* in the region of maximal tensile
stresses as evidenced by the results of the reg-

istration of hydrogen in the fracture zone (Ref


73). In this case, a decrease of sulfur to a degree
less than 0.005 wt% does not protect against the
development of the hydrogen embrittlement.
The form of nonmetallic inclusions also has a
dramatic influence, especially in steels with the
ferrite-pearlite structure. Inclusions of the spherical form are considered to be the safest ones.
The negative role of sulfur can be connected in
that this element has a retarding effect on the
recombination of hydrogen atoms on the steel
surface and so brings about an increase of the
amount of absorbed hydrogen (Ref 74).
Phosphorus and its analogs (Sb, Sn) are the
most dangerous on formation of segregation at
grain boundariesthat is, in the process of thermal treatment, which results in the development
of the temper embrittlement (Ref 7577). In the
absence of segregations the steels fracture in hydrogeneous media connected with the intensive
plastic deformation in the crack top. Fracture
mainly occurs transgranularly (i.e., along a
grain). In this case, in fractograms of the crack
growth up to the critical value, either a viscous
cup fracture or a quasi-spall takes place. During
the formation of segregation another mechanism,
the brittle fracture along grain boundaries, is realized with a consequent abrupt decrease of K1th.
In this case, an intensive reduction of K1th takes
place already at initial stages of the segregation
formation when temper embrittlement has hardly
shown up. It is considered that one of the reasons

Table 1 Influence of alloying elements and impurities on transition temperature, T50; work
of propagation of a ductile crack, A; plasticity loss after hydrogenation, Fw and resistance
to hydrogen embrittlement, s (chromium-molybdenum steel, 0.2% C)
Element (limits of
alloying, mass, %)

Dependence of fracture characteristics on


element concentration (per 0.1% of each)

Optimal content of
the element, mass, %

T50, C

A, MJ/m2

Fw%

s, h

for such a strong influence of phosphorus, antimony, and tin segregation is a buildup of hydrogen concentration on the grain boundaries as a
consequence of the chemical interaction of hydrogen with these impurities (Ref 75).
The steel structure has a determining effect on
the resistance to delayed fracture. The least resistance is typical for the martensitic structure,
especially at low tempering temperatures. An increase of the tempering temperature to 420 C
(788 F) causes the value of rth to increase by
several times. The decrease of the volume fraction of martensite to 20 to 30% means to that
even in the hardened state delayed fracture does
not develop (Fig. 8) (Ref 78). In Ref 54 attention
is drawn to the increased sensitivity of steel Fe0.4C-Cr-Si with the lower bainite structure to delayed fracture, although this structure is characterized by high-impact toughness and adequate
plasticity. Unlike the martensitic structures,
whose failure process develops mainly along
boundaries of former austenitic grains, the bainite structure fails along the grain body (by mechanism of quasi-spall).
On evidence given in Ref 74 the resistance to
the hydrogen embrittlement of low-alloyed pipe
steels increases in the direction from normalization to the controlled rolling, but the best results are ensured by the thermal improving
(quenching high tempering) and quenching
from the intercritical temperature interval. A
comparison shows that the thermal improving
and quenching from the intercritical interval ensure about the sixfold enhancement of the longterm strength as compared with the normalization. Water quenching at 775 C (1425 F) when
the structure contains approximately 30% of
bainite affords the best combination of properties. The marked enhancement of the resistance
to delayed fracture is provided by the high-temperature thermomechanical processing of the

C (0.2 to 0.8)

0.20.3

20 to 60

0.1 to 0.2

32

Solid solution
Si (0.2 to 1.8)

0.40.7

4 to 6(a)
5 to 8(b)
3(a)
7 to 16(b)
4 to 10
1 to 1.5
9

0.12
0.05
0.05
0
0
0.17

3
2
2
7(a) to 1.3(b)
1.2
4

5
15
23
6.5
5.7
7.5

0.04
0.01
0

2.9
1.5
8

80
1
260

0
0
0

36(a) to 10(b)
12
10

250
250

400

0.04
0

35
70

2.4
1.0
2.5
2.5
0.04

0
80

360
500

32

1.2

Ni (0 to 3.0)
Co (0 to 3.0)
Al (0.03 to 0.7

0.51
0.5
0.25

Carbide-forming
Cr (0.5 to 3.0)
Mo (0 to 1.0)

0.4 to 0.5

Ti (0 to 0.15)
Nb (0 to 0.20)
V (0 to 0.4)

0.05
0.02 to 0.06
0.1

1(a)
2 to 5(a)
15 to 30(a)
3 to 5(b)
60(a) to 14(b)
0(a) to 40(b)
20(a) to 15(b)

Modifiers
Rare-earth (Ce)(0 to 0.5)
AIN, VN, NbN (0 to 0.2)

0.1 to 0.3
0.2

10
40

Impurities
S (0.015 to 0.045)
P (0.004 to 0.026)
Sb (0.0006 to 0.027)
Sn (0.0005 to 0.03)
Cu (0 to 1.5)

11.5

1400
5

1200
Stress , MPa

Mn (0.6 to 2.5)

1600

1000
800

600
2

200
1

0.01
0.015
0.01
0
0.5

0
230
180
180
1

(a) Alloying for optimal content. (b) Alloying over optimal content. Source: Ref 143

2
3
Time to failure , h

Influence of martensite content and tempering


temperature on delayed failure of quenched carbon steel from 0.35% C. 1, 100% of martensite, 0.023% of
S, without tempering; 2, the same, tempering at 200 C; 3,
the same, tempering at 420 C; 4, 7080% of martensite,
0.025% of S, tempering at 200 C; 5, 7080% of martensite, 0.005% of S, tempering 200 C; 6, 2030% of martensite, 0.026% of S, without tempering. Source: Ref 78

Fig. 8

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 77


high-strength steel Fe-0.3C-2Ni-Cr-Mn-Si (Ref
79).
Some Specific Cases of Delayed Fracture.
In the scientific literature, including the educational one, it was a widespread opinion that delayed fracture of steels can develop involving no
hydrogenthat is, under the effect of only structure factors (Ref 28, 54, 80). In particular, the
question about the nature of the enhanced sensitivity to delayed fracture of steels in the socalled newly quenched state is long discussed.
Cases of quenched products cracking, which are
met in practice of thermal processing (Ref 52,
54), have been associated with the development
of delayed fracture of untempered martensite.
Initiation and growth of crack provoking delayed
fracture occurs mostly over the boundaries of
primary austenitic grains. The short-term tempering or resting of quenched samples at room
temperature enhances resistance to delayed fracture; therefore, its development calls for further
hydrogenation. Within the span of more than 50
years of investigations, alternate hypotheses for
explaining this phenomenon have been proposed. The structure hypothesis based on the
prevailing role of residual microstresses inherent
in the untempered martensitic structure has received the most acceptance (Ref 79, 80). The
role of internal hydrogen, which is always present in steel as an inevitable impurity, has not
been discussed practically in the literature.
The phenomenon of delayed fracture of
quenched steels has been studied in detail in Ref
8183. Steels Fe-0.3C-Cr-Mn-Si and Fe-0.4CCr-Mn quenched in oil at 860 C (1580 F) have
been chosen as objects of investigation. The pro-

cess of initiation and growth of a crack when


testing on a delayed fracture in a freshly hardened state is quite the same as that in the grain
boundaries. The typical initiating crack found in
the stress concentrator top is shown in Fig. 9.
The work has been methodically performed in
the following way. First, regularities of the initiation and crack growth have been studied thoroughly by methods of mechanical tests in conjunction with the acoustic emission. In
particular, the effect of the deformation-rate inhibition of the crack growth has been established, durations of the incubation period, si, under different conditions of loading have been
determined and so on. Second, the simulation of
the newly quenched state has been performed.
For this purpose, after approximately 70 h exposure at room temperature (to eliminate the sensitivity to delayed fracture) quenched samples
have been hydrogenated by the special procedure.
Subsequent tests have shown that all the characteristics of delayed fracture of quenched
steelsthe threshold stress, rth, the duration of
the incubation period, si, and so on, as well as
the mechanism of fracture, are analogous with
the case of delayed fracture caused by hydrogen.
Therefore, it is beyond the reason to consider
that in the case of the untempered martensite
structure a certain different type of delayed fracture that is not connected with the presence of
hydrogen in steel takes place. The fact that the
kinetics of relaxation of interior microstresses in
the process of resting at room temperature does
not correspond to the kinetics of the change of
the threshold stress, rth, bears witness to this
also. Judging by the change of the interior fric-

80
3

, %

60
40
1
20
2
0
2500
3

B, MPa

2000
1500
2
1
1000

500
340

380

420

460

500

540

T, C
Nucleating grain boundary crack revealed at the
top of the stress concentrator on testings of
quenched steel Fe-0.3C-Cr-Mn-Si for delayed failure.
1100. Source: Ref 81

Fig. 9

Influence of testing media on delayed failure of


steel Fe-18Ni-9Co-5Mo-Ti. 1,1 testings in the
air, u 4045%; 2.2, the same, u 6570%; 3,3,
vacuum testings, P 1.3 Pa. Source: Ref 86

Fig. 10

tion, the microstress relaxation process completes in a time of no more than 10 to 20 h since
quenching. In this state a material still possesses
the sensitivity to delayed fracture. In the process
of holding samples at room temperature, the evolution (desorption) of hydrogen from them into
the atmosphere has been found to occur. The
sensitivity to delayed fracture is just concurrently with termination of hydrogen desorption
(i.e., about 70 h after quenching has been
elapsed). According to the data of the laser spectroscopy, the amount of desorbent hydrogen over
the period of holding comes to approximately 2
106 at.% (Ref 84).
The source of diffusive-mobile hydrogen in
the structure of the quenched steel is inner hydrogen, whose amount at conventional process
of melting makes up no less than 2 to 3 104
at.%. The lattice solubility of hydrogen increases
from approximately 107 at 20 C (68 F) to 4.4
104 at.% at 860 C, (i.e., by about 3 orders
of magnitude), as a result of heating to the austenization temperature. A reverse decrease of
solubility in the process of the fast quenching
results in a part of hydrogen being kept for some
time in the oversaturated state in the -solid solution and has the increased diffusive mobility
and the property of localizing in the region of
maximal tensile stresses. High interior microstresses inherent of the quenched steel structure
make the process of the crack initiation and the
crack growth easier but are not the factors that
control kinetic parameters of delayed fracture.
Another interesting case is delayed fracture of
high-strength maraging steels, which develops
over the certain temperature interval of aging
400460 C (Ref 85). The development of delayed fracture in these steels requires no special
hydrogenation. Therefore, within a long period
of time this phenomenon was also connected
only with the peculiarities of the structure state
forming at indicated aging temperatures. Furthermore, this problem is urgent because aging
in the indicated interval provides for obtaining
the maximum elastic limit of steel. However,
owing to the low resistance to delayed fracture,
this mode of aging usually is not practiced. An
analysis of data in hand has shown that the key
question in explaining the phenomenon of delayed fracture of maraging steels is a question on
the role of the testing environment (air, vacuum)
as well as the role of titanium (Ref 85).
The investigations have been carried out with
steels Fe-16Ni-10W-Mo-Ti and Fe-18Ni-9Co5Mo-Ti (Ref 86, 87). The tests have been performed under static loading at the reduced rate
of e 2.8 105 s1. Dependencies of the mechanical properties of steel Fe-18Ni-9Co-5MoTi on the aging temperature when testing in the
air and in vacuum are presented in Fig. 10. A
crevasse in curves of the change of mechanical
properties on tests in the air (curves 1,1) corresponds to the aging temperature of 420 to 460
C. Plastic properties of steels in this range come
practically to the zero level. An effect of delayed
fracture for steel Fe-16Ni-Mo-10W-Ti is observed within a more narrow temperature inter-

78 / Effect of Materials and Processing


val at 420 C. On testing in vacuum (residual
pressure of 1.3 Pa) the indexes of mechanical
properties abruptly increase (curves 3,3). An effect of the delayed fracture elimination is due to
the change of the crack-initiation mechanism.
On testing in the air the brittle grain-boundary
fracture occurs, on vacuum testing the viscous
inter-granular fracture takes place (Fig. 11). In
the study of the titanium influence it has been
found that the influence of the testing environment takes place only with a definite content of
titanium in steel exceeding 0.5%. When titanium
content is 0.5%, mechanical properties tests in
the air and vacuum tests do not practically differ
from one another (i.e., delayed fracture does not
develop).
It has been found also that resistance to delayed fracture depends on moisture of the environmental atmosphere (the laboratory air),
namely, on a season in the course of which tests
are carried out. The figure shows that an increase
of relative humidity (from 4045% to 6570%)

Fig. 11

results in the deterioration of plasticity and the


reduction of the tensile strength by 30% (curves
1,1, 2,2). Thus, tests have been carried out in
the air-dried chamber, for which purpose the special procedure has been devised. On testing in
the air-dried atmosphere (u 0.04%) plastic
and strength properties of steel Fe-18Ni-9Co5Mo-Ti come to such a degree that corresponds
tests in vacuum (i.e., the effect of delayed fracture disappears). This means that the structure
state formed on aging within the interval of 400
to 460 C possesses the increased sensitivity to
the presence of water vapors in the air environmentthat is, delayed fracture of high-strength
maraging steels is caused by the effect of atmospheric moisture adsorbed on a surface.
Investigations of grain boundaries of fractured
samples by x-ray electron spectroscopy have
shown that on the grain boundaries there are hydrides, first of all, titanium hydrides, TiHx, apart
from apparent components (segregations of alloying elements, oxides, hydroxides, and inter-

Mechanism of failure of steel Fe-16Ni-10W-Mo-Ti at crack initiation stage (aging at temperature of 420 C).
(a) Testings in the air. (b) Vacuum testings (P 1.3 Pa). 1050. Courtesy of A.I. Kovalev

H3O

102

H2O

10-20

Oxide-hydroxide layer

P
Grain 1

~10
H3O

Ni3Ti

Hads

Grain 2

Grain 1

Oxygenless zone

Grain boundary
(a)

Fig. 12

(b)

Grain 2

Ti segregation
Ni3Ti

H2O

Air

Habs
TiHx

Grain boundary

Mechanism of the crack growth on delayed failure of maraging steel. (a) Stage before crack initiation. (b) Stage
of crack growth. Source: Ref 87

metallics). Since the process of hydride linking


of the Ti-H type in the air is practically excluded
(titanium, by virtue of its enhanced activity, in
its interaction with oxygen and moisture in the
air forms first of all the hardly permeable oxide
phase TiO2), the conclusion has been drawn that
titanium hydrides are formed on inner interfaces
(grain boundaries). In other words, the TiHx formation occurs in the zone of tensile stresses
ahead of the crack top as diffusive-mobile hydrogen enters there.
According to the mechanism suggested (Ref
86), the process of grain boundary embrittlement
includes the following successive stages: adsorption of atmospheric moisture at the metal surface, oxidation of the most electronegative steel
components (iron atoms), transport of H3O
ions in the film of adsorbed water toward the top
(mouth) of the growing crack, formation of
atomic hydrogen H0ads on the freshly formed surface by cathodic reduction of H3O ions on
nickel atoms, transition of atomic hydrogen
H0ads into the absorbed state and subsequent diffusion of hydrogen to the tension-stressed zone
ahead of the crack top, and embrittlement of
grain boundaries due to both the formation of
titanium hydrides and segregations of hydrogen
solutes ahead of the crack top. Thus, the final
stage of the process is the formation of diffusivemobile hydrogen, which causes embrittlement
owing to both the formation of titanium hydrides, TiHx, and the direct action (decohesion).
This is the main difference of the studied type of
brittleness from the delayed failure of the
quenched steel, which is caused by internal diffusive-mobile hydrogen formed in the process of
steel heat treatment.
The mechanism of the crack growth, which
illustrates the sequence of stages under consideration is represented schematically in Fig. 12
(Ref 87). The formation of titanium hydrides in
the crystal lattice is attended with the considerable volume effect, which is about 3 times higher
than for the martensite transformation in steel.
Besides, hydrides favor the crack growth owing
to the reduced strength of the hydride-matrix interface (Ref 28). Atomic hydrogen H0ads formed
in the crystal lattice is ionized producing ions
H and H, which quantitative ratio, on data
given in Ref 88, makes up 1:9, approximately.
The rate of diffusion of ionized hydrogen is 1 to
2 orders of magnitude higher than that of atomic
hydrogen. According to data obtained by the
method of the secondary ion mass-spectrometry
(Ref 88, 89) more mobile ions H, which serve
as main carriers of diffusive-mobile hydrogen in
metal, are responsible for the formation of the
hydride phase and the process of embrittlement
of the grain boundaries. Ion H possesses the
lesser mobility in the crystal lattice, and its contribution to the embrittlement is not so considerable essential.
In Ref 90 and 91, delayed failure of high-alloy
steel Fe-24Cr-7Ni-3Mo-Al-Ti containing no
martensitic structure has been studied. This steel
displays the enhanced sensitivity to delayed failure in the so-called newly-smelted state (i.e.,

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 79


under the effect of metallurgical hydrogen determining the role of which in the development
of delayed failure is of great practical importance). The studied steel structure is embrittled
d-ferrite with austenite c-phase interlayers uniformly distributed in it. Unlike the cases discussed previously, the initiation and the crack
growth in the structure under study occur in accordance with a transgranular mechanism, with
defects of the cast structure (pores) of 20 mkm
in size being the source of fracture. As evident
from the registration of the acoustic emission,
the crack grows according to the mechanism of
the discrete (stepwise) growth of separate sites
of the crack front followed by the stop at viscous
sites (austenitic inter-layers). According to observations, the resistance to delayed fracture depends on the period of the natural recreation after
smelting. Within the span of 4 years the metallurgical hydrogen content drops from approximately 18 104 to 6 104 at.%, and it is
a limiting one at which the development of delayed fracture is still possible.
The object under study is of interest as far as
it allows the different mechanisms of the embrittlement of grain boundaries or their volume
to be realized. The thermal treatment has been
carried out under operating conditions including
the high-temperature preheating to 1250 C
(2282 F) for obtaining the one-phase structure
of d-ferrite and postaging (annealing) at different
temperatures. The hydrogen content in thermally
treated samples does not exceed 1.4 104
at.%. Annealing at 680 C (1256 F) makes it
possible to obtain the brittle fracture along the
grain boundaries, which, on data of the auger
spectroscopy, arises from the segregation of sulfur and phosphor impurities. After the 850 C
(1562 F) annealing on the grain boundaries of
ferrite, the abundant precipitation of embrittling
r-phase responsible for the quasi-brittle grain
boundary fracture takes place. Aging at 440 C
(824 F) makes it possible to obtain the brittle
intergranular fracture in accordance with the
transgranular mechanism.
The tests performed on embrittled samples
have once again favored the view that structure
factors bringing about the embrittlement of the
grain boundaries or their volume cannot in themselves result in the development of delayed fracture. As is obvious from this data obtained from
steels with the different structures, the development of delayed fracture is only possible with
the participation of diffusive-mobile hydrogen
independently of the source of its origination,
namely on quenching, smelting, from atmospheric moisture, and so on.

Stress-Corrosion Cracking of LowAlloy Steels


Stress-corrosion cracking (SCC) is a specific
type of delayed fracture that develops under the
concurrent action of tensile stresses and the corrosion medium. The given type of fracture occurs in chemical, gas-oil producing, metallurgic,

shipbuilding, and other branches of industry, and


accounts for approximately 30% of material
damage caused by corrosion. The most dangerous media that can cause SCC under the certain
conditions have been established for most steel
classes used. The greatest body of information
accumulated to date refers to high-alloy steels,
in part, to corrosion-resistant steels with austenitic structure. Stress-corrosion cracking of lowalloy steels has been studied least, although such
steels, especially in the high-strength state, have
an enhanced tendency to corrosion-mechanical
fracture. Low-alloy steel cracking can occur under the action of various actuating media including solutions of acids, nitrates, alkali, and sulphur-hydrogeneous media as well. The low
resistance of high-strength steels to SCC makes
itself evident even in such, at first sight, undangerous media as pure water, moist air, and water
vapors (Ref 92, 93).
One of the characteristic peculiarities of SCC
is a pronounced selectivity of a material relative
to a corrosion medium. Therefore, the results of
one metal-medium system investigation cannot be extended to another system. Besides, the
noncorrodive media, in common conception of
this term, may appear to be the most aggressive
ones, because the corrosion cracking process is
actively developing in them under the effect of
tensile stresses. Another interesting peculiarity
lies in a usual insensitivity of high-pure metals,
including iron, to SCC in any environment. The
susceptibility of iron to cracking under the effect
of stresses appears only in the presence of introduction impurities, namely, carbon and nitrogen.
The response to the cold plastic deformation is
also intriguing. Unlike austenitic corrosion-resistant steels, the cold plastic deformation of carbon steels makes them relatively insensitive to
SCC. Annealing of the cold-deformed 0.15% Ccontaining steel at 600 C (1112 F) restores this
sensitivity to the level of nondeformed steel (Ref
70).
Peculiarities of Crack Initiation and
Growth Mechanism under SCC. The mechanism of initiation and growth of cracks under
SCC depends on the steel structure, the value of
applied stresses, and the testing medium. Fracture along the primary austenitic grain boundaries is the most dangerous from the standpoint
of loss in corrosion-mechanical strength. Carbon
steels, as well as many of the low-alloy steels,
fail in alkaline, nitrate, phosphate, carbonate solutions, and also in aqueous acid solutions by
such a mechanism (Ref 75). Depending which
reaction controls the cracking processelectrochemical, anodic, or cathodictwo main SCC
mechanisms are differentiated: the active-part
corrosion (APC) or the hydrogen-induced corrosion cracking (HISCC). Kinetics of the anodetype cracking is determined largely by the conditions of passivation (i.e., by the peculiarities of
the protection corrosion films formation and by
their properties). This is reflected in the fact that
the region of danger of the given type of SCC
corresponds to the so-called active-passive region of potentials in the polarization curves. The

region of cracking by the HISCC mechanism has


no clearly defined boundaries and extends from
the potential of free corrosion to the potential of
cathodic protection (Ref 94).
The mechanism of part cracking presupposes
the accelerated metal dissolution, which is considered to be connected with the local disruption
of passivity at the crack top i.e., with the protective film breakdown and occurrence of the newly
formed (juvenile) surface. The mechanism of
HISCC involves the following main reactions:
discharge of hydrogen H3O ions from water
solution onto the steel surface, the transport of
hydrogen to the region of the elevated stress concentration ahead of the crack top, and the crack
propagation caused by the local hydrogen embrittlement. The concrete mechanism of the hydrogen embrittlement depends on the structure
and properties of steel, the loading conditions
and the ability of a medium to hydrogenation
(Ref 94).
Whatever the mechanism of corrosive-mechanical fracture is, the three stages of the process are distinguished: incipient defects (microcracks) formation, microcracks growth, and fast
fracture as soon as the main crack reaches the
critical size. Both metallurgical defects of the
surface and defects of corrosion origin may present as incipient defects (Ref 93). In the SCC
theory, particular attention is given to the processes proceeding in the crack top, which are
much different from the processes proceeding on
the surface. In particular, an important role is
given to the newly formed (juvenile) surface,
which appears either under the effect of the external loading or in the process of APC. All the
particular SCC reactions, namely, APC, cathode
precipitation of hydrogen, its absorption, and so
on, are sharply speeded up on such a surface. The
newly formed surface possesses a more negative
potential compared with both the previous surface and the passivated surface of crack edges.
It is suggested that in the corrosion crack there
is a local galvanopair of the newly formed surface-crack edges surface type that contributes to
the accelerated selective crack propagation (Ref
95).
Criteria and Methods of Estimating Sensitivity to SCC. A great body of different criteria,
namely, mechanical, electrochemical, physical,
and so on, has been suggested to estimate the
sensitivity of steels to SCC (Ref 92). All the
known methods of corrosion-mechanical testings can be divided into three main groups:
Testings under constant external load (P
constant)
Testings under constant strain (e constant)
Testings at the constant load rate (e, v constant).
Samples without a notch, with a notch, and with
the precreated fatigue crack can be used in each
of these above-stated methods.
The service conditions are best reproduced by
corrosion-mechanical testings under constant external load. In this case, as the crack initiates and
grows, the stress rises in the rest part of a sample
that contributes to the development of fracture.

80 / Effect of Materials and Processing


By analogy with delayed fracture (see the section
Delayed Fracture in Steels), the sensitivity to
SCC is characterized by the time till a crack initiation, si, durability till complete failure, s, and
the value of the threshold stress, rth. Unlike delayed fracture, it often occurs that dependencies,
s f (r), are not brought out onto the plateau,
when testings for corrosion cracking are performed (i.e., there is no threshold stress). The
conventional threshold stress, rth, which should
be determined in this case, is determined at the
specified (base) duration of testings. The duration of the incubation period si during which the
crack initiation occurs takes usually 80 to 90%
of the whole time prior to fracture (s) (Ref 96).
Testings under constant (fixed) strain are usually carried out according to the scheme of threeor four-point bending of samples fixed in the
rigid attachment. On such testings the external
load is reduced gradually because of the crack
formation and the residual bending strain accumulation. Testing under constant strain render
well the cracking processes in the material structure that develop under the action of residual
stresses.
Tests at constant load rate are mainly performed by the slow strain rate testing procedure
(SSRT). In this case the resistance of a material
to cracking is characterized by the value of tensile strength in the corrosion medium, characteristics of plasticity, W, d, the critical fracture
stress, r*, and the area below the strain curve, r
e, characterizing the fracture work. With the
plasticity characteristics taken into accounts, the
relative (as compared to the air) sensitivity of a
material to corrosion cracking is determined.
Used as criteria in this case are the so-called coefficients of medium effect of the (W0 W1)/
W0 type, where W0 and W1 are the relative narrowing on testings in the air and in the corrosion
medium, respectively. The strain rate e with the
use of the SSRT method is usually selected
within the range of 104 to 106 sec1.
When using samples with a crack, a load, the
load line displacement, the crack length, and the
crack opening displacement (COD) are registered as per the SSRT method. For analyzing the
test results, the criteria of fracture mechanics are
used. The threshold (critical) stress-intensity factor, KISCC, which makes it possible to estimate
quantitatively the steel resistance to cracking in
the corrosion medium is taken as the main criterion. The value of KISCC corresponds to the beginning of the crack growth before the critical
value (Ref 97). It is apparent that if K1  KISCC
or r  rth, the fracture of the construction element inevitably takes place. When the criterion
KISCC is used, the sensitivity of material to the
action of the corrosion medium is characterized
by the ratio KISCC /KIC, where KIC is the stressintensity factor when testing in the air (planestrain fracture toughness) (Ref 98100). In all
the schemes of testings the maximum length of
a crack resulting in fracture, the crack growth
rate, the number of cracks per unit surface, etc.,
can be used as characteristics of a material (Ref
101, 102).

The most reliable values of KISCC are obtained


on the basis of standard kinetic cracking resistance diagrams analogous to that represented in
Fig. 3. The plateau-figurative regions in kinetic
diagrams are connected with the branching and
blunting of a crack. If relative resistance of materials to the action of corrosion media is to be
determined quickly, the SSRT method has an advantage over the constant load testing method for
determining KISCC, although it is considered the
basic one. The fact of the SSRT method high
efficiency is evidenced by the data of many scientists. The console samples made of Fe-0.3C2Ni-Cr-Mn-Si steel, for example, have been
tested in the NaCl 3.5% solution (Ref 103). The
values of KISCC obtained at a constant rate of load
and on standard testing (under constant load)
have turned out practically to be equal to 11.69
and 1196 MPa m1 / 2, respectively. To choose the
proper load rate is of great importance in this
case. At the high load rate the corrosion process
has no time to produce changes at the crack top;
at low rate, on the contrary, the passivation process develops at the crack top, thus preventing
from corrosion cracking propagation (Ref 96).
The linear mechanics of fracture are unsuited
for high-ductility steels. In this case, resistance
to crack growth is characterized by the value of
the J-integral, in particular, its threshold value,
JISCC. A tendency to corrosion-mechanical
cracking is estimated also by electrochemical parameters, namely, by the critical value of the hydrogen index of the medium pH, the electrode
potential, E and the density of anodic current at
which the cracking process begins. When testing
the carbon steel in the 3% solution of NaCl,
strong jumps of the electrode potential (300
mV) and the anodic current at the moment of a
crack propagation have been detected (Ref 104).
This effect is connected with revealing the juvenile surface, which has an enhanced electrochemical activity and it can also be taken as one
of criteria for the sensitivity of steels to corrosion-mechanical fracture. The important information on kinetics of SCC, especially at initial
stages of the process, can be obtained from
acoustic emission measurements (Ref 105107).
Effect of Alloying Elements on Resistance
to SCC. The effect of most alloying elements on
steel resistance to SCC depends on their concentration, the presence of other alloying elements,
impurities, the corrosion medium content, the
value of electrode potential, and other factors
(Ref 108). For high-alloy austenitic steels there
is a certain relationship between the effect of an
element and its position in the periodic system.
The analogous relationship is not observed for
low-alloy steels (Ref 109). The theory of alloying that makes it possible to predict the resistance of low-alloy steels to SCC in different media also has not been developed. Hypotheses
suggested provide, as a rule, an explanation for
the role of alloying elements by only one of
some factors, namely, effect on strength of interatomic bonds, the rate of forming protective
films, the effect on the carbon distribution in
steel, and so on (Ref 108).

Information on the effect of alloying elements


on mechanisms of the crack growth in different
corrosion media is necessary to develop successfully new grades of steels with an enhanced
resistance to corrosion cracking. For lack of this
information, the empirical data are usually used
which are often contradictory. The findings on
the individual and combined effects of 27 alloying elements and impurities on the resistance to
SCC of iron and low-alloy steels, including highstrength steels, have been generalized in Ref
108. The effect of elements depending on the
medium (aqueous solutions of chlorides, nitrates, carbonates, alkali, hydrogen sulfide, etc.),
the temperature, the electrode potential, and the
pH value have been considered. The following
elements have been analyzed: N, Al, B, V, H,
Ge, Ca, O, Co, Si, La, Mn, Cu, Mo, As, Ni, Nb,
Sn, Pb, S, Sb, Ti, C, P, Cr, Ce, Zn. It has been
shown that the majority of these elements, especially in the case of combined alloying, ambiguously affects resistance to SCC. The same
element can have both a positive and negative
effect depending on the corrosion medium. Silicon, for example, enhances steel resistance to
cracking in nitrate solutions but reduces it in alkali ones. However, the previously listed elements can be divided arbitrarily into four main
groups: those that enhance the resistance to SCC
(Ti, Nb, Mo, Al, Co, Cu, Pb, B, Ce, Ca, La),
those that reduce it (N, O, H, S, P, V, Sb, Cr,
Mn, Ge), elements of the alternating effect (C,
Si, Ni), and neutral ones (Sn, As, Zn).
The characteristic properties of distribution of
elements and impurities in the steel structure are
of great importance. Segregations of sulphur and
phosphor impurities, for example, change the
electrochemical characteristics of the grain
boundaries and initiate fracture by the local parth
corrosion mechanism (Ref 108). Steels produced
by vacuum melting possess the enhanced resistance to SCC in so far as they contain 2 to 3
times as few detrimental impurities as standard
steels do (Ref 110). It has been noted that the
effect of alloying elements can be judged by the
mechanism of fracture. Elements with a positive
effect assist in realizing the transgranular mechanism of the crack growth. Those that have a
negative effect cause cracking along the grain
boundaries. The effect of the chemical composition may manifest itself indirectly, namely,
through the change of the structure and mechanical properties of the steel. Therefore, the thermal processing can drastically change degree
and even the direction of some alloying elements
effect.
Role of Structure and Thermal Processing
in the SCC Process. The basic parameters of the
SCC, namely, the threshold stress, rth, corrosion
crack-toughness, KISCC, the crack growth rate,
and so on, depend to a great extent on the following structure factors: phase composition,
austenite grain size, state of grain boundaries,
and nature and amount of nonmetallic impurities
(Ref 111). Figure 13 shows that the crack-tough-

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 81


ness reaches its peak values KISCC in hightempered steels. In the region of high-strength
states realized at low tempering, the parameter
KISCC takes quite low values not exceeding 20
MPa m1 / 2.
As for steels of mean and low strength, the
structures of fine-grain ferrite, martensite, and
bainite possess the best properties. The spheroidized structure with uniformly distributed finegrain carbides possesses somewhat worse properties, and the lamellated pearlite structure
possesses the worst ones (Ref 112). There is also
evidence that compound structures, for example,
consisting of fine-grain ferrite and bainite
wherein the intergranular ductile mechanism of
failure is realized (Ref 113, 114) show the highest resistance to SCC. For the steel Fe-0.3C-MnSi-Ni with the compound structure of martensite
lower bainite, the value KISCC in the NaCl
3.5% solutions is 21 MPa m1 / 2. In the case of
the martensitic structure this index is reduced by
half (i.e., steel with this structure possesses the
much lower resistance to SCC) (Ref 115).
The positive role of residual austenite is observed frequently. In the scope of the SCC model
by the hydrogen embrittlement mechanism this
derives from the fact that the saturability of austenite with hydrogen is essentially higher, and
the hydrogen diffusion factor is three to four orders of magnitude lower than that in the martensite structure. Cracks propagating in martensitic
slow down in the region bordering hydrogen-resistant austenite. However, with the insufficient
stability of residual austenite its negative effect
is a possibility. In this case, as residual austenite
is turned to martensite under deformation at the
crack top, the resistance to SCC reduces (Ref
116).
The use of high-temperature thermo-mechanical processing which raises both the parameters

100
1
2
3
4
5
6
7

KISCC, MPa m1/2

80

60

40

20

1000

1400
B, MPa

1800

Strength level effect of steels Fe-0.3C-2Ni-CrMn-Si (1,2,3), Fe-0.4C-Cr-Ni-Mn (4,5), and Fe0.3C-Cr-Mn-Si (6,7) on value of threshold stress-intensity
factor KISCC in corrosion medium. 1,4,6,8, tests in water;
2,5,7, tests in H2S solution; 3, tests in gaseous H2S. Source:
Ref 93

Fig. 13

KISCC and KIC in the case of high-strength combined-alloy steels (when testing in the NaCl
3.5% solution), holds promise (Ref 117).
Quenching from the intercritical temperature interval can also be profitable. The formation of
the ferritic-martensitic structure as the result of
quenching from the intercritical temperature interval (740 C, or 1364 F) makes it possible to
enhance vastly the resistance of steel Fe-0.5CCr-Mo to the hydrogen SCC in salt water (Ref
118). The two-stage austenitization (high-temperature conventional one) forms a toothed
structure of grain boundaries, which also inhibits
the grain boundary failure by enhancing the resistance to SCC.
The effect of grain size on the resistance of
steels to SCC is then ambiguous. In many works
it was pointed out that the finer the austenitic
grain, the higher resistance to SCC is observed.
For example, on the data given in Ref 119 and
120 the time before failure in the NaCl 3.5%
solution increases by 2 to 3 orders of magnitude
with decreasing the austenite grain in the steel
Fe-0.4C-Cr-Ni-Mo from 172 to 12 mkm in size.
However, there is other evidence. For example,
overheating in the process of the austenitization
of the high-strength steel Fe-0.45C-2Ni-Cr-MoV favors a great increase of KISCC when testing
in distilled water (Ref 121). The high-temperature austenitization is most profitable for cast
steels in so far as they are less prone to the grain
growth as compared to hot-strained steels (Ref
122). Unlike quenching, the overheating in normalizing, as a rule, has a negative effect, especially in the case of low-alloy steels containing
manganese and vanadium (Ref 123).
The problem on the effect of initial structure
is of great importance. It is found that, compared
with hot-strained steels, cast steels have higher
values of the corrosion cracking resistance, although standard mechanical properties of cast
steels are much lower (Ref 122). Studied in Ref
124 is the effect of different methods of melting
(an electric furnace, an arc-vacuum furnace, a
vacuum induction furnace, electroslag remelt)
and their combinations on the corrosion mechanical properties of maraging steel. It has been
shown that combined methods of melting make
it possible to increase considerably (by 25%)
KISCC when testing in the NaCl 3% solution. The
analogous data have been obtained in work (Ref
110). Commercial Fe-0.4C-2Ni-Cr-Mo and Fe0.4C-2Ni-Cr-Mo-Si steels that have been melted
in the air and vacuum were tested in the chloride
medium at K 0.9KI. Vacuum-melted steels
possessed the enhanced SCC resistance, whereas
in air-melted steels, cracks developed practically
at once.
In corrosion-mechanical steel failure one of
the key problems is the problem of nonmetallic
impurities. In this case, both morphological
characteristics of impurities (the quantity, form,
sizes, and distribution) and their chemical activity in the given corrosion medium are of importance. The impurities of sulphides, silicates, and
oxides as well, which play the main role in the
process of the crack initiation, are the most ac-

tive (Ref 111). The local electrochemical corrosion processes developing in the corrosive medium around the impurities, resulting in
inevitable formation of aggressive medium-filled
microcavities. The role of MnS impurities in the
course of corrosion-mechanical failure of the Fe0.2C-Cr-Ni-Mo type reactor steel intended for
the production of high-pressure containers was
studied in (Ref 125). The suggested mechanism
of initiation and growth of cracks includes, as
the first stage, the appearance of a slot between
an impurity and a matrix owing to dissolving
MnS in the electrolyte. Atomic hydrogen is adsorbed at crack lateral surfaces, following which
it diffuses into the metal toward the crack top
and causes local cracking.
In acid media sulfide impurities dissolve and
yield H2S that greatly reduces the resistance to
SCC (Ref 126). Dissolution products of some
impurities enhance the adjacent metal corrosion
that also promotes the cracking process. The enhancement of cracking in dissolving nonmetallic
impurities is supported by the data given in Ref
127 devoted to the study of SCC of the Fe-0.2CMn-Ni-Mo steel in the medium intended for
cooling atomic reactors. As the medium has been
contaminated by dissolution products, the MnS
samples have broken down due to the development of the corrosion cracking process, whereas
in the medium free of the corrosion products, this
process has not occurred. At the same time, as
shown in Ref 128, with cathode deposits present,
the SCC process of pipe steels of Fe-0.1C-2MnSi-V-Nb and Fe-0.1C-2Mn-Si-V, and so on type,
which have been chosen from zones of main
pipeline failure cannot be explained in the context of the suggested model of the sulphide impurity dissolution (Ref 128). On authors data,
the negative effect of sulphides is connected
mainly with their role as concentrators of internal stresses (Ref 129). Nonetheless, in spite of
distinctions between interpretations, it is believed that the enhancement of the corrosion-mechanical steel strength can be attained by reducing the nonmetallic impurity content that is
achieved in part by rational deoxidizing and by
modifying the steels by the addition of rare-earth
and alkaline-earth elements.
Stress Corrosion Cracking of Steels for
Main Gas-Pipe Lines. In many countries,
namely, the United States, Canada, Australia,
and Russia, there are recorded emergency failures of subterranean buried gas-pipe lines because of pipe steels SCC (Ref 130, 131). The
summary data displaying the time of trouble-free
operation of main gas-pipe lines used by three
major Russian companies are given in Fig. 14.
Pipes of different diameter (1020, 1220, 1420
mm) and different chemical composition have
been analyzed. It has been found out that pipes
put into service from 1967 to 1980 have the resistance to SCC much higher than those which
have been laid recently in a period from 1981 to
1989. The time of safe service of the former is
from 9 to 23 years, that of the latter is only from
2 to 14 years. It is significant that all the shortlived pipes have been fabricated from steels of

82 / Effect of Materials and Processing


the Fe-0.1C-2Mn-Si-V-Nb and Fe-0.1C-2MnSi-V type produced in France, Italy, Germany,
and Japan. Also, it has been found that in the
most cases (more than 60% of emergencies) the
failure happens in places with the increased
ground humidity. Cracks are revealed most often
in the lower part of pipes. The mean rate of crack
growth is about 1.15 mm/year (0.05 in./year).
The critical length of cracks that cause failure is
in the range of 140 to 150 mm (5.55.9 in.) (Ref
129).
According to current concepts, main gas-pipe
line steels are susceptible to two types of SCC:
APC and HISCC. The first (classical) type of
SCC is developed at medium acidity of pH 
9.5 (high pH SCC); the second one is developed
at pH  6.5 (low pH SCC). In both cases crack
groups (colonies) orientated along the axis of the
pipe arise (Ref 94). More high density of cracks
in the colony is inherent in cracking by the APC
mechanism. Furthermore, it has been found that
in the overwhelming majority of cases, SCC at
a high pH is observed within the limits of regions
in extent of 20 km (12.4 miles) from the compressor stationthat is, where the temperature
and the pressure of gas have their top values (Ref
132134). As to cracking by the HISCC mechanism, the analogous regularity is not fulfilled,
but in this case, the well-defined connection with
the weld location becomes evident. A major part
of failures at low pH begins in centers located at
a distance of 200 to 250 mm (810 in.) from the
weld that is connected with the increased level
of processing residual stresses. The failure at
high pH develops mainly in the intercrystalline
manner (i.e., along the grain boundaries), at low
pH, it develops in the transcrystalline manner
(Ref 94).

Period of safe service, years

30
25
20
15

C
T
Y

T
10
T

5
0
1966

1971

1976

1981

1986

1991

Year of putting into operation


Period of gas main pipes service till first emergency failure depending on year of puttingthem into operation and manufacturing country. Closedcircle data points, steels Fe-0.1C-2Mn-Si-V-Nb and
Fe-0.1C-2Mn-Si-V made in Germany, Italy, France, and Japan; open circle data points, steels Fe-0.17C-Mn-Si, Fe0.15C-2Mn-Si made in Russia; rhomb, steels Fe-0.14C2Mn-Si-N-V, Fe-0.17C-2Mn-Si-N-V made in Russia. Size
of small circle corresponds to 1020 mm pipe diameter;
middle one, to 1220 mm pipe diameter; and large one, to
1420 mm pipe diameter. Y, T, and C denote companies
Ultransgas (Y), Tyumentransgas (T), and Severgasprom (C),
respectively. Source: Ref 129

Fig. 14

The composition of the corrosion medium


forming under the peeled-off insulation is the
main factor on which the realization of the given
SCC mechanism depends. High pH media are
concentrated carbonate-bicarbonate earth solutions with temperatures up to 75 C (135 F).
These solutions are formed under conditions of
the cathode protection that initiates an increase
of pH (acidifying) of the corrosion medium on
the pipe surface. The crack grows as a result of
the rapid electrochemical metal corrosion at its
top where the main processes stimulating the
crack growth, namely, the stress localization, the
plastic strain, emerging of slip strips on the
newly formed surface, are grouped. Low pH
SCC develops in deleted earth electrolyte media
containing dissolved CO2. It is suggested (Ref
130, 133137) that hydrogen arises in microregions filled with the electrolyte (i.e., in slots, pittings, etc.). As a result of corrosion product hydrolysis, the electrolyte in these microregions is
acidified to a level sufficient to discharging ions
of hydrogen, which then diffuses into the region
of tensile stresses ahead of the crack top. As in
the classic model (high-pH SCC), the plastic
strain at the crack top, which benefits the metal
hydrogenation, is required as the obligatory condition (Ref 135, 138). In this case, as seen from
the direct measurements (Ref 94), the zone of
the increased hydrogen concentration extends
for approximately 10 mm (0.4 in.) beneath the
crack surface.
Pipe surface metallurgical defects arising
from rolling may present as nucleous cracks.
These defects, for example, arise from surface
microcracks generating at the stage of continuous steel casting. Cracks initiate on the continuously casted ingots over definite temperature
intervals, mainly over the interval from 700 to
1000 C (12921832 F) (webbed and cross
cracks), and at temperatures close to the solidus
one (spider-shaped cracks). In the given temperature intervals, low-alloy steels are subject to
the strong embrittlement connected with the
change of the grain boundary chemical composition. The present state of the problem concerning high-temperature steel brittleness and the
mechanism of the surface crack initiation in continuous casting has been analyzed thoroughly
(Ref 139, 140).
Practically all the steels applied in the presentday gas-pipe lines, whatever their chemical composition, structure, and strength level, are prone
to SCC to some extent. That is why the prediction of a steels resistance to SCC takes on great
significance. The results of testing a number of
pipe steels for SCC have been correlated with
their mechanical properties determined in the air
(Ref 141). In addition to the standard mechanical
properties, the mechanical properties of samples
under bending with the preliminary-created fatigue crack have been determined. Any correlation between the resistance to SCC and the standard mechanical characteristics has not been
detected. At the same time, according to the
analysis made, a distinct relation is observed between the threshold stress, rth, in the corrosion

medium and characteristics of the mechanical


properties of samples with a crack while performing bending tests in the air. The resistance
to SCC in the studied aggressive media enhances
with an increasing of such characteristics as failure work, yield strength, and ultimate strength.
On this basis, the conclusion has been drawn regarding the necessity of introducing an additional testing of pipe steels in manufacturing
plants, namely, tests of samples with a crack to
predict the resistance of different types of steels
to SCC.

Effect of Alloying and Impurity


Elements on Hydrogen
Embrittlement Resistance of Steel
Hydrogen-resistant engineering steels should
have a number of the following basic characteristics:
High resistance to a brittle fracture for pre-

vention of development of an origin crack

Low content of nonmetallic inclusions and

phases suitable as hydrogen traps

Low level of internal stresses for elimination

of an opportunity of crack nucleation

Presence of surface protecting films prevent-

ing penetration of hydrogen in bulk


To ensure the first requirement, the structure and
alloying of steel should provide:
Hardenability of steel to gain a homogeneous

structure of martensite or its mixture with


lower bainite
Decreasing the ductile-brittle transition temperature, T50, and growth of an energy for a
ductile crack propagation, ac
Low-grain-boundaries
concentration of
harmful impurities
Influence of alloying elements and impurities
on these basic characteristics has been set (Ref
51). The corresponding data, also involving the
optimal content of the basic and impurity elements providing a high hydrogen resistance of
steel, are shown in Table 1. In this case, tendency
of steel to hydrogen embrittlement was estimated from the relative reduction of its plasticity
under tension:
FW

W0 WH
100%
W0

before (W0) and after (WH) hydrogenation of


plain samples during 5 h, and also from reduction of the average time for failure, st, of the ringnotched ones (h 1 mm, r 0.25 mm) under
approximately 0.6 ry loading.
Carbide-forming elements affect the quantity,
distribution, and morphology of carbonitrides.
The carbides are spheroidized in such a manner
that interfaces of the particles with a matrix are
not collectors for hydrogen. These elements in a
solid solution depress the hydrogen thermodynamic activity; as a result, the occluding ability
of steel is essentially diminished (Ref 142).

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 83


Those additions in steel have dissimilar influence on the resistance to different kinds of fracture. If its content does not exceed an optimal
value (Table 1), resistance to all kinds of brittleness increases (Ref 143). At higher contents, the
resistance of steel to hydrogen is increased, but
it is reduced to other kinds of brittle fracture.
Carbide-non-forming elements affect steel resistance to a ductile, brittle, and intergranular
fracture. Addition of 0.1 % Si, Mn, Co or Al in
a solid solution of an improved engineering steel
reduces its fracture resistance; the transition temperature, T50 rising by 5; and the impact strength
at decreasing by 0.07MJ/m2. Nickel is an exception; 0.1% of it makes T50 lower by 4 to 10.
Similarly, these elements influence hydrogen resistance of steel. Thus, manganese, aluminum,
silicon, and cobalt decrease the time for failure,
st, and increase the loss of plasticity, Fw, under
hydrogenation. Nickel increases steel resistance
to hydrogen.
Thus, when these elements are added in a solid
solution, the direct dependence between the hydrogen and brittleness resistance characteristics,
T50, ac, is observed.
Impurities such as sulfur, phosphorus, antimony, tin, and other elements have great influence on the resistance of engineering steels to
hydrogen embrittlement. Their action is manifested via weakening of grain-boundary binding
(temper-embrittlement processes) and via nonmetallic inclusions formation. According to existing opinions (Ref 144), loaded grain junctions
can be zones for the origination of three-axial
stress state. These grain junctions are energetically preferable for the accumulation of hydrogen atoms having a low chemical potential
value. It is assumed that hydrogen decreases the
cohesive forces between iron atoms at grain
boundaries (Ref 145). Segregation of impurities
(phosphorus, antimony, tin, etc.) at boundaries
of original austenite grains results in recurring
temper embrittlement, which further weakens intergranular cohesive forces. These considerations have been proved by direct experiments
(Ref 146). The results indicate that hydrogen
does not interact with impurities or segregations
at grain boundaries (i.e., the grain-boundary embrittlement actions of impurities and of hydrogen
mutually supplement each other).
Published data indicate that steel improving
with impurities and gases cleanliness enables it
to boost its fracture resistance (Ref 147). The use
of metallurgical refining and remelting processes
greatly increases hydrogen-embrittlement resistance due to the removal of nonmetallic inclusions from the steel, especially sulfides and
stringer oxide inclusions (Ref 148).
Improving the quality of heat-hardenable engineering steels (unalloyed with molybdenum
and tungsten) enables them to boost by 2 to 3
times their resistance to brittle and ductile fracture.
Sulfur is present in steel mainly in the form of
manganese sulfides. Nonmetallic inclusions of
this type are very ductile and, hence, during rolling are rolled out in the form of elongated

streaks. Sulfide streaks have S:Fe:Mn ratio of


2:1:1, which comply with the formula (FeMn)S;
their thickness is 10 to 15 nm, and they are
mainly located at the boundaries of original austenite grains (Ref 149). The sulfide-matrix interfaces are sites for the accumulation of diffusionmobile hydrogen atoms, which at subsequent
stages cause hydrogen blistering in low-alloy
ferritic-pearlitic steels due to hydrogen molecules formation from its ions and greater (than
the yield strength) planar pressure. Under loading conditions the hydrogen also will accumulate
at the sulfide-matrix boundaries and other sites
where there is a high stress level and will cause
sulfide stress cracking, which is characteristic
both to mild and alloy high-strength steels (Ref
150, 151). Shape modification of sulfide inclusions from plate-like to globular lowers the sensitivity of steels to hydrogen cracking, since the
globules are less favorable to hydrogen accumulation than plates are. Calcium and rare-earth
metals are suitable for shape modification of inclusions (Ref 152).
Thus, the nonmetallic inclusions, especially
sulfides, enhance accumulation of hydrogen in
steels and increase the susceptibility of steels to
hydrogen embrittlement. Therefore, in order to
lower the sensitivity of steels to hydrogen embrittlement, it is necessary to decrease the sulfur
content in them and to modify the shape and type
of sulfide inclusions.
Sulfur, therefore, is a harmful impurity in
steels, lowering their resistance to fracture and
hydrogen embrittlement. In mild steels with ferritic-pearlitic structure, the sulfur, forming manganese sulfides of plate-like shape (stringers),
enhances hydrogen blistering, which is manifested even at 0.001% S. Sulfide (hydrogen)
stress cracking of such steels is sharply reduced
with the decrease of sulfur content below
0.007%. An effective way for prevention of blistering and hydrogen stress cracking is modification of engineering steels with rare-earth elements (0.10.3%), especially with cerium
spheroidizing sulfide inclusions.
Phosphorus, Antimony, and Tin. The adverse
effect of these elements on steel properties is
well studied. Impurities enriching grain boundaries during tempering cause reversible temper
brittleness. This phenomenon has a considerable
effect on the level of steel resistance to fracture.
Phosphorus in engineering steels greatly lowers brittle and ductile fracture. Phosphorus facilitates steel hydrogen embrittlement by enhancing temper brittleness development and
establishing chemical and structural heterogeneity.
Quantitative relationships have been found for
the embrittling effect of phosphorus, tin, and antimony on steel: Each 0.01% of the impurities
increases brittle fracture temperature, T50, by 20
to 23 C; (3641 F), decreases crack-development energy, Ap, by 10 J/cm2; and increases
plasticity loss value during hydrogenation by
8%.
Reference 153 describes a study on the influence of phosphorus content on the resistance of

heat-hardenable steel (0.4% C, 1.0% Cr, 0.37%


Mo, Fe-bal) to hydrogen-sulfide cracking, and it
has been shown that increasing its content from
0.009 to 0.06% lowers the threshold stress from
518 MPa (75 ksi) to 140 MPa (20 ksi). Such an
effect of phosphorus is attributed, first, to the fact
that it activates the hydrogen absorption process
and, second, that it changes the shape and distribution of carbides. Phosphorus or manganese
segregation can be formed at a pearlite band in
low-alloy steels (Ref 154). Such formation
causes structural changes, an increase of steel
occlusive capacity by 2 times, and a decrease of
the time for failure under load in a hydrogenating
medium by 2 times.
The adverse effect of phosphorus on steel resistance to brittle fracture is comparable with the
favorable effect of nickel and molybdenum:
Each 0.01% lowering of phosphorus content in
steel is equivalent, in its influence on the T50
level, to steel alloying with 1% nickel or 0.1%
of molybdenum.
Antimony and Tin. The resistance of engineering steels to brittle fracture depends, to a
large extent, on its cleanliness with regard to antimony and tin impurities. These impurities are
brought into steel in considerable quantities
(0.010.03%) during melting when using scrap.
Both elements increase the cold-shortness
threshold and lower the impact strength level
(Ref 155). The degree of its influence may vary
with the steel alloying system to a great extent.
In low-alloy steels with ferritic-pearlitic structure, the addition of antimony, which is a more
electro-positive noble element than iron, improves the corrosion resistance in hydrochloric
and sulfuric acids, assumed to be due to a film
containing antimony forming on the steel surface. The surface layer inhibits penetration of hydrogen into steel (Ref 156).
To a greater extent, antimony and tin lower
the steel resistance to brittle fracture: Each additional 0.01% Sb or Sn raises the T50 by an average of 18 C (32 F). Their adverse effect in
this respect is comparable with the favorable influence of nickel and molybdenum.
It is possible to assume that at an antimony
content above 0.3%, the effect of lowering steel
resistance to brittle and ductile fracture will prevail over the effect of lowering the quantity of
absorbed hydrogen. Hence, additional alloying
with approximately 0.20 to 0.30% antimony is
advantageous for low-alloy steels to be used in
hydrogen-sulfide media. The associated reduction in the level of toughness properties is insignificant as compared with the improvement in
the resistance to hydrogen embrittlement by 2
times.
With respect to these considerations the following conclusions can be drawn:
Antimony and tin impurities lower the resistance of structural steels to brittle and ductile
fracture: Each 0.01% increase in antimony or
tin raises T50 of heat-hardenable steels by 18
C (32 F), and each 0.01% increase in antimony raises T50 of low-alloy steels by 1 to 5
C (1.89 F).

84 / Effect of Materials and Processing


The embrittling effect of antimony and tin is

due to the enhancement of recurring temper


brittleness process: Addition of 0.0015% Sb
or Sn increases the T50 of heat-hardenable
steels by 100 C (180 F).
Antimony additions (up to 0.30.5%) in lowalloy ferritic-pearlitic steels retard electrochemical reactions on a steel surface, reducing in the amount of absorbed hydrogen.
Consequently, these additions increase (by
more than 2 times) the resistance of steel to
SCC.
REFERENCES
1. H. Epstein, J. Chipman, and N.J. Grant,
Met. Trans., Vol 9 (No. 4), 1957, p 597
2. N.N. Kolgatin, Zavodskaya Laboratoriya,
Vol 9, 1957, p 1098 (in Russian)
3. T.C. Poulter and L. Uffelman, Physics, Vol
3, 1932, p 147
4. W. Geller and T.H. Sun, Arch. Eisenhuttenwes, Vol 21 (No. 11/12), 1950, p 423
5. E. Martin, Arch. Eisenhuttenwes, Vol 3
(No. 6), 19291930, p 403
6. A. Sieverts et al., Z. Phys. Chem., Vol 183
(No. 1), 1938, p 19
7. L. Luckemeyer-Hasse and H. Schenk,
Arch. Eisenhuttenwes, Vol 5 (No. 6),
19321933, p 209
8. M.J. Armbruster, Am. Chem. Soc., Vol 65
(No. 11), 1943, p 1043
9. W. Eichenbauer, H. Kunzig, and A. Pebler,
Z. Metallkd., Vol 49, 1958, p 220225
10. M.Z. Hill and E.W. Johnson, Trans. AIME,
Vol 221, 1961, p 622629
11. Th. Heumann, D. Primas, and Z. Naturforschung, Vol A-212, 1966, p 260269
12. V.I. Saly, R.A. Riabov, and P.V. Geld, Fiz.
Met. Metalloved., Vol 35 (No. 1), 1973, p
119122 (in Russian)
13. H. Schenk and K.W. Lang, Arch. Eisenhuttenwes, Vol 37, 1966, p 739744
14. J.R.G. Da Silva, S.W. Stafford, and R.B.
McLellan, J. Less-Common Met., Vol 49,
1976, p 407420
15. J.R.G. Da Silva and R.B. McLellan, J.
Less-Common Met., Vol 50, 1976, p 15
16. C.A. Zapffe, Metals Handbook, American
Society for Metals, 1948, p 1208
17. J. Hewitt, Hydrogen in Steel, London
Chapman, Ed., 1962
18. T.N. Stross and F.C. Tomkins, Chem.
Abstr., Vol 50, 1956, p 230
19. W. Eichenauer et al., Z. Metallkd., Vol 49
(No. 5), 1958, p 220
20. E.W. Johnson and M.L. Hill, J. Met., Vol
11 (No. 9), 1959, p 572
21. A.E. Schuetz and W.D. Robertson, Corrosion, Vol 13 (No.7), 1957, p 437
22. R.C. Frank, J. Appl. Phys., Vol 29 (No. 6),
1958, p 892
23. M. Smialowski, Neue Hutte, Vol 10, 1957,
p 621
24. J. Hobson, J. Iron Steel Inst., Vol 189 (No.
4), 1958, p 315

25. H.I. Read, Hydrogen Embrittlement in


Met. Finish., Reinhold, N.Y., 1961 p 213
26. F. Kazinczy, Jernkontorets Ann., Vol 1,
1955, p 885
27. C.A. Zapffe, J. Iron Steel Inst., Vol 156,
1947, p 123
28. B.A. Kolachev, Hydrogen Embrittlement
of Metals, Metallurgia, Moscow, 1985, p
217 (in Russian)
29. V.I. Shapovalov, Hydrogen Influence on
the Structure and Properties of Iron-Carbon Alloys, Metallurgia, Moscow, 1982, p
232 (in Russian)
30. E. Lunarska, A. Zielinski, and M. Smialowski, Acta Metall., Vol 25 (No. 1),
1977, p 161171
31. J.P. Hirth and H.H. Johnson, Corrosion,
Vol 32 (No. 1), 1976, p 326
32. U.I. Archakov, Hydrogen Resistance of
Steel, Metallurgia, Moscow, 1979, p 152
(in Russian)
33. G.V. Karpenko and R.I. Kripyakevich, Hydrogen Influence on Structure and Properties of Steels, Metallurgia, Moscow,
1962, p 198 (in Russian)
34. L.S. Moroz and B.B. Chechulin, Hydrogen
Brittleness of Metals, Metallurgia, Moscow, 1967, p 255 (in Russian)
35. Hydrogen in Metals, Proc. 2nd International Congress, 1977 (Paris), Pergamon
Press, Oxford, 1978
36. C.J. McMahon and V. Vitek, Acta Metall.,Vol 27 (No. 4), 1979, p 507513
37. M. Cornet and S. Talbot-Besnard, Metall.
Sci., Vol 7, 1978, p 335
38. J.P. Hirth, Metall. Trans. A, Vol 11 (No.
6), 1980, p 861890
39. G.M. Pressouyre, Metall. Trans. A, Vol 10
(No. 10), 1979, p 15711973
40. A.G. Kumnick and H.H. Johnson, Acta
Metall.,Vol 28 (No. 1), 1980, p 3339
41. C.D. Beachem, Metall. Trans., Vol 3,
1972, p 437
42. A. Troiano, Met. Prog., Vol 71, 1960, p
112
43. R. Oriani and P. Josephic: Acta Metall.,
Vol 22, 1974, p 1065
44. L. Engel and H. Klingel, Rasterelectronenmikroskopishe Untersuchungen von Metallschaden,
Karl
Hanser
Verlag,
Munchen-Wien, 1982
45. J.M. Hodge, M.A. Orehoski, and J.E.
Steiner, Trans. AIME, Vol 230 (No. 5),
1964, p 1182
46. J. Mussati and A. Reggiori, Met. Prog.,
Vol 53 (No. 7), 1936, p 51
47. P. Bardenheuer, Stahl u. Eisen, Vol 45,
1925, p 1782
48. W. Eilender and H.Z. Kiessel, VDI, 1932,
Vol 76, p 729
49. E.A. Klausting, Metallurgist, Vol 5, 1937,
p 21
50. E. Houdremont, Handbuch der Sonderstahlkunde, Verlag Springer, Berlin, 1934
51. V.V. Zabilsky et al., Fiz. Met. Metalloved.,
Vol 83 (No. 2), 1997, p 160175 (in Russian)

52. Y.M. Potak, Brittle Fractures of Steel and


Steel Parts, Oborongiz, Moscow, 1955 (in
Russian)
53. K. Oniki, Kinzoku, Vol 46 (No. 4), 1976,
p 1518
54. M.A. Balter, Fractography as Tool for Diagnosis of Fracted Parts, Maschinostroenie, Moscow, 1987 (in Russian)
55. Y. Nakatani, K. Utsubo, and K. Yamada,
Trans. Jpn. Soc. Mech. Eng. A, Vol 64 (No.
626), 1998, p 25612566
56. M.M. Shved, Change of Operating Properties of Iron and Steel under Hydrogen
Action, Naukova dumka, Kiev, 1985 (in
Russian)
57. A.Ye. Andreykiv, V.V. Panacjuk, and V.S.
Charin, Fiz. Met. Metalloved., (No. 3),
1978, p 323 (in Russian)
58. E.A. Steigerwald, Trans. AIME, Vol 215,
1959, p 1048
59. L.M. Bilyi and M.M. Shved, Fiz. Met. Metalloved., (No. 6), 1978, p 106107 (in
Russian)
60. I.S. Yablonsky, Fiz. Khim. Mekh. Mater.,
(No. 6), 1979, p 4756 (in Russian)
61. F.F. Azhogin, A.V. Sacharov, and S.S. Ivanov, Fiz. Khim. Mekh. Mater., (No. 3),
1979, p 3538 (in Russian)
62. K. Soeno, Trans. Iron Steel Inst. Jpn., Vol
21 (No. 7), 1981, p 469476
63. K. Soeno and K. Taguchi, Trans. Iron Steel
Inst. Jpn., Vol 21 (No. 9), 1981, p 618
623
64. K. Soeno and M. Tsuhiya, Trans. Iron
Steel Inst. Jpn., Vol 22 (No. 11), 1982, p
848853
65. V.A. Maritchev, Fiz. Khim. Mekh. Mater.,
(No. 5), 1981, p 2429 (in Russian)
66. S. Yamasaki and T. Takahashi, J. Iron
Steel Inst. Jpn., Vol 83 (No. 7), 1997, p
460465
67. O. Tamura and M. Takemoto, J. Iron Steel
Inst. Jpn., Vol 83 (No. 8), 1997, p 532
537
68. C.L. Briant and S.K. Banerji, Ed., Treatise
on Materials Science and Technology,Vol
25, Embrittlement of Engineering Alloys,
Academic Press, 1983
69. G.G. Hancock and H.H. Johnson, Trans.
Met. Soc. AIME, Vol 236, 1966, p 513
70. H. Liebowitz, Ed., Fracture, Vol 3, Mir,
Moscow, 1971 (in Russian)
71. V.S. Savchenko and K.A. Yutshenko, Fiz.
Khim. Mekh. Mater., (No. 1), 1982, p 54
57 (in Russian)
72. V.N. Zikeev, MiTOM, (No. 5), 1982, p 8
23 (in Russian)
73. G.G. Irzhov, S.A. Golovanenko, and T.K.
Sergeeva, Fiz. Khim. Mekh. Mater., (No.
3), 1982, p 8993 (in Russian)
74. S.A. Golovanenko and T.K. Sergeeva,
Steel, (No. 7), 1984, p 7378 (in Russian)
75. L.M. Utevsky, Ye.E. Glikman, and G.S.
Kark, Reversible Temper Brittleness of
Steel and Iron Alloys, Metallurgia, Moscow, 1987 (in Russian)

Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 85


76. A.I. Kovalev, G.V. Tsherbedinsky, Modern Methods of Investigations of Metal and
Alloy Surface, Metallurgia, Moscow,
1989 (in Russian)
77. V.V. Zabilsky, Metal. Termoobrabotka,
(No. 1), 1987, p 2432 (in Russian)
78. L.I. Gladshtein and N.I. Kelberin, Zavodskaya Laboratoriya, (No. 1), 1995, p 38
41 (in Russian)
79. A.Yu. Kazanskaya, M.A. Smirnov, and
V.V. Zabilsky, Fiz. Met. Metalloved., (No.
8), 1990, p 201203 (in Russian)
80. M.L. Bernshtein and V.A. Zaymovsky,
Mechanical Properties of Metals, Metallurgia, Moscow, 1979 (in Russian)
81. V.V. Zabilsky, O.A. Bartenev, V.V. Velitchko, and S.M. Polonskaja, Fiz. Met.
Metalloved., (No. 4), 1986, p 793800 (in
Russian)
82. V.V. Velitchko, G.M. Micheev, V.V. Zabilsky, and D.I. Maleev, Fiz. Khim. Mekh.
Mater., Vol 27 (No. 1), 1991, p 112114
(in Russian)
83. V.V. Zabilsky and V.V. Velitchko, Metal.
Termoobrabotka, (No. 4), 1993, p 915 (in
Russian)
84. V.V. Velitchko, V.V. Zabilsky, and G.M.
Micheev, Fiz. Met. Metalloved., (No. 1),
1995, p 130140 (in Russian)
85. I.S. Toydorova and V.V. Zabilsky, Fiz.
Met. Metalloved., (No. 7), 1991, p 511
(in Russian)
86. V.V. Zabilsky, V.V. Velitchko, and S.G.
Iliyina, Fiz. Met. Metalloved., Vol 80 (No.
12), 1995, p 108118 (in Russian)
87. V.V. Zabilsky, V.V. Velitchko, F.Z. Gilmutdinov, V.G. Makletzov, and Yu.P. Vorobiov, Fiz. Met. Metalloved., Vol 83 (No.
2), 1997, p 160175 (in Russian)
88. A.V. Vaynman, R.K. Melechov, and O.D.
Smiyan, Hydrogen Embrittlement of Elements of High Pressure Kettles, Naukova
dumka, Kiev, 1990 (in Russian)
89. I.K. Pohodnya, Electron and Ion Spectroscopy of Solids (No. 2), L. Fiermans, J.
Vennik, and W. Dekeyser, Plenum Press,
1988, p 3031 (in Russian)
90. V.V. Zabilsky and V.V. Velitchko, Fiz.
Khim. Mekh. Mater., Vol 25 (No. 3), 1989,
p 1721 (in Russian)
91. V.V. Zabilsky, V.V. Velitchko, and A.Yu.
Kazanskaja, Fiz. Met. Metalloved., (No. 3),
1990, p 143149 (in Russian)
92. V.I. Voronenko, Zashch. Met., Vol 33 (No.
2), 1997, p 132143 (in Russian)
93. O.N. Romaniv and G.N. Nikifortchin, Mechanics of Corrosion Failure of Structural
Alloys, Naukova dumka, Kiev, 1986
94. N.I. Volgina and T.K. Sergeeva, Materialovedenije, (No. 12), 1998, p 2130 (in
Russian)
95. L.N. Petrov, V.N. Tischenko, and A.S.
Tchistov, Fiz. Khim. Mekh. Mater., (No.
4), 1979, p 2025 (in Russian)
96. V.I. Nikitin, Fiz. Khim. Mekh. Mater., Vol
25 (No. 1), 1989, p 31 (in Russian)
97. F.R. Ford, Mech. Behav. Mater. Proc. 3rd
Int. Conf., 1979 (Cambridge U.K.), Pergamon Press, Toronto, 1979, (No. 2), p 431

98. O.N. Romaniv, G.N. Nikifortchin, and


A.Z. Student, Fiz. Khim. Mekh. Mater.,
Vol 19 (No. 5), 1983, p 3 (in Russian)
99. V.P. Voronin, A.Yu. Schulte, and Yu.V.
Tomkin, Fiz. Khim. Mekh. Mater., Vol 19
(No. 5), 1983, p 114 (in Russian)
100. A. Mazur and B. Pawlowski, Corros. Sci.,
Vol 26 (No. 1), 1986, p 7
101. T.J. Kaneko, J. Iron Steel Inst. Jpn., Vol
72 (No. 13), 1986 p 1472
102. J. Kuniya, I. Masaoka, and R. Sasaki, J.
Pressure Vessel Technol., Vol 107 (No. 4),
1985, p 430
103. W. Diton and P. Huiying, Proc. Int. Congr.
Met. Corros., 1984 (Toronto), (No. 3),
1984, p 573577
104. L.N. Petrovm, Fiz. Khim. Mekh. Mater.,
(No. 4), 1984, p 1621
105. W.W. Gerberich, R.N. Jones, M.A. Friesel,
and A. Nozue, Mater. Sci. and Eng. A, Vol
103 (No. 1), 1988, p 185
106. A. Nozue, T. Kishi, and R. Horiuchi, J.
Iron Steel Inst. Jpn., Vol 67 (No. 10),
1981, p 1787
107. G.B. Muravin, L.M. Lezvinscaja, N.O.
Makarova, and G.S. Pavlovskaja, Defektoskopiya, (No. 2), 1990, p 1828
108. B.I. Voronenko, Zashch. Met., Vol 33 (No.
5), 1997, p 472488 (in Russian)
109. B.I. Voronenko, Metal. Termoobrabotka,
(No. 9), 1990, p 58 (in Russian)
110. C.S. Carter, Corrosion, Vol 25 (No. 10),
1969, p 423
111. B.I. Voronenko, Zashch. Met., Vol 33 (No.
6), 1997, p 573589 (in Russian)
112. P. McIntyre and T.M. Banks, Eur. Offshore Steels Res. Seminar, VIP 19/1, 1978
(Cambridge), Abington Publishing, 1980
113. D. Huerta and K.E. Heusler, Werkst. Korros., Vol 45 (No. 9), 1994, p 489
114. O.N. Romaniv, G.N. Nikifortchin, A.Z.
Student, Fiz. Khim. Mekh. Mater., (No. 2),
1985, p 2031 (in Russian)
115. B. Pawlowski, Werkst. Korros., Vol 37
(No. 8), 1986, p 448
116. R.O. Ritchie, M.H.C. Cenedo, and V.F.
Zackay, Metall. Trans. A, Vol 9 (No. 1),
1978, p 3540
117. M. Turdy, V. Suchanek, L. Hyspecka, and
K. Mazanec, Adv. Res. Strength and Fract.
Mater./4th Int. Conf. Fract. 1977 (Waterloo, NY), 1978, p 255
118. D. Zhou, W. Cao, and H. Guo, Heat Treat.
Met., (No. 1), 1992, p 8
119. Y. Hirose, K. Tanaka, and K. Okabayashi,
22nd Congr. Mater. Res., 1978 (Kyoto, Japan), Proc. Kyoto, 1979, p 76
120. Y. Hirose and K. Tanaka, Met. Corros./8th
Int. Congr./7th Congr. Eur. Fed. Corros.
(111th event), 1981 (Mainz, Germany),
Frankfurt/M., European Federation of
Corrosion, 1981, (No. 1), p 553
121. O.N. Romaniv, G.N. Nikifortchin, and
A.Z. Student, Fiz. Khim. Mekh. Mater.,
Vol 17 (No. 3), 1981, p 24 (in Russian)
122. W. Cao and X. Lu, Mater. Sci. and Eng.
A, Vol 110 (No. 151), 1989, p 165

123. W. Radeker and B.N. Mishra, Werkst. Korros., Vol 21 (No. 9), 1970, p 692
124. F.W. Fraser and E.A. Metzbower, Fractogr. and Mater. Sci. Symp., 1979 (Williamsburg, VA), Philadelphia, 1981, p 51
125. K. Rippstein and H. Kaesche, Corros. Sci.,
Vol 29 (No. 5), 1989, p 517
126. Z. Szklarska-Smialowska and E. Lunarska,
Werkst. Korros., Vol 32 (No. 11), 1981, p
478
127. H. Hanninen, Corrosion, Vol 41 (No. 7),
1990, p 563
128. A.G. Gareev, I.G. Abdullin, and G.I. Abdullina, Gazovaja Promischlennost, (No.
11), 1993, p 29 (in Russian)
129. A.F. Matvienko, A.V. Baldin, A.P. Grigorjev, V.A. Kanaikin, V.G. Rjibalko,
V.V. Sagaradze, and Yu.I. Philippov, Fiz.
Khim. Mekh. Mater., Vol 86 (No. 2), 1998,
p 139146 (in Russian)
130. R.N. Parkins, Corrosion, (No. 50), 1994,
p 384
131. T.K. Sergeeva, E.P. Turkovskaja, N.P. Michailov, and A.I. Tchistjakov, State of
Problem of Stress-Corrosion in SNG
Countries and Abroad, IRZ RAO, Gazprom, Moscow, 1998 (in Russian)
132. R.N. Parkins, C.S. Dell, and R.R. Fessler,
Corros. Sci., Vol 24 (No. 4), 1984, p 343
374
133. Public Inquiry Concerning Stress Corrosion Cracking on Canadian Oil and Gas
Pipelines, Report NEB, 1996, No. 11, MH295
134. M. Urednicek, S. Lambert, and O. Vosikovsky, Monitoring and Control Proc. Int.
Conf. on Pipeline Reliability (Calgary,
Canada), 1992, N VII-2
135. T.K. Sergeeva, Proc. Int. Symp. on Pipe
Line Safety and Aging, CNIICHERMET,
1995 (Moscow), Vol 1, p 139164
136. O.M. Ivantzov, Stroitelstvo Truboprovodov, No. 6, 1993, p 1016 (in Russian)
137. B.G. Mazel, Stroitelstvo Truboprovodov,
No. 7, 1993, p 3639 (in Russian)
138. D. Rudy, Prev. of Pipeline Corrosion
Conf., 1994 (Houston, TX), p 1721
139. V.V. Zabilsky and V.I. Lebedev, Tchernaja Metallurgija. Bjulleten NTI, No. 2,
1991, p 1329 (in Russian)
140. R.M. Nikonova and V.V. Zabilsky, Fiz.
Metal. Metalloved., Vol 76 (No. 8), 1993,
p 99106 (in Russian)
141. A.F. Matvienko, A.V. Baldin, A.P. Grigorjev, V.A. Kanaikin, V.G. Rjibalko,
V.V. Sagaradze, and Yu.I. Philippov, Fiz.
Khim. Mekh. Mater., Vol 86 (No. 2), 1998,
p 147155 (in Russian)
142. G. Libowitz, The Solid-State Chemistry of
Binary Metal Hydrides, Vol 6, 1965, p 11
143. V.N. Zikeev, Metal. Termoobrabotka, Vol
5, 1982, p 1823 (in Russian)
144. V.I. Shapovalov, Hydrogen Influence on
the Structure and Properties of Iron-Carbon Alloys, Metallurgia, Moscow, 1982, p
232 (in Russian)
145. R. Oriani, Acta Metall., Vol 24, 1977, p
979988

86 / Effect of Materials and Processing


146. M. Vornet, Mem. Sci. Rev. Met., Vol 74
(No. 5), 1977, p 307316
147. A.P. Guliaev, Clean Steel, Metallurgia,
Moscow, 1975, p 184 (in Russian)
148. V.G. Starchak, A.B. Kuslitsky, U.A.
Kliachko, and I.A. Tamarina, Metal. Termoobrabotka, No. 4, 1977, p 1113 (in
Russian)
149. A. Joshi, Corrosion, Vol 34 (No. 2), 1978,
p 4752

150. J. Hirth, Corrosion, Vol 32 (No. 1), 1976,


p 3437
151. Z. Szklarska-Smilowaska, Werkst. Korros., Vol 32 (No. 11), 1982, p 478485
152. G.M. Itskovich, Deoxidation of Steel and
Modification of Nonmetallic Inclusions,
Metallurgia, Moscow, 1981, p 296 (in
Russian)
153. E. Snape, Corrosion, Vol 24 (No. 9), 1968,
p 261282

154. H. Gondo and M. Iino, Nippon Steel Corporation Studies on Materials for Steel
Pipes for Service in Wet Sour Gas Medium, Technical Report B2 (No. 11), Nippon Steel Corporation, 1979, p 71
155. E. Stephenson, Metall. Trans. A, Vol 11
(No. 3), 1980, p 517524
156. S. Shichi, M. Takeshi, and D. Utaka, Japanese Application kl 12A41 (C23Fg/02),
No. 54, 1979

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p89-98
DOI: 10.1361/hrsd2002p089

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Deflection Methods to Estimate


Residual Stress
H.W. Walton, Consultant

NOT EVERYONE has sophisticated instrumentation such as x-ray or neutron diffraction


readily available for the determination of residual stresses. Nor do they have the time or inclination to use these procedures for checking production parts. Mathematical derivations such as
the one for plate by the Treuting and Read
method are also very time consuming (the
method is described at the end of this article).
To determine if residual stresses from prior
processing are the cause of uncontrollable distortions during processing of thin section parts,
it is a simple matter to slit an as-received piece
and note the amount of deflection resulting from
relaxation of residual stresses, for example, the
net opening of the saw cut. If no deflection occurs, then it can be safely assumed that the problem lies with the processing parameters and not
with the material (e.g., high feeds and speeds or
high chucking pressures during machining, or
surface overheating during grinding resulting
from insufficient wheel dressing).
These simple techniques, sometimes known
as dissection, are old but still very useful. However, because slitting a component is a destructive method, there are limitations, particularly if
the component in question is large.
This article provides a low-cost, easy method
of determining if residual stresses are the cause
of component distortion during manufacture.

Residual Stress
To understand the type of stress that is being
measured, a distinction first needs to be made
between microstresses and macrostresses extending over large volumes. Microstresses resulting from changes occurring at the atomic
level are equilibrated over very small volumes
and cannot be measured by use of deflection
methods. Macrostresses, on the other hand, are
primarily the result of forming operations or
thermal imbalances during casting, welding, heat
treating, and so on.
If the stress induced by nonuniform cooling
exceeds the yield strength of the material, plastic
deformation occurs. On cooling to a uniform am-

bient temperature, the deformed volume being


restrained by the surrounding material is now
placed under significant elastic stress. This
would be considered macrostress and is the
stress most likely to cause deformation during
subsequent processing.
Consider the simple example of bending a bar
to form a permanent set. The surface material at
the elbow or concave surface undergoes permanent plastic deformation while the inside diameter is undergoing compressive plastic deformation. When the applied force is removed, the
underlying elastically stretched material (at the
elbow) is now prevented from relaxing completely. The result is that the plastically deformed surface layers of the elbow are placed
under elastic compression by the underlying
elastically stretched material. The opposite is
true for the inside diameter. It follows that if the
bar is placed in service in this condition, corrosion and development of cracks are more likely
to occur at the surface of the inside diameter.
In general terms, if a surface undergoes permanent plastic deformation by cold rolling, shot
peening, machining, thermal processing, and so
on, which in turn results in the underlying material being elastically stretched or compressed,
the material plastically deformed by tensile
forces is left in elastic compression. Material
plastically deformed by compressive forces is in
elastic tension.
The influence of nonuniform residual stress is
sometimes dramatically illustrated by fishtailing of large hot-rolled steel plate. As the plate
cools after rolling, transformational stresses are
developed and, if an extensive central lamination
is present due to inadequate discard or casting
problems, these stresses can cause the plate to
separate into two halves (Fig. 1).
The stress distributions present before and after the occurrence of a fishtail are shown in the
sketch. In the as-rolled plate, tensile stresses
(tv stress) on the surface are trying to contract
the surface while compressive stresses (tv
stress) are trying to expand the plate core. How
this stress distribution develops is illustrated in
Fig. 2. After fishtailing, the two halves of the
plate attain a new equilibrium stress distribution.

While air cooling after hot rolling, the plate


surface will be cooling faster than the core material, and thermal stresses developed at this
stage will be relaxed by the plastic deformation
of the low strength austenite. At lower temperatures, transformation to ferrite and pearlite begins at the surface (Fig. 3), causing complex interactions and a variety of stress conditions.
Because of the relatively low yield strength of
the various phases and the role of transformation
induced plasticity, the final stresses are low.
However, over a large surface area even low residual stresses can have a profound effect, particularly in the presence of large defects such as
a lamination.
If not fixtured in a vertical position, severe
distortion of quenched flat steel sections may occur. This is due to differences in the cooling rate
and transformation behavior between the top and
bottom surface.
In a carburized product, because of the compositional changes causing transformation to
progress from the inside out and the considerably
higher-strength case, the final residual stress profile is reversed and a beneficial high-compressive
surface stress is developed (Ref 1). Figure 4
shows the development of these residual stresses
during the heat treatment of a carburized component.
It is very important to avoid heating small,
isolated areas on the surface of hardened steel.

Fig. 1
stress

Schematic representation of fishtailing in plate


steels. tv, tensile stress; tv, compressive

90 / Measurement and Prediction of Residual Stress and Distortion


Examples of where this may occur are:
Components ground with inadequate cooling

fluid or faulty equipment (vibrating wheel


heads)
Using a gas flame to enlarge a hardened component for assembly on a shaft
Electrical arcing from inadequately earthed
drive motors
Accidental arcing during magnetization for
magnetic particle inspection
The thermally damaged area is restrained from
expanding by the surrounding material and becomes plastically deformed under compression.
Consequently, on cooling to ambient temperature, the surface is left in a tensile condition, and
cracking ensues either immediately or after the
component is placed in service (more worrying).
Heating at temperature above the original tempering temperature may also lead to a contraction.
Shot peening appears to be an anomaly. At
first sight, the impact of the shot particles seems
to cause plastic deformation in compression. It
should follow that the surface is in tension following the peening treatment. However, the ad-

ditive influence of the multiple small areas of


plastic extension of the surface must be considered. The overall surface undergoes a plastic extension, thus putting the underlying layers in tension. Consequently, on completion of the
shot-peening operation, the surface is restrained
in a beneficial compressive state.
The maximum level of residual stress that can
be generated is approximately equal to the elastic
limit or yield strength of the material measured
under the same stress condition and for the same
cross section. The existence of triaxial stresses
in heavy section components has a major influence on the ease of plastic flow (plane-strain conditions). If surrounding material inhibits plastic
flow by significantly limiting shear stresses, the
levels of residual tensile stress may become
high. Catastrophic fracture occurs when the tensile stress level exceeds the cohesive strength of
the steel.

Manufacturing Implications of
Residual Stresses
There are many examples that show how relaxation or redistribution of residual stresses

contributes to high processing costs. One example is the need to use multiple grinding passes
to avoid distortion in thin section components.
A frequent problem often wrongly attributed to
the heat treatment department is heat treatment
distortion from prior residual stresses in the raw
material.
Premature failure due to harmful residual surface tensile stress in the manufactured product is
to be avoided at all costs.
On the other hand, the presence of residual
stresses may be advantageous, as in the example
of compressive stresses from case hardening and
shot peening.
Using residual stresses to intentionally deform
a component can be illustrated by the use of shot
peening to shape the aircraft fuselage, by using
a peen hammer to straighten a long shaft, or by
simply hot spotting to bend a simple bar of steel.
For example, a small area on one side of a
steel bar is heated rapidly to a temperature, T, of
approximately 600 C (1100 F) (Ref 2). If free
to expand, the heated disc-shaped region increases in length, l, by:
l T
14 600

Cooler outside
surface

8.4 103 lm/mm

Hotter core

Transformation
Consider a slice through half the thickness of a plate.
At the onset of the cooling, both the surface and
the core are austenitic and stress free.

Transformed

Not transformed
The surface will transform first, acccompanied by
an expansion. Restraint by the underlying material
will place the transformed element in compression "C".
Although the adjoining austenitic element "stretches"
and plastically yields, it will be in tension "T".

CT

where is the coefficient of thermal expansion


in lm/mm K. However, the surrounding cooler
material restrains expansion. Because of the relatively low yield strength at 600 C (YS600) of
approximately 30 MPa (4.5 ksi), the heated steel
deforms plastically under compression, increasing the thickness of the heated disc. On cooling,
the disc contracts the same amount and, assuming no stress relaxation occurred, a tensile stress,
r, is now developed in the disc:
r (E l) YS600
(200 103 8.4 103) 30
1650 MPa (240 ksi)

Schematic representation of transformational


expansion in the absence of any restraining effect
by the surrounding material

The already extended underlying element then


undergoes transformation and expands further. This
induces tensile stresses in the material on either side,
but due to the restraining influence of surrounding
material, compressive stresses are induced in the
newly transformed steel.

TCT

Fig. 2

TCT

The transformation front proceeds deeper into the


section, leaving behind a surface that is in tensile.
The magnitude of stress is determined by the relative
yield strength of the phases. Because of the TRIP
effect (transformation induced plasticity), the plasticity
of the transforming steel is considerably higher.

The development of residual stresses during cooling of steels

where E is the modulus of elasticity (MPa). The


stress in the disc cannot be greater than the yield
strength of about 500 MPa (73 ksi). Therefore,
either the disc stretches plastically, or the bar
bends elastically to balance the stresses (or
both).
The manufacture of seamless high-carbonbearing steel tube involves hot rolling, sizing,
straightening, and spheroidize annealing. Each
stage of manufacture introduces various degrees
of residual stress. If residual stress levels are
higher than normal, difficulties may be experienced during subsequent machining and heat
treatment of the rings. As seen from the previous
calculation, the higher the yield strength of the
material, the greater the potential for high levels
of residual stress. As thermal and mechanical
treatment of steel increases the strength of a material, potential problems with residual stress
also increase. For example, a small area of thermal damage on the surface of a hardened high-

Deflection Methods to Estimate Residual Stress / 91


carbon steel (e.g., electric arc discharge) may result in a level of triaxial residual stress close to
the cohesive strength of the material, possibly
leading to immediate or delayed cracking.
Although the stress may be multidirectional,
hoop stress that arises at several stages of manufacture is the primary cause of many of the
problems. When a length of tube is parted off
and slit in a longitudinal direction, any hoop
stress present tends to open the slit (Fig. 5).
Opening of the slit indicates compressive stress
in the inside diameter (ID) of the tube and tensile
stress on the outside diameter (OD). On rare occasions, the slit may close, indicating the reverse
condition. The stress is not always uniform along
the length of the tube. Stress measurements have,
on occasions, indicated a cyclic variation corresponding to differences in cooling of areas in
contact with the cooling bed cross ties.
Shot peening intensity is monitored by using
a method developed by J.O. Almen of General
Motors Company. In this method, 75 mm long
by 18.75 mm wide (3.00 by 0.75 in.) strips of
1070 spring steel are quenched and tempered to
a deep-blue oxide finish (blue tempered) for a
hardness of 45 to 50 Rc and exposed on one side
to the same shot intensity as the component undergoing treatment (Ref 3). Three different
thicknesses are used: N, 0.79 mm (0.031 in.);
A, 1.30 mm (0.051in.); and C, 2.38 mm
(0.0938 in.) to allow for differences in degree
and shot peening required and materials being
treated. For example, peen forming of an aluminum alloy requires considerably less peen intensity than a carburized gear tooth. During treatment, the Almen strip assumes a concave shape.
The lift height is proportional to the level of
compressive stress developed in the upper sur-

face during peening. Peening is continued until


the material has reached saturation; that is, when
the lift height increases by no significant amount
and the compressive stress in the surface layers
corresponds closely to the elastic limit of the
steel strip (Fig. 6).
Consider the residual stress distribution
through the Almen strip. Deflection of the strip
occurs in order to relieve some of the high surface compressive stress. The peened surface area
endeavors to expand by bowing in order to nullify the compressive stress. The magnitude of
bowing is limited by the restraint of the remainder of the material, and equilibrium is reached
when the remaining compressive stress is in balance with the elastic compressive stress developed in the lower surface. The depth and intensity of the compressive layer is proportional to
the lift height at saturation. The Almen strips are
primarily used to ensure that the process is in
control (shot condition, uniformity in application, impeller operation, etc.). A photograph of
three typical strips is shown in Fig. 7.
Metal Improvement Company, Inc., in collaboration with ENSAM, a French advanced engineering school, has developed a software program called PeenStress (Metal Improvement
Co., Inc., Paramus, NJ) that is used to assist in
shot peening callouts (Ref 4). The user selects a
material from a library of about 80 materials,
then selects a shot size and shot intensity and
inputs some basic geometry considerations. Figure 8 is a curve generated on a chromium-silicon
spring wire shot peened with a hardened, 0.023
in. shot to a 10 A intensity. A 10 A intensity
equates to a 0.25 mm (0.010 in.) arc height on
the A-strip.
One of the older standard tests for evaluating

Ar3
Ar1
Austenite

quench distortion is the Navy C-ring test (Ref 5


7). The amount of distortion (deflection) of a
quenched test piece is measured by the change
in the gap width (Fig. 9).

Methods for Measuring Residual


Stresses from Deflection Data
Sectioning to allow relaxation of residual
stress in actual components where stresses are
thought to be present can be performed by several methods, from simple saw cutting to the
more sophisticated compliance method (Ref 8
10). In the latter method, the residual stress profile is calculated from the strains caused by introducing a cut of progressively increasing depth
into the component. The strains are measured by
using suitably positioned strain gages cemented
to the surface adjacent to the cut. The cutting is
performed by using various techniques, however, electrical discharge machining is the preferred method. Usually, two computer-based approaches are used to analyze the data. These are
the forward and inverse solution. The forward
solution derives the measurable strains (compliance functions) that develop from introducing a
successively deeper slot into a part containing an
arbitrary known stress distribution. The inverse
solution develops the original residual stress distribution that best matches the actually measured
strains.
Similar methods based on drilling small holes
in the stressed material have been around for
some years. Strain relaxation is measured using
strain gages or photoelastic coatings. Strain
gages have been used to measure strain relaxation in bevel gears following successive removal
of layers by electrochemical machining (Ref 11).
Saw cutting or slitting is an easy quality control test that gives a global overview of the state
of residual bulk hoop stress in rings. Interestingly, such a technique was used recently to validate the measurements of residual stress levels
in railway wheels using electromagneticacoustic transducers (Ref 12).

Ferrite
Pearlite
Temperature

Core
Bainite
Surface

Martensite

Time, T
Schematic representation of the relative transformation at the surface and in the core of the mild steel plate. At
time, T, the core is still austenitic while the surface has already transformed to a ferrite-pearlite structure. Ar3
and Ar1, upper and lower transformation temperatures, respectively, on heating (refroidissant) a hypoeutectoid steel

Fig. 3

Mathematical Derivations for


Simple Cases Based on the Saw-Cut
Methods
Definitions of the symbols used in the following derivations are given in Table 1.
To develop the mathematical algorithms for
interpreting deflection measurements after slitting a simple shape (plate, round bar, or tube),
simple beam theory is used (Fig. 10).
The basic formula for the state of affairs at
any point (x) along a beam is (Ref 13):
M
r
E

I
c
R

(Eq 1)

where M is the bending moment to which the


beam is subjected at x. Bending moment M is

92 / Measurement and Prediction of Residual Stress and Distortion

%C

Carbon profile through case. As the carbon level


decreases, the transformation temperature during
quenching increases (Ms). Although the cooling rate
is less below the surface, transformation occurs from
the inside out.

Consider a section through the carburized case. At


the onset of quench, both the surface and the core
are austenitic and stress free.

Start of transformation
Even though the cooling rate is lower than at the
surface, the low carbon core will begin to transform,
starting first below the case and progressing inward.
The adjoining austenitic case element will plastically
expand outward to accommodate the resulting tensile
stress.

equal to load multiplied by distance (lbf in., or


kgf m). If M is not obvious from the first principles for a given case, it may be obtained from
the data given subsequently. I is the moment of
inertia of the section of the beam at x, usually in
inches. Moments of inertia are treated in textbooks of elementary mechanics. The moment of
inertia in question is that about the neutral axis
which, in most beam problems, will be a line
through the center of gravity of the section. If
the I of the section is not already known, it is
usually easily calculated by applying one of the
following formulae.
For a rectangle:

bt3
12

Schematic representation of transformational


expansion in the absence of any restraining effect
by the surrounding material

C CC T T

+tv

Following the progressive transformation of the core,


the innermost element of the case transforms. The
resulting expansion of the already extended case
element creates tensile stresses in both remaining
austenitic case and adjoining core.

Sections made of rectangles, such as rectangular


tubes, H-beams, and channels, can be calculated
by subtracting the I of the empty areas.
For a circular cross section:

The continued transformation of the progressively


higher carbon case and the resulting expansion of the
already plastically stretched elements result in high
compressive stresses. Consequently, the underlying
transformed core is pulled into tension.

Stress profile through case. Compressive stress


near the surface of the case can reach levels of
300/500 N/mm2 and is one of the primary reasons for
carburizing.

tv

Fig. 4

(Eq 2)

The development of compressive stress in the case of a quenched carburized steel. Source: Ref 1

pr 4
4

(Eq 3)

The moment of inertia for hollow tubes can


be calculated by subtraction, where: r is stress
in the material. c is distance from the neutral
axis. In symmetrical sections this is the distance
from the midpoint or center of gravity of the section. E is Youngs modulus. (In thick cross sections and/or high-strength steels, this should be
corrected for Poisson contraction where m is
Poissons ratio [0.3 for steel].)
E

E
1 m2

(Eq 4)

R is the radius of curvature of the beam when it


bends under load.
To calculate the stress in a simple beam:
r

Mc
I

(Eq 5)

For a rectangular beam at the surface where the


stress is greatest:

Fig. 5

Schematic of the residual stress distribution in rings manufactured from tube before and after slitting. ID, inside
diam; OD, outside diam

bt3
12

(Eq 6)

t
2

(Eq 7)

so

Deflection Methods to Estimate Residual Stress / 93

rmax 6

M
bt2

(Eq 8)

EI
R

Symbols used in the derivations and formulas


Units

This explains why a beam that is twice as thick


is four times as strong.
The distribution of residual stress in actual
components is unlikely to be linear. However,
for the subsequent approximate analysis, the
beam stress is assumed to vary linearly through
the section.
In sheet and bar, as the sheet fishtails or is cut
in a central planar direction, the bending moment
created by the residual stress is released (Fig.
11). The bending moment may be expressed as:
M

Table 1

(Eq 9)

Symbol

Description

M
I

Bending moment
Moment of inertia

r
c

Stress in material
Distance from neutral axis

E
R
L
r
t
d
c

Youngs modulus
Radius of curvature of beam or displaced section
Length of curved beam
Radius of round bar cross section
Thickness of beam/plate/tube wall
Measured deflection
Poissons ratio

SI

English

Notes

NM
m4

lbf in.
in.4

Pa
m

psi
in.

Pa
m
m
m
m
m
Dimensionless

psi
in.
in.
in.
in.
in.
...

...
I bt3/12 for rectangle (where b and
t are the dimensions of the
cross section) I pr 4/4
for circle
...
In a symmetrical section this
is the distance from the midpoint or center of gravity
...
...
...
...
...
...
0.3 for steel

given that:
c

t
4

(Eq 10)
30 0.015 in.

where t is plate thickness, and assuming the distribution of the residual stress that resulted in the
fishtail varies linearly over the half-thickness of
the sheet, the maximum longitudinal stress at the
surface is given by:
r

Mc
I

(Eq 11)

0.0310.001 in.
Measuring dial

0.0510.001 in.

N strip

Peening nozzle

A strip
C strip

Almen strips

0.09380.001 in.
0.7450.750 in.

Shot stream
4-6 in.

Almen test strips

Substituting Eq 9 and 10 in Eq 11 gives:


r

Et
4R

Hardened
ball supports

10/32
screws

(Eq 12)

Arc height
3.0 in.

If the deflection, d, is small compared to the radius of curvature, R, R may be expressed in terms
of the deflection, d, and the length of the curved
surface, L, by:
R

L2
2d

3.0 in.
Holding
fixture

0.75 in.
Strip removed. Residual
stresses induced arching

Fig. 6

Use of Almen strips to monitor the degree of shot peening. Source: Ref 3

Fig. 7

Almen strips used to monitor different levels of shot peening intensity

(Eq 13)

This is derived as in Fig. 12.


Referring to Fig. 12, the two radii O1 and O2
are drawn as shown, and the tangents 1,3 and 2,4
are drawn to these radii. A chord is drawn between the points of tangency 1 and 2 and the line
2,3 is drawn perpendicular to the chord at 2. The
two isosceles triangles RLk and mdn are similar
and therefore:
R
m

L
d

(Eq 14)

or
R

Lm
d

(Eq 15)

For small angles, L is the arc 1,2 and m is 12 the


arc 1, 2, or

Arc height
Strip mounted for
height measurement

94 / Measurement and Prediction of Residual Stress and Distortion

1
L
2

(Eq 16)

therefore (as in Eq 13):


R

L2
2d

(Eq 17)

where d is the deflection of the strip (Fig. 11a).


According to Eq 12:
Et
r
4R

(Eq 18)

it follows:
r

A simpler technique to measure the circumferential hoop stress in a thin-walled tube is to


slit the tube longitudinally with a carborundum
cutoff wheel (using plenty of coolant to avoid
heating of the surfaces) and measuring the
change in diameter of the tube (Fig. 13b):

Etd
2L2

r Et

1
0

1
D1

Using a carborundum wheel, a ring, either

parted from the tube or machined, is cut axially through the section.
If unknown, the saw curf width (width of saw

(Eq 22)
0.600
0

where D0 is initial diameter and D1 is diameter


after slitting.
Alternatively, the net opening displacement at
the saw cut is measured by subtracting the width
of the saw blade, and the residual hoop stress is
measured using the formula:

(Eq 19)

A similar derivation can be obtained for round


bar (Fig. 11b):

following method is often used to identify the


possible cause:

1
1
r Et

x
D0
D0
p

1.000

2.900
B

1.900

(Eq 23)

E
5.000
(a)

1.65 Erd
L2

(Eq 20)

For thin-walled tubing (Fig. 13), the longitudinal


stress is given by:
r

Etd
L2

(Eq 21)

Equation 21 is the same as Eq 19 except that


t/2 is used in the former case because only half
the sheet thickness deflects.
The method of sectioning a tube to obtain the
previously mentioned approximate measurement
is shown in Fig. 13(a).

where x is the net opening displacement.


A summary of the derivations is given in Table 2.
A Practical Example. The formulas in Table
2 are used extensively for measuring residual
hoop stresses in seamless tube and machined
bearing rings. These stresses may arise from tube
straightening without a stress-relief anneal or
nonuniform rapid cooling following the spheroidized anneal. In the past, contact with the support bars on cooling beds has been recognized
as being a potential cause of cyclic nonuniform
stress along the length of a tube.
Where 52100 bearing steel rings are being machined on multispindle machines and the operator is having difficulty in maintaining size, the

0.25 in.

0.50 in.

2.50 in.
0.5 in.

1.45 in.

(b)
2 in.
1/4 in.

1 1/4 in.

Residual stress , ksi

3/4 in.

2000:5:9
Peen stress
ENSAMMIC

Residual stress distribution


8
Depth, mils

5/8 in.

d = 5.198 mils
h = 0.294 mils
10 A
S 230 shot
V = 124 ft/s
HS/4320 CB/SP

100

(c)
Examples of C-ring test specimens used for
quench distortion studies. (a) Source: Ref 5. (b)
Source: Ref 6. (c) Source: Ref 7

Fig. 9

Dir. Z

x
s (ksi)

102

m (ksi)

203

Pm (in.)

0.001

P0 (in.)

0.008

B
d

A
l

200

W
M = Wx
Max. M = Wl at B

Fig. 8

Calculated residual stress distribution in a shot peened chromium-silicon spring wire. d, depth; h, height. Source:
Peen Stress, Metal Improvement Company, Inc.

Fig. 10

Simple beam theorydefinition of bending


moment. W, weight in lbs or kg

Deflection Methods to Estimate Residual Stress / 95


cut) is determined by measuring a shallow cut
adjacent to the final cut.
The gap width is measured using internal calipers. The net opening (x) equals the gap
width minus the saw curf (width).
On occasions the gap may decrease, indicating a
compressive hoop stress. In extreme situations,
the blade may become nipped and may even disintegrate. Ensure proper protection when using
this method.
For greater accuracy in large cross sections,
the value for E can be corrected for Poissons
ratio by:
E

E
28.57 106 psi
1 m2

these may lead to quenched-in nonuniform


stress. During subsequent grinding operations,
the stress distribution undergoes a redistribution
and may, in extreme cases, lead to the rings distorting. Although heat treating receives the
blame, many similar problems may, in fact, be
attributed to the incoming raw material.

Validation of Net Opening


Calculation
Considerable work has been accomplished in
recent years in the area of heat treatment modeling and, in particular, of software capable of

Elastic constants for several materials are


given in Table 3 (Ref 16). (These must only be
used as a guide. For more rigorous treatment of
elastic and plastic distortion, actual values
should be obtained for each material and condition under consideration.)
From experience in dealing with distortion
problems in manufacturing that are attributable
to residual stress in tube stock, it was determined
that the following criteria could be used as a
guide to the acceptability of the calculated stress
levels.
Residual stress

Acceptability

5000 psi
500010,000 psi
10,000 psi

Acceptable
Borderline
Unacceptable

The criteria used to determine what residual


stress is acceptable must also be based on the
cross section of the ring: the more rigid the design, the higher the stress level tolerated. However, this only applies to holding tolerance during machining. During subsequent heat
treatment, the magnitude of residual stress may
still be sufficient to cause distortion during heating. As the temperature increases, the yield
strength of the steel decreases, allowing relaxation by plastic deformation. This is manifested
in the parts being out-of-round exiting the furnace.
In very thin sections, some of the stress may
be from the machining operation (e.g., feeds and
speeds too high).
In order to lessen the effects of residual
stresses during heat treatment, pre-heating at a
subcritical temperature has been found to be useful. A portion of the stress is allowed to relax
without causing a gross shape change.
Examples of extreme levels of residual stress
in tube material are usually seen when it was
impossible to machine an acceptable part. Stress
relieving one particular batch of tubes (new supplier) resulted in them moving so far out-ofround that they no longer fit in the multispindle
machine collets.
In an attempt to avoid out-of-round rings, die
quenching techniques are often used. However,

predicting distortion in heat-treated products. An


example is DANTE (Deformation Control Technology, Inc., Cleveland, OH), developed by a
collaborative team comprising Ford Motor Company, General Motors Company, The Torrington
Company, Eaton Corporation, IITRI, Colorado
School of Mines, The Department of Energy National Labs, Deformation Control Technology,
Inc., and under the auspices of the National Center for Manufacturing Science. In order to accurately predict the thermal and allotropic stressinduced size changes and distortion in
heat-treated products (including carburized), it is
necessary to incorporate many factors into the
finite element-based model, including the transformational characteristics of the steel, the ele-

2d

Residual stress
distribution

(a)

2d + 2 r

(b)

Fig. 11

Determination of longitudinal residual stress by the deflection method. (a) Rolled sheet. (b) Drawn bar. Source:
Ref 14

t
d
L

(a)

D1

D0

x
(b)
Determination of residual stresses in thinwalled tube by deflection methods. (a) Longitudinal stress. (b) Circumferential stress. t, thickness; d, deflection; D0, initial diam; D1, diam after slitting; x, net
opening displacement. Source: Ref 14

Fig. 13

Fig. 12

Diagrammatic representation of the deflection


formula derivation. Source: Ref 15

96 / Measurement and Prediction of Residual Stress and Distortion


Using DANTE, it has been possible to subject
the net opening deflection method to a more rigorous analysis.
An example of the simulation of the opening
of a slit ring in the presence of an imposed hoop
stress is shown in Fig. 14 and 15.
Two-dimensional modeling was used, and
plane-strain conditions were assumed. The rings
were assumed to be 152 mm (6 in.) OD and man-

vated temperature properties and behavior of the


individual and mixed phases, geometry of the
part, quenching factors, heat-transfer coefficients, and so on. Several papers have been published on the development and application of the
model (Ref 1721). Commercialization of the
software package is being handled by Deformation Control Technology, Inc., Cleveland,
Ohio.

Table 2 List of formulas used for calculating approximate levels of residual stress in simple
geometries by deflection methods
Method

Formulas

Longitudinal stress in plate (Fig. 11a)

Etd
r 2
2L

Longitudinal stress in solid bar (Fig. 11b) where r is the radius of the bar

1.65 Erd
L2

Longitudinal stress in thin-walled tube (Fig. 13a)

Etd
L2

Circumferential (hoop) stress in tube from change in diameter where D0 and D1


are the initial and final diameters, respectively (Fig. 13b)

r Et

Circumferential stress in tube from net opening displacement x

r Et

Table 3

1
0

1
D1

1
1

x
D0
D0
p

Example: Determining Biaxial


Residual-Stress State

Elastic constants for some of the common metals


Modulus elasticity

Shear modulus

Materials

GPa

106 psi

GPa

106 psi

Poissons ratio

Aluminum alloys
Copper
Steel (plain carbon and low alloy)
Stainless steel
Titanium
Tungsten

72
110
200
193
117
400

10.5
16.0
29.0
28.0
17.0
58.0

28
41
76
66
45
157

4.0
6.0
11.0
9.5
6.5
22.8

0.31
0.33
0.33
0.28
0.31
0.27

Source: Ref 16

Fig. 14

ufactured from annealed 52100 bearing steel


with two different cross sections. A residual
stress profile was superimposed on the rings by
creating a 93 C (200 F) linear temperature gradient across the cross section from ID to OD.
For the ring in Fig. 14, simulating axial slitting
of the ring resulted in a net opening of 2.9 mm
(0.114 in.) with an accompanying redistribution
of the residual stress.
Using Eq 23, the average hoop stress in the
ring prior to slitting was calculated to be 172
MPa (25 ksi). In thinner cross section rings, the
net opening will be larger for the same level of
residual stress, as shown in Fig. 15. For a cross
section of 9.5 mm (0.375 in.) and a similar residual stress condition, the net opening is calculated to be 6.3 mm (0.247 in.). Equation 23
gives a calculated hoop stress of 179 MPa (26
ksi). These calculations used a Youngs modulus
value of 29 106 psi and Poissons ratio of
0.33.
It is interesting to note that one of the many
techniques used to exercise DANTE was based
on the carburizing of a modified Almen strip
(Ref 22, 23). Excellent agreement between the
predicted and measured transverse deflections
were obtained.

Simulation of the effects of residual stress on the net opening displacement of a cut 0.75 in. thick ring

Treuting and Read developed a method for determining the biaxial residual-stress state on the
surface of a thin sheet (Ref 24). The method assumes the metal behaves in an elastically homogeneous manner and that the stress varies, not
in the plane of the sheet, but only through the
thickness. To apply the method, the sheet specimen is cemented into a flat parallel surface, and

Deflection Methods to Estimate Residual Stress / 97

Fig. 15

Simulation of the effects of residual stress on the net opening displacement of a cut 0.75 in. thick ring

the thickness is reduced a certain amount by


careful polishing and etching. The sheet specimen is then released from the surface and measurements are made of the longitudinal radius of
curvature Rx., the transverse radius of curvature
Ry, and the thickness, t. Figure 16 illustrates the
orientation of the principle stresses and the curvature of the sheet. The measure values of radius
of curvature are expressed in terms of two parameters, Px and Py.
Px

1
m

Rx
Ry

Py

1
m

Ry
Rx

rx

(t

t)2

ry

(t

E
6(1 m2)
dPx
4(t0 t)Px 2
dt

t0

Px dt

E
6(1 m2)

t)2

dPy
4(t0 t)Py 2
dt

t0

Py dt

Values of dP/dt are obtained from the slope of


the curves of P versus t, and the integrals are
evaluated by determining the area under the P
versus t curve over the appropriate limits.
REFERENCES

Measurements of Rx and Ry are made for different amounts of metal removal, and Px and Py are
plotted against the sheet thickness, t. The residual stresses in the x and y directions of the sheet
are determined for any value of t by the following equations.

1. G. Parrish and G.S. Harper, Production Gas


Carburizing, Pergamon Press, 1985
2. F.W. Jones, JISI, May 1969, p 556562
3. Shot Peening Applications, 7th Ed., Metal
Improvement Company, Inc., 1995

z
z
y
y

x
x

t0

(a)

Fig. 16

Rx

Ry

(b)
(a) Coordinate system for measuring biaxial stress in thin sheet. (b) Curvature produced by removing material
from top surface

4. D. Breuer, Metal Improvement Company,


private communication, 1999
5. Heat Treating, Cleaning and Finishing, Vol
2, Metals Handbook, 8th ed., ASM International, 1964, p 41
6. Tenaxal, Ucon Quench AThe Fast Safe
Way to Quench Steel Alloys, Product Information Bulletin, Tenaxal, Inc., Milwaukee, WI, revised 1972
7. H.J. French, The Quenching of Steels,
American Society for Steel Treating, 1930,
p 133
8. B. Prime, Residual Stress Measurement by
Successive Extension of a Slot: The Crack
Compliance Method, Los Alamos National
Laboratory, Publication LA-UR- 983857,
1998
9. W. Gremaud, I. Cheng, M. Finnie, and B.
Prime, The Compliance Method for Measurement of Near Surface Residual
StressesAnalytical Background, J. Eng.
Mater. Technol., Vol 116, p 550555
10. I. Cheng, M. Finnie, M. Gremaud, and B.
Prime, Measurement of Near Surface Residual Stresses Using Electric Wire Machining,
J. Eng. Mater. Technol., Vol 116, p 17
11. Kovac, Residual Stress Measurements in
Bevel Gear after Different Production
Phases, J. Mater. Eng. Perform., Vol 3 (No.
1), Feb 1994, p 6164
12. E. Shramm, J. Szelazek, and A.V. Clark,
DynamometerInduced Residual Stress
in Railroad Wheels: Ultrasonic and Saw Cut
Measurements, National Institute of Standards and Technology Publication NISTIR
5043, Report Number 30, March 1995
13. J.E. Gordon, The New Science of Strong
Materials, Penguin Books, 1968, p 258
261
14. G.E. Dieter, Mechanical Metallurgy, 1st
ed., McGraw-Hill Book Company, New
York, 1961
15. R.L. Anderson and E.G. Fahlman, A

98 / Measurement and Prediction of Residual Stress and Distortion


Method for Measuring Internal Stress in
Brass Tubes, J. Inst. Met., Vol 32, 1924, p
367383
16. G.E. Totten and M.H. Howes, Steel Heat
Treatment Handbook, Marcel Dekker Inc.,
New York, 1997, p 256261
17. D. Shick et al., Development of a Carburizing and Quenching Simulation Tool: Determination of Heat Transfer Boundary Conditions in Salt, Second International Conf.
on Quenching and Control of Distortion
Proc., ASM International, 1996
18. D. Brammann et al., Development of a Carburizing and Quenching Simulation Tool: A
Material Model for Low Carbon Steels Undergoing Phase Transformations, Second

International Conference on Quenching and


Control of Distortion Proceedings, ASM
International, 1996
19. C. Anderson et al., Development of a Carburizing and Quenching Simulation Tool:
Numerical Simulation of Rings and Gears,
Second International Conference on
Quenching and Control of Distortion Proceedings, ASM International, 1996
20. B.L. Ferguson, A.M. Freborg, and G.J. Petrus, A Software Tool to Simulate Quenching of Alloy Steels, Heat Treating Progress,
to be published
21. M.T. Lusk, Y.K. Lee, H.J. Jou, W.E. Elliott,
and G.M. Ludtka, An Internal State Variable
Model for the Low Temperature Tempering

of Low Alloy Steels, First International


Conference on Thermal Process Modeling
and Computer Simulation, Jiatong University, Shanghai, March 2000
22. V.C. Prantil, M.L. Callabresi, G.S. Ramaswamy, and J.F. Lathrop, Simulating Distortion and Residual Stresses in Carburized
Thin Strips, ASME J. Eng. Mater. Technol.,
to be published
23. M. Henriksen, D.B. Larson, and C.J. Van
Tyne, On the Analysis of Distortion and Residual Stress in Carburized Steels, ASME J.
Eng. Mater. Technol., Vol 114, 1992, p
362367
24. R.G. Treuting and W.T. Read, J. Appl.
Phys., Vol 22, 1951, p 130134

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p99-117
DOI: 10.1361/hrsd2002p099

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Measurement of Residual Stresses


C. Ruud, Pennsylvania State University

THIS ARTICLE PROVIDES an insight into


the principles, practices, and limitations of residual-stress measurement procedures for steel. It is
not meant to provide sufficient detail for the performance of the various methods described, but
references are cited where such procedural details may be found (e.g., Ref 15 or for general
inquiries, www.residualstress.com on the Internet).
There have been many methods and techniques proposed for the measurement of residual
stress, but only a few may be applied in practice
on components ranging from small to very large,
such as bridges and aircraft. These few include
x-ray diffraction (XRD) and blind hole drilling
with electrical resistance strain gages. For some
of the methods described, the component in
which residual stresses are to be measured must
be brought to the measuring instrument, but for
others the measurement devices are portable and
may be brought to the component (Fig. 1). However, in some cases it may be feasible to remove
a section from the component and bring that section to the residual-stress measuring device.
Great caution must be observed in this sectioning
because it will change the stress field by relieving and/or inducing stresses.

have resulted in disagreement between residualstress measurements made by the mechanical


stress relaxation type methods discussed in the
section Introduction to Destructive Procedures
and the XRD methods discussed in the section
Nondestructive Techniques in this article. The
disagreement is nearly always because the volume of the component in which the stress is measured is not the same in the two methods, and
thus a different portion of the stress gradient is
measured by each. Ruud et al. (Ref 6) showed
that the hole-drilling method resulted in measured stresses approaching that of the stresses
measured at the surface by XRD when the holedrilling results were extrapolated to the surface.
Furthermore, because of the steep gradients, unless the stress field caused by the process is well
understood, stress measurements must be performed at many locations in the manufactured
solid in order to establish the magnitude and distribution of the stress field of interest. Many researchers in residual-stress techniques have fo-

Any manufacturing process that changes the


shape of a solid, or where severe temperature
gradients exist during the process, causes residual stress. By their very nature, processes that
change the shape of a solid cause nonuniform
plastic deformation in the solid, which leads to
residual stress. These processes include forging,
rolling, drawing, machining, and so forth. Also,
processes that produce high thermal gradients in
a solid often lead to residual stress. These processes include quenching, casting, welding, and
so forth. Furthermore, processes that induce localized phase changes produce residual stress.
These processes include martensitic hardening.
The residual stresses caused by manufacturing
processes usually show very steep residual stress
to distance gradients as shown in Fig. 2. Many
of the other articles in this Handbook discuss and
describe in detail the causes and mechanisms of
residual-stresses development in manufacturing
processes. The steep gradients typical of residual
stresses induced by manufacturing processes

) dxd > 200mmMPa )


y
650

MPa

Welding
d y
~ 200
dx

3 mm

650

MPa

Machining
d y
~ 3000
dx

y
0

0.1 mm

y
650

Measurement of residual stresses on oil-drilling


platform component weldments using a portable
x-ray diffraction instrument

MPa

Drawing
d y
~ 600
dx

Drawn
cup
0

Fig. 1

Fig. 2

3 mm

x
x

Residual-stress magnitudes and distributions typical of a 650 MPa yield strength steel

100 / Measurement and Prediction of Residual Stress and Distortion


cused on enhancing the accuracy of
residual-stress measurement and have ignored
the fact that from a practical standpoint, tens to
hundreds of stress measurements are needed to
define the stress field of interest. Thus, many of
the residual-stress measurement techniques require too much time to perform and are thus impractical. This includes some of the techniques
developed for XRD, strain gaging, and other
methods. Measurement times on the order of a
second are available with XRD and some other
methods, and automated stress mapping has been
performed with such techniques (Ref 7). Another
concern in the measurement of residual stresses
in manufactured components is that the area or
volume over which the stresses are resolved
must often be on the order of 1 mm (0.04 in.) or
less. This is because the residual-stress gradients
are usually quite steep, and measurement resolution larger than this tends to average the
stresses to such an extent that high stresses are
not detected.
Some characteristics of steel that can cause
error in residual-stress measurement by the various methods described in this article include
phase composition, plastic strain, grain size,
crystallographic texture, and others. For example, in a mixed ferrite/austenitic structure the residual stresses in the ferrite are invariably different than those in the austenite (see Table 1).
Also, Wimpory et al. (Ref 8) described the influence of varying amounts of cementite in a ferrite
matrix. The possibility of one or more of these
microstructural characteristics causing error in
the residual-stress measurements performed by
the methods described in this article should be
assessed by an expert in the method selected, and
details of the errors and their causes are not discussed in detail in this article.
The subsequent sections of this article discuss
and describe:
The need for measurementwhat problem is

the engineer or metallurgist trying to solve by


obtaining information about the residualstress field?
The nature of the residual-stress fields in
steelsexamples of the magnitudes and distributions.
The strain basics for residual-stress measurementselastic strain measured, not stresses.
Table 1 Sample of residual-stress readings
from a 316 stainless steel pipe weldment
Distance from
the weld
fusion line

Stress
(a) in
austenite

Stress
(a) in
ferrite

in.

mm

ksi

MPa

ksi

MPa

0.04
0.07
0.11
0.15

1
1.8
2.8
3.8

21
19
16
17

145
131
110
17

46
67
62
65

317
462
427
448

Note: The ferrite places tensile stresses on the lattice of the austenite,
while the austenite tends to compress the ferrite; therefore, the more
compressive the stress in the ferrite, the less compressive the stress in
the austenite. (a) The precision of these measurements was 3.0 ksi
(21 MPa).

The destructive procedures of residual-stress

measurementthese procedures are all


based on sectioning or removal of material to
cause a redistribution of the residual stress,
which is measured as a strain change.
The semidestructive methods of residualstress measurementthese methods are
based on the same principle as the destructive
methods or on the perturbation of the residual-stress field by other means.
The nondestructive methods of residual-stress
measurementthese methods do not permanently disturb the residual-stress field, but
directly measure the atomic lattice strain
caused by the stress or measure some physical
property perturbed by the lattice strain.

Need for Residual-Stress


Measurements
Before the engineer or metallurgist commits
to measuring residual stresses in some component or workpiece, he or she must be sure that
the reason for the measurement is clearly understood. The major reasons that residual stresses
are of concern are:
Failures that are suspected as being caused by

fatigue, stress corrosion, corrosion fatigue, or


hydrogen embrittlement
Assessment for the continued serviceability
of a component, for example, life assessment;
this is usually focused on a concern for inservice failure
Distortion occurring during processing of a
component
Distortion of components during storage or in
service
It is extremely important that the investigator
understand the mechanism for the inducement of
the residual-stress field of concern. As implied
by the other articles in this Handbook, most
cases of suspected harmful residual-stress fields
are induced by manufacturing processing or repair procedures, although sometimes abusive
service conditions or an accident may have
caused them (Ref 9). When manufacturing processes or sometimes repair procedures are
judged the most likely source of the residual
stresses, it is often possible to predict the magnitude and distribution of the residual stresses.
Such information can be obtained through consulting the literature or the application of computer modeling (Ref 1018 or the Internet at
www.residualstress.com). A preconceived
model of the residual-stress field will aid the investigation in determining the best method to
measure residual stress and the location and
number of measurements that need to be made.
Nevertheless, sometimes the cause of the residual-stress field is not evident, and the investigator is compelled to perform measurements as
a means to determine the cause. In such cases,
measurement methods and location must be selected without the aid of a priori knowledge of
the stress field, and it is prudent to consult the

literature and experts in the field of residualstress measurement and manufacturing processes.

Nature of Residual Stresses


Residual stresses are the inevitable consequence of thermomechanical processing of steel.
The resulting stress fields usually are nonuniform and show high stress gradients. For example, Fig. 2 illustrates the residual-stress magnitudes and distributions typical of a steel with
a 650 MPa yield strength. Because of the high
stress gradients, tens to hundreds of residualstress measurements with resolution on the order
of 1 mm may be required to identify precisely
the maximum stress and its location.
The characteristically nonuniform, high stress
gradient nature of residual stresses require that
either the induced stress field is well understood
and predictable, or many residual-stress measurements must be performed on one or more
components in order to reveal the nature of the
stress fields. Often, scientists attempt to gain an
understanding of a residual-stress field by making a few measurements on one or two components, and from this often arise erroneous conclusions regarding the nature of the stress field.
A few measurements may be useful if the scientist or engineer knows the distribution of the
stresses a priori. However, this is seldom the
case, and tens to hundreds of measurements are
required on a single component or many samples
to really understand the residual stresses induced
by a given manufacturing process. This means
that the measurement method must be as rapid
and labor-efficient as possible. Some of the new
semidestructive hole-drilling procedures and
XRD and ultrasonic (for special cases) instrumentation meet this criteria. Stress mapping is
offered with some XRD instruments to map
stresses over the surface of a component (Ref
19). Also, because the stress gradients are often
very high, the measurement method must be able
to resolve the stresses in dimensions on the order
of 1 mm or less. Here, some of the hole-drilling
procedures and recently developed XRD instruments can offer the best resolution.
Finally, the component in which residual
stresses are to be determined is often too large
to be brought to a laboratory, and removing sections is not a rational solution. Note that sectioning often disturbs the existing stress field to the
extent that it is not possible to reconstruct the
original stress field from the residual stress measured in pieces removed from the original whole.
Hole-drilling, XRD, and ultrasonic instrumentation are available as portable devices that may
be brought to the component in the field (Ref 7,
9, 20, 21).

Stress Measurement
A number of procedures and methods have
been applied to determine the residual stresses

Measurement of Residual Stresses / 101


extant in a metallic component, usually as a result of manufacturing processing. However,
stress is never the quantity measured because
stress is a quantity that is applied to a metal and
can only be measured in the process of its application. What is invariably measured to determine residual stress is elastic straineither the
elastic strain resulting directly from the existing
residual stress in the metal or the elastic strain
change resulting from relief of some portion, or
all, of the existing residual stress. The stress that
is causing, or has caused, the strain is then calculated using the applicable elastic constants for
the metal.
The methods described in the section Nondestructive Procedures in this article measure,
directly or indirectly, the strain response of the
metal to the residual stress in situ, while the
methods described in the sections Destructive
Measurement Procedures and Semidestructive
Procedures measure the strain change caused in
relieving some or all of the residual stress in the
metal.

Destructive Measurement
Procedures
The first concern in selecting a destructive residual-stress measurement procedure is whether
it is reasonable to destroy one or more components or samples in order to determine the residual stresses. Usually this implies that one or a
few of the components are a small portion of the
total number produced. Also coupled to this decision is whether the one or more components in
which the residual stresses are to be measured
are representative of all the others. In other
words, how great is the expected variation of the
residual-stress field from part to part?
As mentioned at the beginning of this article,
the need for the residual-stress information about
a component must be clearly understood. This
includes whether the triaxial stress field must be
established or if the uniaxial or biaxial condition
of the stress field along specific directions is sufficient. Examples of these three situations are
discussed in this section. However, in any case,
the fact that residual stresses are usually not uniform in any direction and show high stress gradients must be kept in mind when stress-measurement criteria are selected.
Destructive methods of residual-stress measurement are fundamentally stress-relaxation
procedures; that is, the information is obtained
by relaxing the residual stress in some finite-volume element of the component and measuring
the resulting strain change. The strain change is
then used, along with applicable assumptions
about the nature of the stress field, to reconstruct
the original stress field. Assumptions about the
nature of the stress field include the magnitudes
and gradients in the stress field and whether it is
sufficient to assume that the gradients are one-,
two-, or three-dimensional. In particular, the gradients that exist will dictate the size of the element that is to be isolated and made stress-free;

that is, the higher the stress gradient is, the


smaller the finite element must be in the direction of that gradient. It must be emphasized that
the larger the element and the higher the stress
gradient, the less quantitative and more qualitative are the measurement results. Electrical-resistance strain-gage technologies are emphasized
as the dominant method of strain measurement
due to their economic, procedural, and precision
advantages over other methods. However, modern XRD equipment when available has all of
these advantages as well and can be used to measure the stresses existing before and after sectioning.
A generic destructive stress-relief procedure is
described first, along with the issues generally
involved in each procedural step. This section on
destructive methods concludes with some discussion of qualitative chemical methods of residual-stress measurement.

Generic Destructive Procedure


Once the decision is made to measure the residual stresses destructively, the following steps
are usually applied in a typical stress-relief technique for residual-stress measurement.
Stress-Field Conditions. The engineering
problem for which the residual-stress information is needed must be analyzed. This need is
often generated by failures of the component in
service or by anticipated failures due to problems
with similar components. Distortion of a product
in storage or during manufacturing can also be a
concern. The shape of the componentthat is,
cylinder, disk, plate, and so forth, or some irregular shapemust be considered. This consideration, along with the process or processes by
which residual stresses were introduced, must be
analyzed. The justification for assumptions regarding the condition of the residual-stress field
can be established from these considerations.
This may lead to simplifying assumptions about
the stress-field condition such as axial symmetry
for a cylinder in which the dominant residualstress field is caused by quenching during heat
treat processing or about stress uniformity in the
surface plane of a plate where stress gradient
with depth is the major concern.
These assumptions and considerations lead to
the methodology, that is, equations, to be used
for computational reconstruction of the stress
fields from the measured strains.
Strain-Measurement Technique. With the
stress-reconstruction approach established, the
method of strain measurement and consequently
the number and/or spatial frequency of measurements can be determined. The strain-measurement technique selected will greatly affect
the resolution of the stress measurement because
of the spatial precision inherent in the technique.
There are a number of techniques that have
been used to measure the strain induced by the
relief of stresses due to sectioning or material
removal in destructive residual-stress measurement. These include mechanical gages, often dial
gages, employed with specially made jigs and

fixtures, reflected light schemes, photoelastic


coatings, and electrical-resistance strain gages.
However, since the 1960s, the use of the latter
has become dominant due to the variety, availability, and precision of these gages. They are
available as uniaxial, biaxial, and rosette gages
of many sizes. The section Strain-Measurement
Methods in this article provides some detail regarding these methods. Also, since the late
1970s, extensive use of XRD has been applied
to provide rapid and numerous stress measurements on sectioned components to gain information regarding the internal stress field (Ref 2,
2225).
Preparation for Strain Measurement. With
the strain-measurement technique selected, the
measurement location must be established and
the component and/or element prepared for the
measurement by, for example, attaching strain
gages. A prestress-relief reading must now be
made before stress relaxation and isolation of the
element is initiated.
Isolation of Gaged Element. With the measurement technique in place, material removal to
isolate the gaged volume must be performed.
The technique for material removal, or sectioning, must be carefully considered because mechanical chip-removal processes such as lathe
turning, milling, sawing, grinding, and so forth
introduce surface residual stresses that can be as
great as the yield strength of the strain-hardened
material and several thousandths of an inch (tens
of micrometers) in depth. The section Sectioning and Material-Removal Methods in this article discusses methods to isolate the gaged element.
Post-Stress Relaxation Measurement. After
the residual stresses have been relaxed and thus
the elements isolated, strain measurements are
repeated, the final reading is subtracted from the
initial to obtain the strain change resultant from
the residual-stress relaxation. The resultant
quantities are then used in the residual-stress reconstruction equations to obtain the original
stress state of the component. The stress-reconstruction equations were selected as a result of
the assumptions made about the stress-field conditions described previously and in more detail
in the following sections.

Stress-Field Condition Assumptions


Engineers and research scientists have approached the measurement of residual stresses
using destructive methods with the aid of assumptions about the stress-field conditions, including that the stresses in only one axis are of
interest in order to simplify the measurement and
reconstruction of the stress field (Table 2). These
have included certain uniaxial, biaxial, and triaxial stress-field assumptions.
Uniaxial Conditions. A procedure applicable
only to the measurement of residual stresses in
rods, cylinders, and tubesthat is, components
with axial symmetrywas reported by Heyn
(Ref 26). In this work, it was assumed that the
stresses were axially symmetric and that only the

102 / Measurement and Prediction of Residual Stress and Distortion


longitudinal stresses were of interest. Thus, the
change in length of the components (cylinders,
rods, and tubes) was measured after removal of
an axially symmetric layer from the outside radius, or it was bored out of an inside radius. The
length of the component was measured after
each layer-removal process and entered into
various equations described in their paper. This
procedure is called the Bauer-Heyn method (Ref
26) and is applicable only to measuring longitudinal stresses in axially symmetric components. It does not measure radial or circumferential stresses.
A second procedure assuming a uniaxial stress
field, or that only the stresses in one axis are of
interest, was proposed by Stablein (Ref 27). Here
the component was a bar with a rectangular cross
section, and the residual stresses acting along its
length and varying through its thickness (smallest dimension) were measured. The material
from one face of the bar, one of the two faces
with the largest area, was removed by milling.
Equations used to reconstruct the original stress
field are described in Ref 27. The depth of the
removed layer must be significantly greater than
the depth of plastic deformation caused by milling (see the section Sectioning and MaterialRemoval Methods) and sufficient to cause a
measurable bend in the material. The length of
the opposite face of the bar from where the material was removed is measured before and after
layer removal to determine the effect of the removal of the stressed layer. Presently, this is usually done using electrical-resistance strain gages
(see Strain-Measurement Methods), but was
performed in the past by sensitive mechanicalgaging techniques. This measurement can also
be accomplished by measuring the bend in the
bar with suitable mechanical gages and fixtures,
for example, a cantilever-beam approach. This
procedure is applicable only to components of
rectangular cross section where the stresses parallel to the length are to be measured as they vary
through the thickness.
Biaxial Conditions. A procedure applicable
to axially symmetric components is the Mesnager-Sachs boring-out technique (Ref 28, 29). The

Table 2

technique is applicable to cylindrical components with an axially symmetric distribution of


stresses. Here the change in length and diameter
of the component are measured as material is
removed by axially boring-out material from the
inside to produce a hollow cylinder. Presently,
biaxial electrical-resistance strain gages (see the
section Strain-Measurement Methods) are
usually attached to the outside of the component
to measure the dimensional changes in the axial
and tangential directions. The strain change results are entered into equations described in Ref
28 and 29.
A procedure applicable to measuring biaxial
residual stresses homogeneous over the planar
surface of a flat metal plate or sheet was proposed by Trenting and Read (Ref 30). It was
based on removing uniformly thin layers of the
metal on one side of the sheet or plate and measuring the changes in curvature as the layers
were removed. It was assumed that the stresses
were constant over the plane of the sheet or plate
and varied only through the thickness. Electricalresistance strain gages (see Strain-Measurement Methods) or mechanical gaging may be
used to measure the change in curvature.
Another procedure for measuring biaxial
stresses homogeneous over the planar surface of
a metal plate was developed by Gunnert (Ref
31). This procedure assumes that a biaxial stress
condition was uniform throughout the depth of
a circular groove that was milled around an elemental volume of material to render it stress
free. Thus, the strain change on only one surface
was measured. The mechanical-gaging technique involved measuring the distance between
each of four sets of shallow holes drilled in the
element before the groove was milled into the
surface using a core drill. The distance between
each set of holes was measured before and after
the groove was produced and provided the information necessary to calculate original, assumed biaxial residual-stress condition parallel
to the surface of the plate. Theoretically, only
three sets of holes are required to measure the
biaxial stresses, but Gunnert used a fourth set to
improve the accuracy. A mechanical gage,

termed an extensometer, was used to measure the


distance between each set of holes. It should be
noted that this technique could be used to measure the gradient in the biaxial stress condition
by pausing in the milling operation at selected
depths and measuring the distance between the
holes at each groove depth. Also, as with many
of the older techniques that originally applied
mechanical strain-gage apparatus, electrical-resistance strain gages or modern XRD techniques
could be used. A procedure involving the drilling
of a blind hole and electrical-resistance strain
gage is somewhat similar to Gunnerts original
technique and is described in the section on
semidestructive methods.
A more accurate procedure was later used to
measure residual stresses in pipe weldments.
Here the component (pipe) was divided into a
network (grid) of squares, and biaxial electrical
strain gages were placed in the center of each
grid square on the outside diameter of the pipe.
The pipe was then sectioned as shown in Fig. 3
into elements assumed to be stress free, and the
strain induced by the stress relief was read from
the gages. In placing the gages only on the outside diameter, the biaxial stress field was assumed to be uniform with depth; however, had
gages been placed on the inside and outside diameter a more complete measurement of the
stress field could have been obtained, albeit assuming a linear variation in the residual stress
from the outside to the inside surface.
A variation on the Gunnert procedure described previously was later published (Ref 1,
32). This variation was also used for application
to plates and implied that the triaxial stress field
could be measured by the technique. It assumed
that there was a homogeneous residual biaxial
stress field that varied with depth through the
plate thickness. Four holes were drilled in a
square pattern through the plate thickness. The
distances between all of the holes was then measured at selected hole depths. Next, a circular
groove was milled in steps of, for example, 2
mm around the drilled holes using a core drill as
in Gunnerts original procedure (Ref 31). The
distance between the holes at the various depths

Summary of destructive residual-stress measurement procedures

Component shape

Rods, cylinders, tubes


Rectangular cross-section bar
Plate, sheet

Cylinder, plate
Plate, weldment

Stress-field-condition assumptions

Stress direction measured

Method

Uniaxial stresses, axial symmetry


Biaxial stresses, axial symmetry
Uniaxial stresses varying through thickness
Homogeneous planar, biaxial stresses
varying through thickness
Homogeneous planar, biaxial stresses
uniform through thickness
Homogeneous planar, biaxial stresses
varying through thickness
Planar biaxial stresses varying through
thickness
Planar biaxial stresses varying through
thickness
Triaxial

Longitudinal
Longitudinal, radial
Longitudinal
Biaxial in the plane of the
component
Biaxial in the plane of the
component
Biaxial in the plane of the
component
Biaxial in the plane of the
component
Biaxial in the plane of the
component
All

Bauer-Heyn
Mesnager-Sachs
Stablein
Treuting and Read

Various
Triaxial

All
All

Moore and Evans


Johanssen

Gunnert
Gunnert
Rosenthal and
Norton
Moore and Evans
Chen

Section and reference

Triaxial Conditions (Ref 26)


Biaxial Conditions (Ref 28, 29)
Triaxial Conditions (Ref 27)
Strain-Measurement Techniques (Ref
30)
Strain-Measurement Techniques (Ref
31)
Strain-Measurement Techniques (Ref 1,
32)
Strain-Measurement Techniques (Ref
33)
Strain-Measurement Techniques (Ref 2)
Preparation for Strain Measurements
(Ref 34)
Strain-Measurement Techniques (Ref 2)
Strain-Measurement Techniques (Ref
25)

Measurement of Residual Stresses / 103


was then measured at each core depth until the
core was milled completely through the plate.
The cored plug was then assumed to be completely free of residual stress. The measured distances were then used to reconstruct the original
biaxial stress condition of the plug at each cored
depth. The mechanical measuring gage used in
this technique was similar to that used in Gunnerts first technique (Ref 31).
Another approach to measuring residual
stresses that has a broader application with respect to the shape of the component and the
stress-field distribution was proposed by Rosenthal and Norton (Ref 33). It is applicable to
plates and plate-shaped weldments.
The procedure involved cutting two narrow
blocks having the full thickness of the plate, each
with its long axis parallel to one of the assumed
biaxial principal, residual-stress directions in the
surface of the plate (see the near side of Fig. 4).
Thus, the long axes of the blocks are perpendicular to each other and parallel to the face, the
largest area surface of the plate. The smallest
dimension of the block should be several times
smaller than the plate thickness and its largest
dimension at least twice the thickness. The block
then can be further sectioned in order to determine the biaxial stress variation through the
thickness of the component. This proceeds by
first cutting the block at the location representing
the midthickness of the component plate, then

removing thin slices parallel to the original surface and from the bisected block as shown in the
upper part of Fig. 4.
The change in strain of the blocks is measured
using shallow holes or dimples in the original
surfaces of the component. These gage points are
located along the long axis of the block on the
original faces of the plate (see Fig. 4). The distance between these gage marks is measured before and after removal of the blocks from the
plate and after each sectioning of the block (Ref
33).
This procedure assumes a constant biaxial
stress field over the length of the blocks, which
is not the case in welded plates in the direction
transverse to the weld. The far side of the block
in Fig. 4 shows a sectioning procedure that
would reveal the stresses parallel to the weld
along a gradient transverse to the weld. The far
side of the block in Fig. 4 shows a sectioning
procedure that would reveal the stresses parallel
to the weld along a gradient transverse to the
weld. Also, electrical-resistance strain gages
could be used instead of measuring the distance
between shallow holes or dimples.
Another approach to a constant biaxial stress
field in a flat plate, varying only through thickness, was described by Moore and Evans (Ref
2). They relied on XRD for the measurement of
the strains from which the stress was calculated,
and the procedure consisted of removing layers

0
40

10 20 20 20

50

from the surface of the plate and measuring the


strains existing at each layer. The process assumed that the stress perpendicular to the surface
was zero. It should be noted that in order for this
procedure to be valid, the areas measured by
XRD would have to be free of plastic deformation caused by layer removal (see Sectioning
and Material-Removal Methods).
Triaxial Conditions. In reality, in most components in which residual stresses have been induced, usually due to manufacturing processes,
the stress field is triaxial and varies from point
to point (element to element) in all three directions. Thus, a number of destructive procedures
and stress-field condition assumptions have been
applied in order to measure the three-dimensional residual-stress-field condition existing in
most components of practical engineering interest. Two of these are described in this section.
Chen (Ref 34) revised Rosenthal and Nortons
approach to deriving the triaxial residual-stress
condition (Ref 33) as follows. The typical
method of residual-stress measurement is by mechanically removing part of a body and measuring the change of stress in the rest of the body.
The method of Rosenthal and Norton instead
deals only with a small element that has been cut
free from a plate. The sectioning procedure consists of removal of a narrow block from a plate
with gages attached, followed by splitting the
block in half with gages attached on the top and
bottom surface of the block, and then successive
slicing of both halves from the midsection to the
outer surface, as shown in Fig. 4.
When the half-block is sliced to a thickness of
0.1 in., gages are removed and stresses are measured (at least two points) on the surface by
XRD. The basic assumptions are:
Partial stress relief occurs in the direction of

90
Initial
sectioning
270

Weld

the long axis of the block and a total stress


relief occurs in the direction transverse to the
long axis.
The small amount of stress relaxed in the remainder of the block follows a linear law
through the thickness when a thin slice of
metal is removed.
Variation of transverse stress along the axis
of the weld is small in the middle portion of
the plate weldment.

45
10

Saw cut
180
Fine cutter
(i.e., green abrasive saw 0.5 mm thick)

10

Further
sectioning

5
10

Final
sectioning

Gage

Weld
10

10
10

Rosette gage

Residual-stress measurement of a girth-welded pipe by strain gaging and sectioning. Note that strain gages
shown in the final sectioning should be placed on the pipe prior to initial sectioning; and for a more complete
analysis several of the layer sections detailed in the final sectioning step should be strain gaged and sectioned. Dimensions
given in mm

Fig. 3

Welded steel plate, the near side of which shows


the two narrow blocks suggested in Rosenthal
and Nortons procedure (Ref 33). The far side shows several
blocks sectioned to reveal the stresses parallel to the weld
with a gradient transverse to the weld.

Fig. 4

104 / Measurement and Prediction of Residual Stress and Distortion


The determination of residual stresses may be
divided into the following steps:
1. Determination of e1 and e,
t which represent
the amount of strain relaxed in longitudinal and
transverse directions by cutting one longitudinal
and one transverse block free from the plate.
This was done by determining the strain relief
between two indentations on the top and bottom
surface of each block using a mechanical gage
(see Fig. 4) and subtracting the strain-gage readings on the blocks from the initial reading from
the plate. Note that this measurement can be performed using electrical-resistance strain gages or
XRD (Ref 7, 22, 23).
2. Determination of strain relaxed on the top
and bottom surfaces by splitting each block in
half and then successively slicing the blocks.
3. Determination of the strain relieved (using
at least two points) on the top and bottom surface
using sensitive mechanical gages to measure the
distance between indentations and using XRD
after the thickness of the top and bottom halves
of the blocks had been reduced to 0.1 in. Note
that this restriction was necessary because they
are using conventional scanning x-ray instrumentation. With modern x-ray instruments, measurements can be made on any size of specimen
(Ref 7, 19, 20, 2224).
4. An enhancement of the Rosenthal and Norton procedure (Ref 33) was suggested by Chen
(Ref 34) where the determination of the residual
stress remaining in the top and bottom slices was
measured using XRD. Here the lattice strain was
measured in the remaining slices in at least two
places near the gage points using XRD, and the
absolute residual stress, S*t and S*1 remaining
was calculated as described in the section Stress
Measurement in this article. These XRD-determined stresses measured for each block were averaged and designated S*
t or S*
1 for that face. The
original stress S present before slicing was determined by:
St S*t Eet

(Eq 1)

S1 S* Ee1

The total strain e may be obtained by dividing


St or S1 by the modulus of elasticity, E.
5. Computation of stress relieved by cutting
the blocks from the plate was accomplished using the values of e and e obtained in steps 1
and 3, the amount of longitudinal and transverse
stress (St and S)
1 relaxed by cutting the blocks
from the plate:
S1

E
mE
(e1 me)
(et me)
t
1
1 m2
1 m2

St

E
mE
(et me)
(e1 me)
1
t
1 m2
1 m2
(Eq 2)

where m is Poissons ratio. Equation 2 is valid


for the case where the length of the block is at
least twice the thickness of the block. In case of
shorter blocks:

S1

E
E
(e1 me)
t
1 m2
1 m2

me
t

m 2 b2
e1
1 b

E
E
St
(et me)
1
1 m2
1 m2

me1

m 2 b2
et
1 b

(Eq 3)

where b is the correction factor for a shorter


block. When plotted for the top and bottom faces
of the block and joined by a straight line, these
values give the stress S relaxed by cutting the
block from the plate.
6. Computation of stress (S) relieved by splitting and slicing the blocks was done using:

2 d (1 ) 2(e e )

(S)
a tE

1 det

et et
2 dl
0.5 (1 l)
0

3(1 )

(5 4)et (1 )eb0
0

(Eq 4)

eb eb0
2 dl
0.5 (1 l)
a

(5 4)eb0 (1 )et 0

(Eq 5)

where l is the thickness of the block; is the


fraction of the total thickness removed; e,
t e
b are
the relaxed strain measured on top and bottom
surface for position ; and et 0, eb0 are the relaxed strain on top and bottom surface when
splitting the block in half.
7. The total stresses relaxed across the thickness of the block in the longitudinal and transverse directions were obtained by:
S1 S1 S1

(Eq 6)

St St St

8. Determination of shearing stresses in the


longitudinal and transverse directions were determined by:
sxy syx

Sw S v
2

(Eq 8)

Sy
sxy
syz

Y 0
y
x
z

(Eq 9)

Sz
sxz
syz

Z 0
z
x
y

(Eq 10)

Differentiation of Eq 8, 9, and 10 with respect


to x, y, z, respectively, yields:
2Sx
2sxy
2sxz

0
x2
xy
xz

(Eq 11)

2Sy
2sxy
2syz

0
y2
xy
yz

(Eq 12)

2Sz
2sxz
2syz

0
z2
xz
yz

(Eq 13)

Subtracting the summation of Eq 11 and 12 from


Eq 13, results in:
2Sz
2Sx
2Sy
2sxy
2
2
2 2
z
x
y
xy

1 deb
(S)
(1 ) 2(eb eb0)
a bE
2 d
3(1 )

Sx
sxy
sxz

X 0
x
y
z

(Eq 7)

where s is the shear stress and Sw and Sv are the


stresses measured in the direction w and v making angles 45 and 45, with the longitudinal axis. However, if the biaxial stress condition
is assumed, only the strain relief in one 45 direction need be determined since they would be
equal. This requires that either a block at a 45
angle to the longitudinal block be removed, or
that electric residual-strain-gage rosettes be
used, or XRD measuring at the 45 angle.
9. Determination of stress in the thickness direction was determined by the equilibrium equations:

(Eq 14)

Variation of Sx along the x-axis is small in the


middle portion of the plate; therefore, the first
term in Eq 14 can be neglected, and Eq 14 can
be approximated using:
2Sz
2
2

(Sy1 Sy0)
sx2y2
z2
Dy21
Dx2Dy2

(Eq 15)

where Sy0 is the value of Sy at the axis, Sy1 is the


value of Sy1 at a distance Dy1 on either side of
the x axis, and sx2y2 is the value of sxy at a distance Dx2 and Dy2 from the y axis and x axis,
respectively.
Therefore, the stress in the thickness direction
Sz can be obtained through double integration
with the boundary condition that Sz vanishes at
both the top and bottom surfaces. In this way,
the entire triaxial residual-stress state was determined. Note that this procedure assumes that the
residual stresses are uniform along the length of
each block.
A variation on Rosenthal and Nortons
method (Ref 33) using electrical-resistance
strain gages or a nondestructive technique such
as XRD is as follows. The largest face of the
blocks described by the thickness of the component (plate) and the longest dimension of the
blocks is divided into a two-dimensional grid of
elements. An electrical-resistant strain gage is
placed on each element, or a nondestructive measurement such as XRD is performed, and the
block is sectioned along the grid lines to produce
elements that are assumed to be stress free. Note
that if a nondestructive technique such as XRD
is used, the plastically deformed surface created
by removing the block from the original component must be removed. This is best done using
electropolishing (see the section Sectioning and

Measurement of Residual Stresses / 105


Material-Removal Methods). If electropolishing of these cut faces is done to remove the plastic deformation and resultant residual stress induced by a mechanical cutting procedure, and
XRD is applied, then the blocks need not be sectioned. The measured XRD stress will provide
the absolute residual-stress-field condition in the
block, andcoupled with the strain relieved by
the original removal of the block from the
platethe entire triaxial residual-stress condition of the plate can be obtained.
This variation of Rosenthal and Nortons
method (Ref 33) provides the information necessary to derive the biaxial stress condition of
each block, which can in turn be used to derive
the triaxial condition of the original plate. The
stresses on each of the measured faces of the
blocks must be measured in three directions to
provide the information necessary to obtain the
principal stresses in the block faces in each element. Note that the strain change caused by the
sectioning of the blocks must be added to the
strains measured in each element.
This procedure can be applied to a weldment
with a single weld through its center as described
by Rosenthal and Norton (Ref 33), or to a more
elaborate stress field, for example, where two orthogonal welds existed in the component (plate).
With a single weld, only the block intersecting
weld needs to be sectioned into elements because
the residual-stress field in the block parallel to
the weld is likely to be constant along the direction parallel to the weld.
Moore and Evans (Ref 2) proposed mathematical procedures for the reconstruction of the
original three-dimensional residual-stress fields
in cylindrical and flat plate components, and
Constantinescu and Ballard (Ref 3) proposed a
modification of Moore and Evans work. They
proposed using XRD as the measurement technique and presented stress-reconstruction equations for the following conditions:
Solid cylinder bar, rotationally symmetric

are then used to calculate the original stresses


rh(r) and rz(r), as well as the radial stress rr(r).
The theory of elasticity provides nine partial
differential equations: the three equations of
equilibrium and the six equations of compatibility. Unique solutions are possible, depending on
boundary conditions, and for the case considered
they give the following working formulas:
rr(r1)

r1

rhm(r)
dr
r

rz(r1) rzm (r1) 2

metric stresses
Hollow cylinder bar, rotationally symmetric
stresses
Flat plate, biaxial stresses
The Moore and Evans procedures were summarized in the Society of Automotive Engineers
Handbook (Ref 4) as described in the following
paragraphs.
Solid Cylinder Bar, Rotationally Symmetric
Stresses. It is presumed that the residual-stress
distribution has both rotational and longitudinal
symmetry, except near the ends where measurements are avoided. Stresses are therefore functions of the radius, r, and do not depend on the
angle, h, measured around the cylinder, nor on
the distance z taken parallel to the axis. By repeatedly removing thin concentric shells, the
stresses on the exposed surface in depth can be
obtained. The circumferential and longitudinal
measured stresses rhm(r) and rzm(r), respectively,

rzm(r)
dr
r

r1

rh(r1) rhm(r1) rr(r1)

rr(r1, h)

(Eq 17)

r1

g0(r)
2r1
2 g1(r)cos h dr
r
r

rh(r1, h) rhm(r1, h)

g0(r)
2r1
2 Re[g1(r)eih] dr
r
r

rh(r1, h) rhm(r1, h)

(Eq 19)

r1

g0(r)
6r1
2 Re[g1(r)eih] dr
r
r

(Eq 20)

(Eq 25)

2h0(r)
8r1h1
2 cos h dr
r
r

(Eq 26)

2r1
g1(r) sin hdr
r2

(Eq 27)

srh(r1, h)

r1

2r1
Im [g1(r)eih]dr
r2

(Eq 21)

(Eq 22)

where Re and Im indicate real and imaginary,


respectively, of whatever follows.
The measured stresses rhm (r1, h) and rhz (r1,
h) are related to the g and h functions by a Fourier series relation:


rhm (r1, h)

gn(r)enih


rzm (r1, h)

hn(r)enih


r1

srh (r1, h)

r1

If stresses are measured at four points around


the cylinder 90 apart, and, regardless of symmetry, the g and/or h functions from Eq 23 become:
h 0

rm0 (r) g0(r) g1(r) g1(r)

h 90

rm1 (r) g0(r) ig(r) ig1(r)

h 270 rm3 (r) g0(r) ig1(r) ig1(r)


(Eq 28)

in which the summation was expanded using


values of n 1, 0, and 1. This is a reasonable approximation, providing the stresses do
not drastically vary with h.
Adding the equations cancels the g1 and g1
terms, giving:
g0(r) 14[rhm (r) rhm (r)

r1

2h0(r)
8r1
2 Re[h1(r)eih] dr
r
r
R

r1

g0(r)
6r1
2 g1(r) cos h dr
r
r

h 180 rm2 (r) g0(r) g1(r) g1(r)

r1

(Eq 24)

(Eq 18)

rz(r1, h) rzm(r1, h)

rz(r1, h) rzm(r1, h)

where r is the original radius and r1 is the radius


at depth of interest.
Solid Cylinder Bar, without Rotationally Symmetric Stresses. Stresses are again assumed to be
independent of z but allowed to vary in the circumferential h direction. Complex variable
methods give the following general solutions:

stresses

Solid cylinder bar, without rotationally sym-

(Eq 16)

rr(r1, h)

rhm (r) rhm (r)]


2

(Eq 29)

Subtracting the equations in pairs and then adding the results cancels the g0 and g1 terms, giving:
g1(r) 14{[rhm (r) rhm (r)]
0

i[rhm (r) rhm (r)]


1

(Eq 30)

Equations 29 and 30 can be used in the general


solutions, Eq 19 to 22. However, if the stresses
are symmetric with respect to a plane through
the axis, then rhm (r) rhm (r) and Eq 30 becomes:
1

(Eq 23)

If the stresses are symmetric with respect to a


plane through the axis of the cylinder, the general solutions simplify because the g1(r) and
h1(r) functions will be real. This is a reasonable
approximation when the residual stresses are induced by heat treatment or straightening and the
equations become:

g1(r) 14[(rhm (r) rhm (r))]


0

(Eq 31)

Equations 29 and 31 are then used in Eq 24


to 27, along with similar formulas for the h function, determined from Eq 23 in the same way. It
is again interesting to note that Eq 16 to 18 are
included in the general solution of Eq 19 to 21,
when the stresses are rotationally symmetric.
The shear stress of Eq 22 or 27 exists only as a
result of asymmetry of the h stresses. When sym-

106 / Measurement and Prediction of Residual Stress and Distortion


metry is obtained, the g1(r) terms are zero, as is
srh(r1, h).
Hollow Cylinder Bar, Rotationally Symmetric
Stresses. If R1 is the inside radius of the cylinder,
similar analysis generalizes the previous equations to the following final formulas:

1 r r
R21

2
1

r1

r2
rhm(r)
dr
R12
r

rz(r1) rzm(r1) 2

(Eq 32)

(Eq 33)

r21 R21
2
2 rr(r1)
1 R1

(Eq 34)

The above equations include those of the solid


cylinder when R1 is zero.
Flat Plate, Biaxial Stresses. It is assumed that
the residual stresses in a flat plate of uniform
thickness depend only on the distance from one
of the flat surfaces of the plate, except, of course,
near the edges. It is also assumed that the principal stresses are rx and ry, lying in the plane of
the flat surfaces, and that the stress normal to the
flat surfaces, rz, is zero at all points sufficiently
distant from the edges.
From the authors assumptions and conditions
of equilibrium, the true stresses rx(z1) at depth
z1 can be expressed in terms of the measured
stress rxm (z1) by the relation:

6z1

z1

z1

rxm(z)
dz
z

rxm(z)
dz
z2

(Eq 35)

where H is the original thickness of the plate and


z1 is the distance from lower surface to uncovered depth of interest.
A similar expression holds for the y direction.
Equation 35 holds, even if rx and ry are not prin-

in.
I

II

III
I
4 ft
2 ft

1 in.

A 2 by 2 by 4 ft (0.6 by 0.6 by 1.2 m) weldment


showing the layers proposed by Johanssen (Ref
25) where the thickness (T) of the layers are 0.5 in. (12.5
mm) and rI ry, rII rx, and rIII rz.

Fig. 5

z1

z1

sxym(z)
z

sxym(z)
dz
z2

(Eq 36)

sxym (z) is determined from:

sin2 2sxym(z1) sin cos

r1

rx(z1) rxm(z1) 2

ram (z1) rxm (z1) cos2 rym(z1)

r2
rzm(r)
dr
R12
r

rh(r1) rhm(r1)

1/ 2

sxy(z1) sxym(z1) 2
dz 6z1

rr(r1)

cipal stresses, but in this case a shear stress also


exists, expressed by:

(Eq 37)

where is the acute angle that the measured


stress ram (z1) makes with the x axis. When measurements are taken 45 apart, sxym (z1) becomes:

rx

2
y2

ry

sxym(z1) r45(z1)
12[rxm(z1) rym(z1)]

Johanssen showed that the change in stresses


resulting from the removal of material can be
determined by Dr(x) and m(x). These were to be
measured at a number of positions xi, i 1,,
N, on the lower side of the plate. Drx(x) and m(x)
are the difference between the stress and deformation measured prior to and following the removal of material. Changes in the internal stress
conditions are thus calculated directly and need
not be calculated as accumulated stress changes
resulting from several layers of material being
removed.
By introducing Aireys stress function u(x, y)
that is defined as:

(Eq 38)

meaning that three x-ray stress measurements are


required after each layer removed.
Johanssen proposed a procedure (Ref 25) for
the determination of the three-dimensional residual-stress field in thick plate (plate weldments)
components using XRD techniques to measure
the strains on the surfaces of the plate and plate
sections, and upon removal of layers of surfaces.
The procedure includes the measurement of the
biaxial stress field existing on the top surface of
the component (see Fig. 5), assuming that the
stress perpendicular to the surface is zero. Material is removed from this surface by, for example, milling and electropolishing, or electropolishing alone (see Sectioning and
Material-Removal Methods) and the biaxial
stresses remeasured at the new depth. Each time
material is removed, the forces that the removed
layer exerted on the remaining component must
be accounted for, and the subsequent measurements corrected for this change in the stress field.
Johanssen based his method on the following assumptions:
When a layer of material is removed, the re-

sulting changes in the stress condition will be


linear elastic; that is, Hookes law is applicable.
The residual-stress distribution is constant in
the z-direction, except at the surface, and rz
is a principal stress in the z-direction (see Fig.
5).
Upon material removal, it is assumed that the
strain ez remains unchanged. Together with
the previous assumption, this implies that the
change in stresses can be treated as a plane
problem.
It is assumed that the stresses are symmetrical
with respect to the y-z plane. This assumption
is, however, not necessary. The procedure can
be developed to include asymmetrical stress
states.
Johanssens justification for his procedure to
measure the three-dimensional stress field in the
weldment shown in Fig. 5 is as follows (Ref 25).

2
2
and sxy
x2
xy

(Eq 39)

Equilibrium and compatibility are automatically satisfied if is a solution to the biharmonic


differential equation:
 2 2u 0

(Eq 40)

The homogeneous boundary conditions are:


ry(x, 0) 0

(Eq 41)

y 0
sxy(x, 0) 0

(Eq 42)

sxy(a, y) 0

(Eq 43)

x a
rx(a, y) 0

(Eq 44)

and at y 0 there should be agreement with


experimental measurements; that is:
rx(x, 0) Drx(x)

(Eq 45)

Pickel (Ref 35) described a method for analytical solution of problems with this type of
boundary condition, using infinite, related series;
however, in this instance an approximate method
is used with trial solutions that are both more
convenient and numerically more stable.
r
The force T is divided into Tx(x) and Ty(x) (see
Fig. 6) and as a solution to Eq 45 to 47, u is
taken as u ux uy where ux is the solution
for the loading case where only Tx(x) differs
from zero (and therefore the boundary condition
ry(x, b) 0 must be applied), and uy is similarly
the loading case where only Ty(x) is not equal to
zero. This ensures that the trial solutions will be
sufficiently general and not involve any limitar
tions on T (x).
The solutions for ux and uy are then adjusted
so that they satisfy Eq 43 and as many of the
boundary conditions as possible. By direct substitution, it is easily verified that the trial solutions in the following satisfy all the boundary
conditions except rx(a, y) 0; however, this

Measurement of Residual Stresses / 107


condition can be filled and the measurement can
be fit using an approximate method, such as the
least-squares method. If this can be performed
sufficiently accurately, these trial solutions are
close to the exact solution, since it has been
shown that the solution is uniquely determined
by the stated boundary conditions.
For the loading case Tx(x):
y3
y2
Cnx
C0x

2
6
2
n1 n

(1 nb coth nb)ny sinh ny]cos n

(Eq 46)

is valid where n np/a and 2a and 2b are the


width and thickness of the plate after removing
material.
Equation 39 gives:

exy

Cnx[(nb sinh ny ny cosh ny)


(1 nb coth nb)ny sinh ny]cos nx
sxxy

2x

xy

Cnx[(nb(coth
n1

bb 1)(2 cosh ny ny sinh ny) nb


(sin hny ny cosh ny)]coshnx
(Eq 47)

vx
y

(Eq 52)

one obtains:
(Eq 48)

v(1 v)
y2
Cx
C0xy
E
2

vx(x, y)

Cnx
n1

1v
E

(Eq 49)

Cnx nby sinh ny 2b

n1

cosh ny

v
nby sinh ny
1v

Since plane strain has been assumed xz can be


calculated from:

(1 nb coth n y cosh ny

xz m(rx ry)

1
v

sinh ny
n
1v

(Eq 50)

where m is the Poissons ratio. Further, the strain


exy can be obtained from:
exy

2x
Cxy C0x
y2

(sinh ny ny cosh ny)]sinh nx

[nb(sinh ny ny cosh ny)

rxx

2x

x2
n1

[2nby sinh ny (1 nb coth nb)

x(x, y) Cx

rxy

1 m2
m
ry
rx
E
1 m

(Eq 51)

where E is the elastic modulus.


Since the condition for compatibility is satisfied, the displacement in the y-direction is
uniquely determined, excepting rigid body displacements, from the relationship:

cos x

sinh ny y cosh ny

(Eq 53)

A similar treatment is used for the loading case


r
T (x). In that case, the solution is given by:
y3
y2
C0y

6
2

y(x, y) Cy

Cny
n2

n1

[(1 nb coth nb)(sinh ny ny


cosh ny) 2n by sinh ny]cos nx

(Eq 54)

from which are obtained:


yx

2y
Cy C0y
y2

n1

Cny[2nb cosh ny 2n by sinh ny


(1 nb coth nb) (sinh ny

Removed layer

ny cosh ny)]cos nx

x
x(x)

Ty (x)

yy

2y

x2

(Eq 55)

Cny

n1

[(1 nb coth nb)(sinh ny ny

T (x )

cosh ny) 2nby sinh ny]cos nx


Internal forces



xy
2

Tx(x)

syxy

(Eq 56)

n1

Cny[nb sinh ny n2by cosh ny)


x

x(x)

(1 nb coth nb) ny
sinh ny]sin nx

(Eq 57)

Finally, the displacement is given by:


x(x)corr

v(1 v)
y2
Ey
C0yy
E
2

vy(x, y)

1 v
E

Cny b sinh ny

n1

nby cosh ny
(x)

x(x)

v
1 v

(b sinh ny nby cosh ny)

(1 coth nb) y sinh ny


x(x)
Residual-stress distributions, forces, and distortion of a plate before and after layer removal. Top: residual-stress
distribution in the x-direction in the center of the plate in the x-z plane; center: same as top after removal of a
layer with the forces Ti(j) caused by the residual stresses tending to distort the plate; bottom: same as center with the
distortion displacement shown (Ref 25)

Fig. 6

2
v

cosh ny
n
1 v

y sinh ny cos nx

(Eq 58)

108 / Measurement and Prediction of Residual Stress and Distortion


By letting the stresses and displacements be
equal to the measured values, the following relations are obtained:
rxx(xi, 0) ryx(x2, 0) Drx(xi)
i 1, 2,

(Eq 59)

i 1, 2,

(Eq 60)

and satisfying the boundary condition stated in


Eq 44 at a suitable number of positions:
rxx(a, yi) rxy(a, yi) 0

r(z1) rm(z1) 2

m (xi, 0) m (xi, 0) m(xi)


x

using the exact equations of the previous sections. The method is summarized for two of the
previous cases as follows.
Flat Plate (see Flat Plate, Biaxial Stress). A
generalized solution is written:

i 1, 2,

(Eq 61)

Equations 53 to 61 form an overdetermined system, from which the unknown coefficients Cnx
and Cny can be determined using, for example,
the method of least squares. The desired correction of the stresses measured at the surface after
removing material can finally be expressed as:
rxcorr rxx(x, b) rxy(x, b)

(Eq 62)

z1

dz 6z1

z1

rm(z)
z

rm(z)
dz
z2

(Eq 64)

The subscripts x, y, or xy have been dropped,


because the form of Eq 35 and that of Eq 36 are
exactly the same. r(z1) represents the true stress
in any direction at depth z1 before a layer was
removed, and rm(z1) represents the measured
value at that depth.
The correction in stress c(z1) at z1 is the difference between the true and measured value,
given by:
H

c(z1) r(z1) rm(z1) 2

and

dz 6z1

z1

z1

m[rxcorr

b)

ryy(x,

b)]

(Eq 63)

Sikarskie (Ref 36) proposed a stress-reconstruction procedure (series method) when thin
layers were removed from the surface of a component. He described procedures applicable to
flat plates and solid cylinders. The procedure
works well for shallow depths (a few percent of
specimen diameter or thickness), or in instances
where the stress gradient over the total depth removed does not change too rapidly and is of essentially one sign. The practicality of this
method depends on the fit of the measured
stresses in depth by a Taylors series referred to
the surface values of stress and successive derivatives at the surface. When the method is applicable, very convenient relations are obtained,
which describe the stress correction in terms of
the influencing factors, for example, layer depth,
stress magnitude, stress gradient, and specimen
size. Judgment is necessary, however, in using
the series approximation that does not arise when

(Eq 65)

z1

Fig. 7

Stresses in flat plate after layer removal

H z1
H

H z1
H

r1

where, again, the subscripts r, h, and z have been


dropped because the form is the same.
When rr(r1) is desired:
r(r1) rr(r1)

rm(r) rhm(r)

(Eq 71)

When rz(r1) is desired,


r(r1) rz(r1)
rm(r1) rzm(r1)

c(r1) r(r1) rm(r1)

(Eq 72)

(Eq 66)

(Eq 67)

where Dz1 H z1. This correction is seen to


be approximately proportional to the magnitude
of the surface stress and thickness of the removed layer (Fig. 7). It is inversely proportional
to the specimen thickness.
By solving for Dz1, the question of proper
slice is given by:
1 Hc(z1)
4 rm(H)

(Eq 68)

Thus, for example, if the measured stress is to


be in error by less than 5%, c(z1)/rm(H)
0.05 and the appropriate slice depth is:

(Eq 73)

Again, expanding the integrand of Eq 70 in a


Taylors series and integrating term by term, a
final form for the correction is obtained:

R r1
1

R
2

c(r1) k rm(R)

R r1
R

[rm(R) Rrm(R)]
Dz1
H

(Eq 70)

rh(r1) is calculated from rhm(r1) using Eq 18.


The correction term in stress, c(r1), is written
as before:

where rm(H), rm(H), are true surface stress and


successive derivatives with respect to z at the
surface.
For shallow depths, only the first terms of the
series can be used and:
c(z1) 4rm(H)

rm(r)
dr
r

k 2

1
[2rm(H) Hrm(H)
3

2H 2rm(H)]

r(r1) rm(r1) k

rm(r) rzm(r)

[rm(H) 2Hrm(H)]

Dz1

Removed layer

The integrands are then expanded in a Taylors series referred to the surface values, after
which the integration is performed term by term.
The final form for the correction is:
H z1
c(z1) 4rm(H)
H

For a plate 0.4 in. (10 mm) thick, for example,


the slice depth is 0.005 in. (0.13 mm).
If the stress gradient is high, then the next term
in the correction series should be included and a
quadratic in Dz1 solved. This requires an estimate of rm(H) based on experience.
Solid Cylinder (see Solid Cylinder Bar, Rotationally Symmetric Stresses). A generalized
solution from Eq 16 and 17 is written:

k 1

rzcorr
ryx(x,

(Eq 69)

rm(r1) 0

rm(z)
z

rm(z)
dz
z2

1
(0.05)H
4

Dz1

1
6

[2rm(R) 2Rrm(R) R2rm(R)]

R r1

(Eq 74)

where rm(R), rm(R), and so forth, are the surface


stress and successive derivatives with respect to
z at the surface.
Insight into the factors that influence the correction holds exactly as previously discussed, as
do the limitations of the method.
Ruud et al. (Ref 2224) applied a modification of Johanssen method to measure the triaxial
stress condition of thick plate 214Cr-1Mo plate
weldments. Ruud et al. actually measured the
strains in all directions and calculated the
stresses, but did not correct for layer removal due

Measurement of Residual Stresses / 109


to the complex nature of the stress field. Ruud et
al. also measured the residual-stress condition of
expanded tubing including 304 stainless steel
tubing (Ref 37) but focusing on the residual
stresses on the inside surface of the heat-exchanger tube components.

Residual stresses parallel


to the milling direction, ksi

10

20

30
0

Distance from the workpiece surface, mils


Residual stresses at the surface and near the surface due to milling a medium-carbon steel work-

Fig. 8
piece

40
20

Residual stresses, ksi

0
20
40
60
80
100
120

10

12

14

Distance from the workpiece surface, mils

Sectioning and Material-Removal


Methods
As discussed in the previous sections in this
article, many procedures require that the component (sample or part) be sectioned and/or
some material be removed from it to measure the
residual stresses. This is especially true for the
measurement of internal residual-stress fields
where the component nearly always has to be
sectioned to reveal the internal stress field. The
two exceptions to the necessity of sectioning and
material removal, neutron diffraction and ultrasonic methods, are described in the section
Nondestructive Procedures in this article.
Sectioning or material removal is required by a
particular residual-stress measurement procedure
(see the section Stress-Field Condition Assumptions) or method (see the section StrainMeasurement Methods). Mechanical chipremoval processes are usually applied because
of their economy and speed. All chip-removal
processes, including lathe turning, drilling, milling, sawing, grinding, and so forth, introduce
surface residual stresses that can be as high as
the yield strength of the strain-hardened metal
and several thousandths of an inch (tens of microns) in depth (Ref 3843). Figures 8 to 10
show the residual stresses in steels caused by
various machining processes. Furthermore, some
steels are especially prone to strain hardening
for example, austenitic stainless steelsand extra care must be used with these materials when
selecting a material-removal technique. Figure
11 shows plots of the residual stresses in 304
austenitic stainless steel caused by various grinding methods. Note that these plots are only a
sample and may not be typical. If the size of the
element in which the strain change is measured
is smaller, or thinner in the case of surface-depth
gradients, than about 0.1 in. (23 mm), then a
chemical or electrochemical material-removal
technique must be used to remove the surface
residual stresses caused by mechanical chip re-

Residual stresses in a 440 C (825 F) stainless


steel workpiece induced by facing

Fig. 9

100
80

20
Residual stresses, ksi

Residual stresses, ksi

0
20
40
60
80

Strain-Measurement Methods

Abusively
machine
ground

Handheld grinder
60

As discussed in the section Stress Measurement, all residual-stress-determination methods


measure elastic strain, not stress, and the residual
stress is calculated from the strain values. Several methods for the measurement of strain have
been applied in residual-stress studies and have
been mentioned previously. These methods include:

40
Handheld
flapper
wheel

20
0

100
120

20
140

10

12

14

Distance from the workpiece surface, mils

Distance from the workpiece surface, mils

Fig. 10

Residual stresses in an alloy steel workpiece


induced by turning

Samplings of residual-stress distributions induced in 304 stainless steel workpieces by


common grinding procedures

Fig. 11

moval. These techniques may be used solely or


in conjunction with, and after, the chip-removal
method. It should be noted that material-removal
techniques such as electrodischarge milling
(EDM) induce residual stresses (Ref 44) as do
chip-removal methods. Other methods, such as
laser, flame, or plasma cutting that cause heating
of the element, must be applied with caution because they may reduce the stress field by annealing before it can be measured.
Thus, the only methods for material removal
from a component surface that do not induce residual stresses are electrolytic or chemical polishing. Electropolishing is described in some detail in Ref 45, and guidelines are provided for
application to various ferrous alloys. In electropolishing, the electrolyte and operating conditions depend on the alloy being polished, as
shown in Ref 4 and 45. Electropolishing combined with XRD is used extensively to reveal
residual-stress gradients on machined, ground,
and hardened surfaces. However, application of
these techniques requires that the subsurface
stresses be corrected for the removal of prior surface layers (Ref 4).
Another concern when reducing components
to a more convenient smaller size to place them
on or in a measurement device is that the stresses
of interest are not changed by the sectioning.
Generally, plates should be cut to a length and
width of at least three times the thickness to
avoid end effects. Cylinders, both thin-walled
and solid, should be a minimum of three diameters in length.
Where the manufacturing process affects the
entire thickness of a component, such as heat
treatment or forging, it might not be advisable
to section without means of measuring the stress
change extensively over the entire component
before sectioning. In other words, selection of
the stress-measurement procedure and methods
should avoid sectioning unless techniques to
measure the effects of sectioning are applied before sectioning is initiated. On the other hand,
when the processes that have induced the residual stress produce only shallow stress fields, then
the three-times rule suggested previously is applicable.

Mechanical gages
Electrical-resistance gages
Optical gages
Birefringent methods
Diffraction methods (x-ray and neutron)
Ultrasonic methods
Magnetic methods

110 / Measurement and Prediction of Residual Stress and Distortion


Diffraction, ultrasonic, and magnetic methods
are discussed in the section Nondestructive Procedures in this article.
Mechanical Gages. The application of mechanical gages such as those described by Heyn
(Ref 26), Stablein (Ref 27), Gunnert (Ref 31),
and others generally preceded the availability of
electrical-resistance strain gages and are not discussed due to their general lack of precision,
poor spatial resolution, and inefficiency.
Electrical-Resistance Strain Gages. Most
bonded electrical-resistance strain gages are
made from either metallic wire or foil materials.
There are also the recently developed semiconductor gages. A variety of sizes, shapes, and configurations are available including single-element gages and rosettes with two, three, or four
elements. Electrical-resistance strain gages are
available in sizes as small as about 1 mm and
thus provide a resolution of strain measurement
on that order. Information on electrical strain
gages is available in numerous sources including
Ref 46 and in reviews by Crites (Ref 47) and
Masubuchi (Ref 14) as well as suppliers (e.g.,
Measurements Group Inc., Raleigh, NC).
Changes in temperature tend to cause an apparent strain; therefore, some type of temperature compensation is needed. Frequently, a
dummy gage, which is not subjected to the
strain, is exposed to the same temperature as the
actual gage to provide a basis for comparison. A
temperature-compensated gage can also be used.
Gages must be bonded securely to the specimen. Various types of cements have been developed. Sometimes gages must be protected
from metal chips produced during machining as
well as from the oil or water. A number of systems have been devised for protecting gages under various conditions.
Brittle coatings are a simple inexpensive
strain gage that will only provide qualitative indications of residual stress. A brittle lacquer is
applied to the area where the stresses are to be
measured by material removal. After the lacquer
has cured (dried), a change in the stress field is
induced, and if the change is sufficient strain will
be produced in the lacquer, causing it to crack.
If the material removal is in the form of a hole
drilled in the lacquer, radial cracks indicate a tensile residual stress in the plane of the component
surface and circular cracks indicate compressive
stress.
Optical Gages. Light reflective methods can
be used to magnify the movement of a reflective
surface in a well-fixtured component that is held
securely in place during material removal to
change the stress field and therefore induce strain
change. This method can also be used if the component can be removed and replaced precisely in
a fixture and the position of the reflected light
measured before and after removal and replacement during which a change in the residualstress field was induced.
Laser Methods. Other techniques applying laser light have been proposed as well. These have
included shearography (Ref 48), interferometry
(Ref 49), speckle-correlation interferometry (Ref

50), and others. Vikram and others (Ref 50) suggested that a small volume of the material could
be stress relieved by heating via a laser to induce
a change in the stress field and the strain change
measured by an optical technique to reveal the
residual stresses existing in the volume before
heating. However, it must be recognized that
heating a volume of steel sufficiently to change
the residual-stress field will result in tensile residual stresses in the heated volume (Ref 51),
and this would likely be detrimental to the component in which stresses were being measured.
Birefringent Methods. Under the action of
stresses, transparent materials become doubly refracting (birefringent), and if a beam of polarized
light is passed through a model (under stress)
made of such a material, a colored picture is obtained from which the stress distribution can be
determined. This technique is called the photoelastic technique (Ref 46). The actual component
in which the stresses are to be measured is coated
with a photoelastic polymer. When residualstress changes are induced in the component,
strain changes are caused and transmitted to the
polymer coating, which then becomes birefringent. This can be observed and measured using a reflection polariscope (Ref 14). Instructions
for analyzing fringe patterns in this application
(nearly the same as those obtained in ordinary
photoelasticity) are provided by the manufacturer.
The photoelastic coating may be applied by
brushing a liquid polymer on the surface of the
specimen and polymerizing it by applying heat.
Alternatively, a prefabricated flat or contoured
sheet of polymer may be bonded to the part at
room temperature (Ref 52). The maximum strain
that can be measured ranges between 3 and 50%,
depending on the type of polymer used; the
strain sensitivity usually decreases with the increase in the maximum measurable strain.

Chemical Methods
A number of qualitative methods to detect residual stresses that may lead to stress-corrosion
hydrogen-induced cracking in steel have been
applied to specimens representing components
to be manufactured. Magnesium chloride solutions have been applied extensively to the study
of stress corrosion in austenitic stainless steels,
including some recent work by Bouzina et al.
(Ref 53).
Masabuchi and Martin (Ref 54) studied the
susceptibility of SAE 4340 steel weldments to
hydrogen-induced stress cracking. The test procedure was to immerse the weldment specimens
in a 4% H2SO4 aqueous solution charged with
H2 and to which two drops of a 5% P solution
of CS2 was dissolved. A direct current (dc) was
applied between a specimen and a lead anode to
provide a current density of 0.35 to 0.8 A/in.2
The crack patterns that developed were related
to the surface tensile residual-stress distribution
in each specimen.
Stress-corrosion cracking induced by residual
stresses in carbon and low-alloy high-strength

steels has been investigated by several researchers (Ref 5456). One procedure consisted of immersing the specimens in a boiling aqueous solution of 60% Ca(NO3)2 and 4% NH4NO3 for 31
h. The crack patterns that developed were related
to the surface tensile residual-stress distributions
in the specimen.
A number of standard practices for testing the
susceptibility of metals to stress-corrosion cracking have been published by ASTM, including:
ASTM G 38 Standard Recommended Prac

tice for Making and Using C-Ray Stress Corrosion Test Specimens, 1979
ASTM G 58 Standard Practice for the Preparation of Stress Corrosion Test Specimens
for Weldments, 1978
ASTM G 39 Standard Practice for Preparation and Use of Bent-Beam Stress Corrosion
Test Specimens, 1979
ASTM G 30 Standard Recommended Practice for Making and Using U-Bend Stress
Corrosion Test Specimens, 1979
ASTM STP 425 Stress Corrosion Testing,
1967

However, these tests for the most part do not


reveal the residual stress, but rather the susceptibility of the metal to cracking under known
stresses in the specified corrosion medium.

Semidestructive Procedures
Nondestructive methods of residual-stress
measurement are characterized as methods that
in no way affect the serviceability or reduce the
mechanical strength or other properties of the
component in which stresses are measured. Between the nondestructive and destructive methods, which have a severe effect on the serviceability, strength, and properties, are the
semidestructive methods. These are methods
that have a small to negligible effect on the components in which stresses are measured, or methods in which the component may be repaired after the measurement.
The methods that are considered semidestructive are those that require small holes to be
drilled or rings to be trepanned in the component
or indentations to be made in the surface. The
first two methods provide quantitative data and
the third only qualitative data.

Blind Hole Drilling and Ring Coring


The hole-drilling method was proposed in the
1930s (Ref 57) and is based on measurement of
the change in surface strain caused by stresses
relieved by machining a shallow hole in the testpiece. The principle is that removal of stressed
material results in the surrounding material readjusting its stress state to attain equilibrium.
The method has been standardized in ASTM E
837 (Ref 58). The ring-core method (Ref 59) is
also based on the strain caused by disturbing the
stress field, but in this case a relatively stressfree island of material is isolated by making a

Measurement of Residual Stresses / 111


shallow ring around a strain gage. This method
is also called trepanning. These two methods are
the least destructive mechanical stress-relief
techniques and are relatively simple and economical. They, as do nearly all stress-relief techniques, rely on electrical-resistance strain gages
to measure the strain change due to metal removal. Rosettes of strain gages are available especially for hole drilling. The size of the rosettes
has been progressively reduced over the last few
decades and are now available in sizes less than
about 10 mm from a number of manufacturers.
As with most residual-stress techniques, hole
drilling and ring coring have been applied
mostly to steels. Most applications have been
done on flat plate or cylindrically round parts
(Ref 5, 60, 61). Stresses can be determined at
various depths into the surface of the material,
down to a depth equal to the diameter of the hole
or core (Ref 62, 63). Kelsey (Ref 64), however,
observed that stresses with depth cannot be measured accurately to greater than half the hole diameter. The thickness of the layers in which
measurements are resolved is about 10 to 20%
of the hole or core diameter.
The equipment necessary to perform the measurement is reasonably inexpensive, portable,
and can be used in a manufacturing shop environment. However, experienced technologists
are needed to take the readingsfrom selecting
the area in which stresses are to be measured to
preparing the surface, applying the strain gages,
and reading and interpreting the data. Due to the
possibility of residual stresses being induced by
the hole-drilling or coring technique, prior calibration of the application is recommended in all
cases, with the possible exception of certain applications where holes are produced by abrasive
jet machining (AJM) (Ref 65).
Rendler and Vigness (Ref 66) developed calibration constants for cold-rolled steel that they
proposed as generally applicable to all metals
provided that the elastic constants were known.
However, they seem to have overlooked variations in the strain-hardening coefficients and the
accompanying residual stress, which exist between alloys and even between tempers of the
same alloy. Dini et al. (Ref 67) showed that direct experimental determination of the necessary
constants for any isotropic material with known
elastic constants can be eliminated by using data
available for cold-rolled steel and calculating
these constants using a formula presented. Despite the success that some researchers have
claimed in circumventing the development of
calibration constants, experimental calibration is
strongly recommended. This is best done by applying strain gages and drilling the holes in testpieces prior to stressing them known amounts
(Ref 65, 66). The AJM technique should be applied to any material with high propensity to
work-hardening during machining, for example,
austenitic stainless steel (Ref 68).
The following are general limitations and/or
concerns of hole drilling and ring coring:
Areas of high stress gradients should be
avoided because the stress gradient must be

assumed to be constant across the hole or ring


diameter.
Areas where stresses are greater than onethird the yield strength of the material are
likely to produce erroneous results due to local plastic yielding during metal removal.
The thickness of the part or specimen must
be at least four times the hole or core diameter.
Strain hardening of the steel in the vicinity of
the hole may result during metal removal,
which can cause tens of ksi (10 ksi 69
MPa) error.
Heating may result during the metal removal.
Holes or cores must be spaced at least eight
times their diameter apart.
The area in which stresses are to be measured
must be accessible to a rather bulky drilling
or coring alignment device.
Preparation of the surface for strain-gage adherence may induce residual stresses that introduce substantial error to the subsequent
measurement (Ref 69).

In conclusion, the drilling and ring-coring


methods are nearly nondestructive variations of
the destructive mechanical stress-relief techniques and require only rather simple equipment
and instrumentation. The state-of-the-art is relatively well developed compared to many nondestructive methods, some of which require considerable research and development work before
they will ever be suitable to general application
in terms of alloys and stress-field conditions.
Technological advancements in hole drilling and
ring coring have largely been due to advancements in the more general area of mechanical
stress-relief methods and research in new metalremoval techniques for metal fabrication.

Indentation Methods
Engineers and scientists have proposed the
use of indentors, such as those used to perform
hardness measurements, as a means to measure
or detect surface-residual stresses since the
1930s. Kokubo in 1932 reported that stresses applied under bending load changed the apparent
Vickers hardness values in carbon-steel rolled
sheets, both as-rolled and annealed (Ref 70). He
showed that tensile stresses tended to decrease
the apparent hardness and compressive stresses
tended to increase the hardness. The stresses applied in tension and compression were sufficient
to cause 0.3% strain.
In the 1950s, Sines and Carlson (Ref 71) proposed a method that required various amounts of
external loads to be applied to the component in
which residual stresses were to be measured
while hardness measurements were made. The
loads were made to cause both tensile and compressive applied stresses. The quality, that is,
whether the residual stress was compressive or
tensile, was then revealed by comparing the effect of the applied stress, and whether it was tensile or compressive, on the hardness measurement. At about the same time Pomey et al. (Ref

72) proposed that residual stresses could be measured by pressing a ball-shaped penetrator into
the component in which residual stresses were
to be measured and establish the relationship between the pressing load while it was progressively increased, and the electrical resistance at
the interface between the penetration and the
component. He maintained that a smaller decrease in electrical resistance indicated that portions of material under the ball were plastically
yielding and that the corresponding load on the
ball could be related to the existing residual
stress.
Later, Chiang et al. (Ref 73) provided a critique of several existing indentation analyses and
proposed an interpretation of indentations exhibiting hemispherical plasticity. Nevertheless, the
applications illustrated in this paper focused on
brittle materials and not metals.
There have been numerous papers published
proposing various approaches to interpreting the
indentation loads and shapes so as to estimate
the residual-stress field on the surface and nearsurface regions of materials. However, indentation methods have not earned the degree of confidence of XRD or hole-drilling methods for
general applications and thus are rarely applied.

Spot Annealing
Another method that has been proposed to
measure residual stresses in metal surfaces is to
reduce the residual stresses in a small volume by
annealing the metal in the volume. It has been
proposed that this annealing be performed by intense laser light (Ref 50). This technique was
envisioned to be similar to residual-stress relief
by removal of the material as accomplished in
the hole-drilling techniques. However, as Cullity
discussed in Ref 51, such localized heating
would induce high surface residual tensile
stresses in the heat-affected region.

Nondestructive Procedures
The methods described in the section StrainMeasurement Methods all measure the change
in some dimension (strain) of the component
produced by the removal of a finite volume of
stressed metal from that component. Thus, they
measure the strain induced by removing material
so as to perturb the residual-stress field. On the
other hand, nondestructive procedures measure
a dimension in the crystal lattice of the metal or
some physical parameter affected by the crystal
lattice dimension. Whenever a mechanical force,
resulting in stress that is less than the yield
strength, is placed on a solid metal component,
that component distorts and strains elastically.
That elastic strain results in a change in the
atomic lattice dimension, and this dimension, or
change, is measured by the nondestructive
stress-measurement procedures. For example,
the diffraction methodsx-ray and neutron
measure an actual crystal dimension, and this

112 / Measurement and Prediction of Residual Stress and Distortion


dimension can be related to the magnitude and
direction of the stress that the metal is subject to,
whether that stress is residual or applied. Subsequently, in this section the following methods
of stress measurement are described:

X-ray diffraction
Neutron diffraction
Ultrasonic velocity
Magnetic Barkhausen noise

X-Ray Diffraction
X-ray diffraction techniques exploit the fact
that when a metal is under stress, applied or residual, the resulting elastic strains cause the
atomic planes in the metallic crystal structure to
change their spacings. X-ray diffraction can directly measure this interplanar atomic spacing;

Ns

X-ray source

S1
Film

Np1
Np2

S2

Specimen
surface

O
2

One-angle arrangement for the single-exposure technique. N, specimen normal; b, angle


that the incident beam makes with Ns, Np1, and Np2 the
normals to the diffracting planes 1 and 2, respectively; w1
and w2, angles between Ns, Np1, and Np2, respectively; g,
the angles between the incident beam and the diffracting
plane normals; Ro, camera radius; O, point of incidence of
the x-ray beam of the specimen; 1 and 2, diffracting planes
at various attitudes to the specimen surface; S1 and S2, measured parameters representing the distance from a reference point of known distance from the incident beam and
the diffracted x-ray beam position. S1 and S2, directly related to the Bragg angles, h1 and h2.

Fig. 12

Miniature x-ray diffractometer for the one-angle technique arrangement of XRD stress measurement. This device incorporates a Ruud-Barrett Position-Sensitive Scintillation Detector and is capable of being
inserted in a 4 in. (100 mm) inside diameter and measuring
residual stress (Ref 22).

Fig. 13

from this quantity, the total stress on the metal


can then be obtained.
Since metals are composed of atoms arranged
in a regular three-dimensional array to form a
crystal, most metal components of practical concern consist of many tiny crystallites (grains),
randomly oriented with respect to their crystalline arrangement and fused together to make a
bulk solid. When such a polycrystalline metal is
placed under stress, elastic strains are produced
in the crystal lattice of the individual crystallites.
In other words, an externally applied stress or
one residual within the material, when below the
yield strength of the material, is taken up by interatomic strain. X-ray diffraction techniques
can actually measure the interatomic spacings
that are indicative of the elastic strain in the specimen. Stress values are obtained from these elastic strains in the crystals by knowing the elastic
constants of the material and assuming that stress
is proportional to strain, a reasonable assumption
for most metals and alloys of practical concern.
Reference 74 describes XRD and instrumentation in detail.
There are three basic techniques for measuring
stresses, based on XRD. They are the doubleexposure or two-angle technique (DET), the single-exposure or one-angle technique (SET), and
the sin-square-psi or multiangle technique. The
angle of exposure referred to is that between the
incident x-ray beam and the specimen surface
normal. It should be noted that in any XRD
stress-measurement technique, x-ray peaks in the
far back-reflection range, that is, peaks with
Bragg angles (h) of near 90, are much preferred
because they show the greatest effect with a
given amount of applied or residual stress. This
is illustrated by:
r (2h1 2h2)

cot h1
E
1
p
2
1 m sin2w 180
(Eq 75)

where h1 is the Bragg angle of the planes diffracting at w1 and h2 is the Bragg angle of the
planes diffracting at w2. As h1 increases, its cotangent decreases; therefore, a larger difference
(2h1 2hw) would result from a given ru to
maintain an equality.
For a residual-stress measurement, the diffracting angle, h, of interatomic planes of at least
two different psi (w) angles with respect to the
surface normal must be measured (Fig. 12).
These planes are crystallographically equivalent
(same Miller indices, hkl ) and in the unstressed
state of the metal would have the same interatomic (d) spacing for the planes labeled 1 and
2 in Fig. 12 (Ref 51, 75, 76). In a stressed material, however, the two or more orientations of
diffracting planes are selected so that they are at
different angles to the surface; thus their normals
are at different psi (w) angles to the surface normal. Then, depending on the angle of these
planes to the stress vector, their interplanar
atomic spacing is increased or decreased by
varying amounts.

The most common sources of errors and misapplications in stress measurements by x-rays
are related to stress constant selection, focusing
geometry, diffracted peak location, cold-working crystallography, texture, grain size, microstructure, and surface condition. The source, significance, and correction techniques for these
errors will not be elaborated upon here; details
may be found in Ref 77.
A point of interest in the error sources concerns cold working and microstresses. As stated
previously, microstresses are usually considered
to be those manifested by strain variation across
single metallic grains. This strain variation is detected in XRD as broadening of the x-ray peak
a distinctly different phenomenon from the peak
shift caused by residual stresses. However, microstrain variation can be measured simultaneously with stress. This microstrain phenomenon has been proposed as a means of judging
cold work, dislocation density, and fatigue damage (Ref 78).
Despite the facts that x-rays provide stress
readings only to a depth of less than 0.001 in.
(0.025 mm) and that the error sources listed previously must be considered, the noncontact XRD
method is presently the only time-proven, generally applicable, truly nondestructive method
for measuring residual stresses. Its reliability has
been proved and documented by thousands of
engineers and scientists since the 1960s, beginning with Bolstad et al.s classic work at Boeing
using x-ray film cameras (Ref 79). This documentation includes measurement of stresses in
the Brooklyn Bridge (Ref 20) and tempering
evaluation of carburized steels. The Society of
Automotive Engineers considers the method of
sufficient practical importance to have printed
three handbook supplements on the subject (Ref
4), and another is under revision.
Even so, this nondestructive technology has
been largely restricted to the laboratory because
of the general lack of knowledge regarding the
state-of-the-art instruments and the limitations of
the more widely known and available conventional scanning XRD equipment. Instrumentation for bringing this technology into the field
and manufacturing area has advanced rapidly
since the 1980s, especially toward increased
portability, compactness, and speed of operation.
As shown in Fig. 13 and 14, instrumentation
has been developed and is commercially available for stress measurement in situ on the inside
diameter of 10 mm (4 in.) diam pipe, and soon
for 5 mm (2 in.) diam steel tube (Ref 22, 23).
Position sensitive x-ray detectors have been
largely responsible for these improvements to
both laboratory-based and field-deployable residual-stress-measuring instruments (Ref 9, 19,
2123, 8082). Also, with the speed of data collection being less than 0.1 s with conventional
x-ray tube sources, XRD stress measurement can
be performed on moving components (Ref 11).
Nevertheless, many engineers have been frustrated in applying XRD to residual-stress measurement. This has been largely due to crystallographers inexperienced in residual-stress

Measurement of Residual Stresses / 113


measurement, attempting to apply conventional
scanning x-ray diffractometers and techniques to
residual-stress measurement. For example, in
conventional XRD analysis and crystallography,
sharp resolution of the diffracted spectra is very
beneficial. However, in XRD stress measurement the need to measure psi (w) angles that are
not zero defocuses the beam, and attempts to refocus leads to significant error in the stresses
read (Ref 83). In XRD stress measurement, the
ability to repeatedly measure the position of a
defocused diffracted x-ray peak is more important than sharp resolution (Ref 84). Thus, it is
recommended in most cases that XRD residualstress measurement be performed by trained
technologists using x-ray instrumentation specifically designed and built for stress measurement, not conventional scanning diffractometers.
Software packages specifically for residualstress measurement used with conventional
scanning diffractometers do not in most cases
eliminate the mechanical and focusing problems
of applying these instruments to residual-stress
measurement. It is necessary to mount the component (or specimen) in which stresses are to be
measured on the conventional scanning diffractometer, which usually requires sectioning of the
component and which complicates and adds error to the measurement procedure.

Neutron Diffraction
Neutron diffraction (ND) is capable of measuring the elastic strains induced by residual
stresses throughout the volume of relatively
thick steel components with a spatial resolution
as small as 1 mm3. Such capabilities provide for
residual-stress measurement inside components
without sectioning or layer removal. Principal
ND methods, as with XRD methods, measure the
spacing between crystallographic planes in a
component, and this spacing is affected by residual and applied stress. The spacing between a
selected set of crystallographic planes (u) is related to the angle of incidence and diffraction of

21/ 4 in. (5.7 cm)

the neutron radiation (h), which are equal, and


the wavelength of the monochromatic radiation
(k) by Braggs Law:
k 2d sin h

(Eq 76)

The elastic strain (e) induced by the residual


stress perpendicular to the diffracting crystallographic plane then is related to d by:
e

d do
sin h0

do
sin h sin h0

(Eq 77)

where do is the distance between the unstressed


crystallographic planes. If the orientation of
principal stresses are known in the component,
the stress in any principal direction may be calculated by:
rA

E
(1 m)

m
(eA eB eC)
(1 2m)

(Eq 78)

where rA, rB, and rC are the principal stresses,


eA, eB, and eC are the strains measured in the
corresponding principal directions, and E and m
are the Youngs modulus and Poissons ratio,
respectively. If the principal stress directions are
unknown, strains in at least six directions must
be measured to determine the residual stresses
acting on the volume of material in which strains
are being measured.
Limitations and Applications. For residualstress measurements in most alloys, especially
steels and cast irons, the unstressed spacing (do)
between crystallographic planes at the exact
point of strain measurement is not known and
not easily measured. This means that do or ho in
Eq 77 cannot be precisely established, and this
leads to various degrees of error in the accuracy
and precision of ND residual-stress measurements. This condition is aggravated by the fact
that the elemental composition, and thus do, vary
considerably within a component and markedly

51/ 16 in. (12.9 cm)


41/ 8 in. (10.5 cm)

within the phase (e.g., martensite, austenite, ferrite) of the alloy at various locations. Additional
limitations are that the component must be
brought to a nuclear reactor, each strain measurement requires over an hour, a single stress
determination in one small volume of the component requires at least three strain measurements, and the measurements are very costly.
Nevertheless, the ND methods have been applied
to residual-stress measurements in steel weldments (Ref 85), cylindrical steel forgings, plastically deformed steel plate (Ref 86), steel rocket
case forgings (Ref 87), and many other types of
components.

Ultrasonic Velocity
The principle underlying the measurement of
stress and thus elastic strain by ultrasonic (acoustic) techniques is the phenomenon of an approximately linear change in ultrasound velocity with
applied stress. In addition, it has been shown that
under certain narrow conditions residual stress
can be measured by exploiting this phenomenon.
Stress is measured by inducing a sound wave in
the frequency of several megahertz into the
metal specimen and measuring the time of flight
or some other velocity-related parameter. Since
many other characteristics of metals besides
stress-induced elastic strain affect velocity, these
must be sorted out; however, neither the technology nor the fundamental knowledge for such
sorting is usually available. The great interest in
ultrasonic techniques for residual-stress measurement stems from their promise for three-dimensional nondestructive measurements within
the material.
Principle. A number of velocity-related phenomena have been used in various methods to
measure stress effects by ultrasound. All utilize
the deviation of the reaction of the metal from
the linearity of Hookes law of elasticity, r
Me, where r is stress, e is strain, and M is elastic
modulus. This has been referred to as the anharmonic property of the solid and may be represented by a power series r Me Ce2 De3
, where C is the third-order anharmonic
constant, D is the fourth, and so on. Most research done for stress measurement has used expressions in which terms past the third-order
constant, C, are dropped. Of the several anharmonic property effects that may be used to measure stress, the following are probably the most
exploited:
Velocity dependence on the elastic modulus
Dispersion of frequency amplitudes in sur-

face waves

11/ 4 in.
(3.3 cm)

Birefringence of orthogonally polarized shear

waves

Harmonic generation in surface waves


1.6 in. (12.9 cm)

Fig. 14

Engineering drawing of the miniature XRD stress measurement head showing major dimensions

A very simplified form of the anharmonic


stress strain law has been written as r Me
Ce2 and rewritten as r e(M Ce). The term
in parentheses is approximately related to the velocity of sound as qV 2 M Ce, where q is
the density of the medium and V is the velocity

114 / Measurement and Prediction of Residual Stress and Distortion


of sound. This may be approximately rewritten
in terms of velocity dependence on strain as:
V

q 1 2M e
M

(Eq 79)

Then, to solve for strain, e 2(V Mq


2M)/C (Ref 88).
A simple view of the dependency of ultrasonic
velocity on the elastic modulus and density may
be shown by rewriting the equation pV 2 M
Ce in terms of V, differentiating, and dividing
by V to yield an expression for DV/V. The result
will readily show that a fractional change in elastic modulus or density would affect the velocity.
The density of metal, for which the Poisson ratio
is near 0.3, obviously changes as a compressive
or tensile stress is placed on the specimen, and
it is reasonable that the speed of sound would
then change.
Limitations and Applications. Ultrasonic
technology offers a number of types of wave
modes in which to probe metals; these include
bulk waves, such as longitudinal and shear, and
surface waves usually confined to Rayleigh type.
Each mode offers many unique parameters for
extracting information. As has been discussed,
the primary effect of stress-induced strain on ultrasonic propagation in metals is on velocity.
This may be detected in a number of ways, including measurements of wave velocity, shear
wave birefringence, and dispersion. However,
there are other characteristics of metals that affect the ultrasonic velocity to the same degree as
stress. These include crystallographic texture,
microstresses, multiple phases, coherent precipitates, composition gradients, and dislocation
density and distribution.
Crecraft (Ref 89) discussed velocity effects,
manifested as texture-induced birefringence, and
the marked change seen with ultrasonic frequency. He also reported birefringence due to
cold-work in nickel-steel specimens, but did not
attempt to separate the cold-work effects in
terms of texture, dislocation density, and so
forth. In the early 1950s, Bradfield and Pursey
(Ref 90) and Pursey and Cox (Ref 91) reported
showing the influence of small degrees of texture
on ultrasonically measured elasticity in polycrystalline bars. They showed how the true isotropic elastic constants can be determined using
measurements of both longitudinal and shear
wave speeds along several directions. They presented stereographic charts that illustrated the relationship between elastic behavior of cubic
crystals and results of x-ray texture determinations.
McGonagle and Yun (Ref 92) noted the coldwork effects in a paper comparing XRD results
with Rayleigh wave velocity measurements. Boland et al. (Ref 93) also recognized that other
material properties can affect ultrasonic velocity
and recommended that methods be developed to
distinguish stress-induced velocity changes from
those from other sources.
James and Buck (Ref 94) pointed out that
since the third-order elastic constants for most

structural materials are not readily available


from the literature, ultrasonic stress measurement must be calibrated relative to the particular
material being investigated. In the same paper
they discounted the possible effect of mobile dislocations on the sound velocity in structural engineering metals with high yield strengths due
to the short dislocation loop lengths prevalent.
However, they did mention that crystallographic
preferred orientation (texture) during deformation or fatigue is capable of severely modifying
the elastic constants on which the sound velocity
depends.
Papadakis (Ref 95) noted marked velocity
changes for ultrasonic waves in various steel microstructures, and Moro et al. (Ref 96) measured
the effect of microstructural changes caused by
tempering on the ultrasonic velocity in low-alloy
steel. Tittman and Thompson (Ref 97) evaluated
the near-surface hardness of case-hardened steel
with Rayleigh waves; because hardness in this
case is a combination of composition, microstress, and macrostress, the velocity change was
due to a combined effect.
The temperature sensitivity of ultrasonic
stress measurements has also been cited as an
important source of error. Salama et al. (Ref 98,
99) proposed that this dependence be used as a
means to measure stress, but also noted the
marked effect of dislocations and did not address
a methodology of separating the stress from the
dislocation effect.
Much of the work cited above is concerned
with attempts to measure the effects of a variety
of material properties on the changes in ultrasonic velocity. However, there apparently is no
comprehensive study that demonstrates the capability of quantitatively separating stress effects
on ultrasonic propagation from other variables
found in structural metals, such as dislocation
density or crystallographic texture. Furthermore,
most of the studies cited observed velocity
changes in bulk waves. Velocity measurements
on these waves must be measured through the
thickness of a component, and, as most metallurgists recognize, obtaining uniform properties
through thicknesses greater than a few millimeters, especially in steels, is difficult. The subtle
property variations to which ultrasound velocity
is sensitive and the inherent lack of homogeneity
in engineering metals present additional serious
problems for through-thickness stress measurements.
In spite of the microstructural variations in
manufactured steel products, success in the application of ultrasonic methods to residual-stress
measurement has been achieved in specific
cases. One is in the measurement of hoop
stresses in railroad wheels (Ref 100). Here
changes or variations of the residual stress in the
hoop direction is of concern while that in the
radial or axial direction can often be assumed to
be constant or negligible. Some techniques then
for the measurement of the residual hoop stresses
has relied on normalizing the hoop velocity
against the axial (Ref 101). Also, European railroads have monitored ultrasonic velocity along

the wheel rims during use and attributed changes


to residual-stress changes (Ref 102). Schramm
in his paper mentioned a number of approaches
for the application of ultrasound to the measurement of residual stresses in railroad wheels, and
these examples may find application in the measurement of residual stress in other axially symmetric shapes (Ref 100). Ultrasonic residualstress measurements have also been applied to
rails as reported by Egle and Bray (Ref 103) and
Bray and Leon-Salamanca (Ref 104).

Barkhausen Noise Analysis


The Barkhausen noise analysis technique
(BNA) is concerned with measuring the number
and magnitude of abrupt magnetic reorientations
made by expansion and contraction of the magnetic domains in a ferromagnetic metal. These
reorientations are observed as pulses somewhat
random in amplitude, duration, and temporal
separation and therefore are roughly described as
noise.
Applications. A few applications of BNA to
ferromagnetic metallic components have been
made, Gardner (Ref 105) mentions a number of
applications that include helicopter rotor blade
spans, autofrettaged gun tubes, gas turbine engine components, and rolling-element antifriction bearing components. In these examples, the
change in residual stresses caused by known service histories was measured.
Chait (Ref 106) qualitatively measured the residual-stress condition of a high-hardness laminar composite steel weldment and compared
some of the BNA data with XRD stress readings.
Sundstrom and Torronen (Ref 107) applied their
BNA method to a number of microstructural
measurements, including evaluation of grainsize measurement for low-carbon ferritic and ferritic-pearlitic steels, evaluation of anisotropy in
deep drawing and textured steels for electrical
applications, measurement of the degree of aging
in rimmed carbon steels, and pearlite morphology in steel wires. They have also measured iron
loss in magnetic material used for transformers
and have proposed using BNA for residual-stress
measurements, pointing out that quantitative results can be obtained if the material and its fabrication history are known and calibration is possible.
Most studies and applications of BNA to
stress measurement have focused on the uniaxial
stress state. However, Sundstrom and Torronen
(Ref 107) implied that the instrumentation they
used could simultaneously measure stress in two
directions to give biaxial stress conditions for
magnetic inspection of roller-bearing components, including BNA for monitoring residualstress change.
Summary. The BNA method certainly has
been demonstrated to be sensitive to the stress
condition in ferromagnetic materials (Ref 108).
Nevertheless, its possibilities for application are
limited by the condition that the material must
be ferromagnetic, the narrow total range of stress
sensitivity (i.e., 40 ksi, or 5.6 MPa), and the

Measurement of Residual Stresses / 115


shallow depth of measurement. The latter condition might be relieved by using magnetomechanical mechanical acoustic emission (MAE)
(Ref 109), an ultrasound analog to BNA. However, the sensitivity of either of these techniques
to other characteristics of metallic components
and the consequent need for calibration with a
nearly identical specimen severely restrict the
general applicability of BNA and MAE. Many
misapplications have been made that have severely damaged the reputation of the BNA methods (Ref 107, 110). Such restrictions can be removed only if the basic phenomena responsible
for the effect of microstructural properties on
BNA and MAE are understood and quantified in
terms of the signal.
Barkhausen noise analysis is not recommended where variations in elemental composition, phase composition, grain size, strain hardening, crystallographic texture grain shape, grain
orientation, carbide size and distribution, and
other microstructural characteristics accompany
variations in residual stress. A recent evaluation
of BNA by Allison and Hendricks (Ref 111)
confirms the uncertainty of BNA residual-stress
measurements.

5.

6.

7.

8.

9.

Summary and Recommendations


Only the destructive stress-relief, semidestructive hole-drilling, or nondestructive XRD
methods of residual-stress measurement are generally reliable over a broad range of steel alloys
displaying residual-stress fields induced by the
various manufacturing processes. Measurement
of residual stresses can be very expensive and
time consuming, and it is often worthwhile to
consult experts in the field before deciding on a
measurement method. Before an engineer or scientist who is inexperienced in residual-stress
measurement selects a method and attempts to
measure stresses he/she should consult someone
experienced in residual-stress measurement and
analysis.

10.

11.

12.

13.

REFERENCES
1. R. Gunnert, Method for Measuring TriAxial Residual Stresses, Document No.
X-184057-OE, Commission X of the International Institute of Welding, 1957;
Weld. Res. Abroad, Vol 4 (No. 10), 1958,
p 1725
2. M.G. Moore and W.P. Evans, Mathematical Corrections in Removal Layers in XRay Diffraction Residual Stress Analysis,
SAE Trans., Vol 66, 1958, p 340345
3. A. Constantinescu and P. Ballard, On the
Reconstruction Formulae of Subsurface
Residual Stresses after Matter Removal,
Fifth Int. Conf. Residual Stresses, Vol 2,
Institute of Technology, Linkopings University, Sweden, 1997, p 703708
4. Residual Stress Measurement by X-Ray
Diffraction, SAE J784a, Handbook Sup-

14.

15.

16.

plement, Society of Automotive Engineers, 1971


A.J. Bush and F.J. Kromer, Residual
Stresses in a Shaft after Weld Repair and
Subsequent Stress Relief, SESA Spring
Meeting Paper No. A-16, Society for Exp.
Stress Analysis, 1979
C.O. Ruud, P.S. DiMascio, and J.J. Yavelak, Comparison of Three Residual Stress
Measurement Methods on a Mild Steel
Bar, Exp. Mech., Vol 25 (No. 4), Sept
1985, p 338343
J.A. Pineault and M.E. Brauss, In-Situ
Measurement of Residual and Applied
Stresses in Pressure Vessels and Pipeline
Using X-Ray Diffraction Techniques, Determining Material Characterization, Residual Stress and Integrity with NDE,
PUP-Vol 276, NDE-Vol 12, American Society of Mechanical Engineers, 1994
R. Wimpory, G.M. Swallow, and P. Lukas,
Neutron Diffraction Residual Stress Measurements in Carbon Steels, Fifth Int.
Conf. Residual Stresses, Vol 2, Institute of
Technology, Linkopings University, Sweden, 1997, p 676681
M.E. Brauss and J.A. Pineault, Residual
Strain Measurement of Steel Structures,
NDE for the Energy Industry, NDE-Vol
13, D.E. Bray, Ed., American Society of
Mechanical Engineers, 1995
C.O. Ruud and P.S. DiMascio, A Prediction of Residual Stress in Heavy Plate Butt
Weldments, J. Mater. Eng. Sys., Vol 31
(No. 1), Jan 1981, p 6265
C.O. Ruud and M.E. Jacobs, Residual
Stresses Induced by Slitting Copper Alloy
Strip, NDC of Materials VI, Plenum Press,
1994, p 413424
E.F. Rybicki and R.B. Stonesifer, Computation of Residual Stresses Due to Multipass Welds in Piping Systems, ASME
78-PVP-104, American Society of Mechanical Engineers, 1978
M. Ehlers, H. Muller, and D. Loke, Simulation of Stresses and Residual Stresses
Due to Immersion Cooling of Tempering
Steel, Fifth Int. Conf. Residual Stresses,
Vol 2, Institute of Technology, Linkopings
University, Sweden, 1997, p 400405
K. Masubuchi, Analysis of Welded Structures: Residual Stresses, Distortion, and
Their Consequences, 1st ed., Pergamon
Press, 1980
M. Bijak-Zochowski, P. Marek, and M.
Tracz, On Subsurface Distributions of Residual Stress Created by Elasto-Plastic
Rolling Contact, Fifth Int. Conf. Residual
Stresses, Vol 2, Institute of Technology,
Linkopings University, Sweden, 1997, p
430445
D. Green and S. Bate, Calculation of Residual Stresses Using Simplified Weld
Bead Modelling Technique, Fifth Int.
Conf. Residual Stresses, Vol 2, Institute of
Technology, Linkopings University, Sweden, 1997, p 508513

17. H. Michaud, F. Mechi, and T. Foulquies,


Three Dimensional Representation of the
Residual Stress Distribution in Steel Wires
or Bars, The Fifth Int. Conf. Residual
Stresses, Vol 2, Institute of Technology,
Linkopings University, Sweden, 1997, p
534538
18. K.W. Mahin, W.S. Winters, T. Holden,
and J. Root, Measurement and Prediction
of Residual Elastic Strain Distributions in
Stationary and Traveling Gas Tungsten
Arc Welds, Practical Applications of Residual Stress Technology, ASM International, 1991, p 103109
19. J. Pineault and M. Brauss, Automated
Stress MappingA New Tool for the
Characterization of Residual Stress and
Stress Gradients, Proc. Fourth Int. Conf.
Residual Stresses, Society for Experimental Mechanics, 1994, p 4044
20. M. Brauss, J. Pineault, S. Teodoropol, M.
Belassel, R. Mayrbaurl, and C. Sheridan,
Deadload Stress Measurement on Brooklyn Bridge Wrought Iron Eye Bars and
Truss Sections Using X-Ray Diffraction
Techniques, ICB-9751, Proc. 14th Annual International Bridge Conf. and Exhib., Engineering Society of Western PA,
1997, p 457464
21. M.G. Carfaguo, F.S. Noorai, M.E. Brauss,
and J.A. Pineault, X-Ray Diffraction
Measurement of Stresses in Post-Tensioning Tendons, IABSE Symposium (San
Francisco, CA) 1995, Extending the Life
Span of Structures, Vol 71/1, International
Association for Bridge and Struct. Eng.,
ETH Honggerberg, CH-8093, Zurich,
Switzerland, 1995, p 201206
22. C.O. Ruud, P.S. DiMascio, and D.J.
Snoha, A Miniature Instrument for Residual Stress Measurement, Adv. X-Ray Anal.,
Vol 27, Plenum Press, 1984, p 273283
23. C.O. Ruud, R.N. Pangborn, P.S. DiMascio,
and P.J. Snoha, Residual Stress Measurement on Thick Plate Low-Alloy Steel Narrow Gap Weldments by X-Ray Diffraction, ASME 84-PVP-128, American
Society of Mechanical Engineers, 1984
24. C.O. Ruud, J.A. Josef, and D.J. Snoha, Residual Stress Characterization of ThickPlate Weldments Using X-Ray Diffraction,
Weld. Res. J., Supplement, March, 1993, p
875915
25. P. Johanssen, Determination of Residual
Stresses in Thick Welded Plated Utilizing
X-Rays and Layer Removal Techniques,
Part 1: Theory for Reconstruction of Original Stress State, Internal Report, Dept. of
Materials and Solid Mechanics, The Royal
Institute of Technology, Stockholm, Sweden, April 1978
26. E. Heyn, Internal Strains in Cold Wrought
Metals, and Some Troubled Caused
Thereby, J. Inst. Met., Vol 12, 1914, p 1
37
27. E. Stablein, Stress Measurements on Billets Quenched from One Side, Stahl Eisen,
Vol 52, 1932, p 1517

116 / Measurement and Prediction of Residual Stress and Distortion


28. M. Mesnager, Methods de Determination
dees Tensions Existant dans un Cylindre
Circulative, C.R. Hebed. Seances Acad.
Sci., 1926, p 169
29. G. Sacks, Evidence of Residual Stresses in
Rods and Tubes, Feitsch. Metallkde., Vol
19, 1927, p 352357
30. R.G. Treuting and W.T. Read, Jr., J. Appl.
Phys., Vol 22 (No. 2), Feb 1951, p 130
134
31. R. Gunnert, Residual Welding Stresses,
Method of Measuring Residual Stresses
and Its Application to a Study of Residual
Welding Stresses, Alquist E. Wicksell,
Stockholm, 1955
32. R. Gunnert, Method for Measuring Residual Stresses in the Interior of a Material,
Document No. X-16257, Commission X
of the International Institute of Welding,
1957; Weld. Res. Abroad, Vol 6, 1960, p
1024
33. D. Rosenthal and J.T. Norton, A Method
for Measuring Triaxial Residual Stresses
in Plates, Weld. J., Vol 24 (No. 5), Research Supplement, 1945, p 295s307s
34. R. J.-S. Chen, Determination of Residual
Stresses in a Thick Weldment, M.S. Thesis, Department of Engineering Science
and Mechanics, The Pennsylvania State
University, Nov 1983
35. G. Pickel, Application of the Fourier
Method to the Solution of Certain Boundary Problems in the Theory of Elasticity,
J. Appl. Mech., 1944, p 176182
36. D.L. Sikarskie, On a Series Form of Correction to Stresses Measured Using X-Ray
Diffraction, AIME Trans., Vol 39, 1967, p
577580
37. A.R. McIlree, C.O. Ruud, and M.E. Jacobs, The Residual Stresses in Stress Corrosion Performance of Roller Expanded Inconel 600 Steam Generator Tubing, Proc.
Int. Conf. Expanded and Rolled Joint
Technology, Canadian Nucl. Society,
1993, p 139148
38. Y.W. Park, P.H. Cohen, and C.O. Ruud,
The Development of a Model for Plastic
Deformation in Machined Surface, Mater.
Manuf. Proc., Vol 8 (No. 5), 1994
39. Y.W. Park, P.H. Cohen, and C.O. Ruud,
Sensitivity of Shear Process on Metal Cutting to the Development of Residual
Stress, Rev. Prog. NDE, Vol 14, 1995, p
11831188
40. Y.C. Shin, S.J. Oh, and C.O. Ruud, Investigation of Residual Stresses of Machined
Surfaces by an X-Ray Diffraction Technique, NDC of Materials IV, Plenum Press,
1992, p 408418
41. J.E. Hoffman, D. Lohe, and E. Macherauch, Influence of Machining Residual
Stresses on the Bending Fatigue Behaviors
of Notched Specimens of Ck45 in Different Heat Treating States, Residual Stresses
in Science and Technology, Vol 2, DGM
Informationsgesellschaft Verlag, Germany, 1987, p 801814

42. E. Brinksmier, Residual Stressed in Hard


Metal Cutting, Residual Stresses in Science and Technology, Vol 2, DGM Informationsgesellschaft Verlag, Germany,
1987, p 839846
43. E. Schreiber and H. Schlicht, Residual
Stresses after Turning of Hardened Components, Residual Stresses in Science and
Technology, Vol 2, 853860, DGM Informationsgesellschaft Verlag, Germany,
1987, p 853860
44. F. Ghanem, H. Dishom, C. Braham, R. Fatallak, and J. Leider, An Engineering Approach to the Residual Stresses Due to
Electric-Discharge Machining Process,
Proc. Fifth Int. Conf. on Residual Stresses,
Vol 1, Ericson, Oden, and Anderson, Ed.,
ICRS-5, Institute of Technology, Linkopings University, Sweden, 1977
45. Electropolishing, Surface Cleaning, Finishing and Coating, Vol 5, 9th ed., Metals
Handbook, American Society for Metals,
1982, p 303309
46. Hetenyi, Ed., Handbook of Experimental
Stress Analysis, John Wiley & Sons, 1950
47. N.A. Crites, Your Guide to Todays Strain
Gages, Prod. Eng., Vol 33 (No. 4), 17 Feb
1962, p 6981; Equipment and ApplicationTodays Strain Gages, Prod. Eng.,
Vol 33 (No. 6), 19 March 1962, p 8593
48. Y.Y. Hung, Shearography: A New Optical
Method for Strain Measurement and Nondestructive Testing, Opt. Eng., Vol 21 (No.
3), 1982, p 391394
49. K. Li, Interferometric Strain/Slope Rosette
Technique for Measuring Displacements/
Slopes/Strains/ Residual Stresses, Proc.
of the 1997 NSF Design and Manufacturing Grantees Conf., University of Washington (Seattle, WA), National Science
Foundation, Arlington, VA, 1997, p 571
572
50. C.S. Vikram, M.J. Pedensky, C. Feng, and
D. Englehaupt, Residual Stress Analysis
by Local Laser Heating and Speckle-Correlation Interferometry, Exp. Tech., Nov/
Dec 1996, p 2730
51. B.D. Cullity, Elements of X-Ray Diffraction, 2nd ed., Addison-Wesley, 1978, p
469472
52. C.O. Ruud, Residual and Applied Stress
Analysis of an Alloy 600 Row 1 U-Bend,
NP-5282, R.P 53033, Electric Power Research Institute, 1987, p 211
53. A. Bouzina, C. Braham, and J. Ledion,
Evaluation and Prediction of Real Stress
State Stress Corrosion Cracking Specimens, Fifth Int. Conf. Residual Stresses,
Vol 2, Institute of Technology, Linkopings
University, Sweden, 1997, p 10601065
54. K. Masubuchi and D.C. Martin, Investigation of Residual Stresses by Use of Hydrogen Cracking, Weld. J., Vol 40 (No.
12), Research Supplement, 1961, p 553s
556s
55. C.E. McKinsey, Effect of Low-Temperature Stress Relieving on Stress Corrosion

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

Cracking, Weld. J., Vol 33 (No. 4), 1954,


Research Supplement, p 161s166s
W. Radeker, A New Method for Proving
the Existence of Internal Stress Caused by
Welding, Schneissen Schneider, Vol 10
(No. 9), 1958, p 351358 (see also abstract
by Claussen, Weld. J., Vol 38 (No. 4)
1959, p 300
J. Mathar, Determination of Initial Stresses
by Measuring Deformation around Drilled
Holes, Arch. Eisenhuttenwes., Vol 6, 1934,
p 277281; Trans. ASME, Vol 56 (No. 4),
1934, p 249254
Determining Residual Stresses by the
Hole Drilling Strain-Gage Method, E
837, Annual Book of ASTM Standards,
ASTM, 1983
H. Wolf and D.C. Sauer, New Experimental Technique to Determine Residual
Stresses in Large Turbine-Generator Components, presented at the Amer. Power
Conf. (Chicago, IL), 1 May 1974
H. Wolf and W. Bohn, Origin, Measurement and Assessment of Residual Stresses
in Large Forging for Turbines and Generators, Arch. Eisenhuttenwes., Vol 42 (No.
7), 1971, p 509511 (in German)
H. Wolf, E. Stucker, and H. Nowack, Investigations of Residual Stresses in the
Turbine and Generator Industry, Arch. Eisenhuttenwes., Vol 48 (No. 3), March
1977, p 173178
J. Lu and J.F. Flavenot, Applications of the
Incremental Hole-Drilling Method for
Measurement of Residual Stress Distributions, Experimental Techniques, Presented
at the SEM Spring Conference (Portland,
OR), Society of Exp. Mechanics, 510
June 1988
M.T. Flaman and B.H. Manning, Determination of Residual Stress Variation with
Depth by the Hole-Drilling Method, Exp.
Mech., Vol 25, 1985, p 205207
R.A. Kelsey, Measuring Non-Uniform Residual Stresses by the Hole Drilling
Method, Proc. Soc. Exp. Stress Anal., Vol
14 (No. 1), 1956, p 181184
A.J. Bush and F.J. Kromer, Simplification
of the Hole-Drilling Method of Residual
Stress Measurement, ISA Trans., Vol 12
(No. 3), 1973, p 249259
N.J. Rendler and I. Vigness, Hole-Drilling
Strain-Gage Method of Measuring Residual Stress, Exp. Mech., Vol 6 (No. 12) Dec
1966, p 577586
J.W. Dini, G.A. Beneditti, and H.R. Johnson, Residual Stresses in Thick ElectroDeposits of a Nickel-Cobalt Alloy, Exp.
Mech., Feb 1976, p 5660
F. Witt, F. Lee, and W. Rider, A Comparison of Residual Stress Measurements Using Blind-Hole, Abrasive-Jet and Trepanning Methods, Exp. Tech., Vol 7, Feb
1983, p 4145
P.S. Prevey, Residual Stress Distribution
Produced by Strain Gage Surface Prepa-

Measurement of Residual Stresses / 117

70.

71.

72.

73.

74.
75.
76.
77.

78.

79.

80.

81.
82.

83.

84.

rations, 1986 SEM Conference on Experimental Mechanics, 1986


S. Kokubo, Changes in Hardness of a Plate
Caused by Bending, Sci. Rep. Tohoku Imperial Univ., Jpn., Series 1, Vol 21, 1932,
p 256267
G. Sines and R. Carlson, Hardness Measurements for the Determination of Residual Stresses, ASTM Bull., No. 180, Feb
1952, p 3537
J. Pomey, F. Goratel, and L. Abel, Determination des Cartraintes Residuelles dans
les Pieces Einentees, Publication Scientifiques et Techniques du Ministere de lair,
1950, p 263
S.S. Chiang, D.B. Marshall, and A.G.
Evans, The Response of Solids to Elastic/
Plastic Indentations I: Stresses and Residual Stresses, J. Appl. Phys., Vol 53 (No.
1), 1982, p 291311
C.O. Ruud, X-Ray Analysis and Advances
in Portable Field Instrumentation, J. Met.,
Vol 31 (No. 6), June 1979, p 1015
H.P. Klug and L.E. Alexander, X-Ray Diffraction Procedures, 2nd ed., John Wiley
& Sons, 1974, p 768770
C.S. Barrett and T.B. Massalski, Structure
of Metals, 3rd ed., McGraw-Hill, 1966, p
474476
C.O. Ruud and G.D. Farmer, Residual
Stress Measurement by X-Rays: Errors,
Limitations, and Applications, Nondestructive Evaluation of Materials, J.J.
Burke and V. Weiss, Ed., Plenum Press,
1979, p 101116
R.N. Pangborn, S. Weissman, and I.R.
Kramer, Dislocation Distribution and Prediction of Fatigue Damage, Metall. Trans.,
Vol 12A (No. 1), 1981, p 109120
D.A. Bolstad, R.A. Davis, W.E. Quist, and
E.C. Roberts, Measuring Stress in Steel
Parts by X-Ray Diffraction, Met. Prog.,
1963, p 8892
D.S. Kurtz, P.R. Moran, K.J. Kozaczek,
and M. Brauss, Apparatus for Rapid Psi
Squared Stress Measurement and Its Applications to Titanium Alloy Jet Engine
Fan Blades, Fifth Int. Conf. Residual
Stresses, Vol 2, Institute of Technology,
Linkopings University, Sweden, 1997, p
744749
B.B. He and K.L. Smith, Strain and Stress
Measurements with Two-Dimensional Detector, Adv. X-Ray Anal., Vol 41, 1977
M.R. James and J.B. Cohen, The Application of a Position-Sensitive X-Ray Detector to the Measurements of Residual
Stresses, Adv. X-Ray Anal., Vol 19, Gould,
Barnett, Newkirk, and Ruud, Ed., 1976, p
695708
H. Zantopulou and C.F. Jatczak, Systematic Errors in Residual Stress Measurement
Due to Specimen Geometry, Adv. X-Ray
Anal., Vol 14, Plenum Press, 1970
C.O. Ruud, D.J. Snoha, and D.P. Ivkovich,
Experimental Methods for Determination
of Precision and Estimation of Accuracy in

85.

86.

87.

88.

89.
90.
91.
92.

93.

94.

95.

96.

97.

98.

99.

XRD Residual Stress Measurement, Adv.


X-Ray Anal., Vol 30, 1987, p 511522
J.H. Root, C.E. Coleman, J.W. Bowden,
and M. Hayashi, Residual Stresses in Steel
and Zirconium Weldments, J. Press. Vessel Technol., Vol 119, 1997, p 137141
M. Hayaski, S. Ohkido, N. Minakawa, and
Y. Murii, Residual Stress Distribution
Measurement in Plastically Bent Carbon
Steel by Neutron Diffraction, Fifth Int.
Conf. Residual Stresses, Vol 2, Institute of
Technology, Linkopings University, Sweden, 1997, p 762769
J.H. Root, R.R. Hosbaus, and T.M.
Holden, Neutron Diffraction Measurements of Residual Stresses Near a Pin Hole
in a Solid-Fuel Booster Rocket Casing,
Practical Applications of Residual Stress
Technology, C.O. Ruud, Ed., ASM International, 1991, p 8393
G.A. Alers, Ultrasonic MethodsOverview, Proc. of a Workshop on Nondestructive Evaluation of Residual Stress,
NTIAC-762, 1975, p 155161
D.I. Crecraft, Ultrasonic Measurement of
Stresses, Ultrasonics, 1968, p 117121
G. Bradfield and H. Pursey, Philos. Mag.,
Vol 44, 1953, p 295
H. Pursey and H.L. Cox, Philos. Mag., Vol
45, 1954, p 295
W.J. McGonagle and S.S. Yun, Measurement of Surface-Residual Stress by Nondestructive Methods, Proc. Fifth Int. Conf.
Nondestructive Testing (Montreal, Canada), May 2125 1967, p 159164
A.J. Boland et al., Development of Ultrasonic Tonography for Residual Stress
Mapping, EPRI RP 5042, Final Report,
Electric Power Research Institute, 1980
M.R. James and O. Buck, Quantitative
Nondestructive Measurement of Residual
Stresses, Crit. Rev. Solid State Mater. Sci.,
Aug 1980, p 61105
E.P. Papadakis, Ultrasonic Attenuation
and Velocity in Three Transformation
Products of Steel, J. Appl. Phys., Vol 35
(No. 5), 1964, p 14741482
A. Moro, C. Farina, and F. Rossi, Measurement of Ultrasonic Wave Velocity of
Steel for Various Structures and Degrees
of Cold-Working, NDT Int., Aug 1980, p
169175
B.R. Tittman and R.B. Thompson, Measurement of Physical Property Gradients
with Elastic Surface Wave Dispersion,
Proc. Ninth Symp. on NDE, Southwest Research Institute, 1980, p 2028
K. Salama and C.K. Ling, The Effect of
Stress on The Temperature Dependence of
Ultrasound Velocity, J. Appl. Phys., Vol
51 (No. 3), March 1980, p 15051509
K. Salama and G.A. Alers, Third-Order
Elastic Constants of Copper at Low Temperature, Phys. Rev., Vol 161 (No. 3), Sept
1967, p 673680

100. R.E. Schramm, J. Szelazek, and A. Van


Clark, Jr., Ultrasonic Measurement of Residual Stress in the Rims of Inductively
Heated Railroad Wheels, Mater. Eval.,
Aug 1996, p 929933
101. H. Fukuaka, H. Tada, K. Hirakawa, H.
Sakawoto, and Y. Toyo, Acoustoelastic
Measurements of Residual Stresses in the
Rim of Railroad Wheels, Wave Propagation in Inhomogenous Media and Ultrasonic Nondestructive Evaluation, Vol 6,
G.C. Johnson, Ed., ASME, 1984, p 185
193
102. E.R. Schneider, R. Harzer, D. Bruche, and
H. Frotscher, Reliability Assurance of
Railroad Wheels by Ultrasonic Stress
Analysis, Proc. Third European Conf. on
Residual Stress Analysis (Frankfurt, Germany), 46 Nov 1992
103. D.M. Egle and D.E. Bray, Ultrasonic Measurement of Longitudinal Rail Stresses,
Mater. Eval., Vol 378 (No. 4), March
1979, p 4146
104. D.E. Bray and T. Leon-Salamanca, ZeroForce Travel-Time Parameters of Ultrasonic Head-Waves in Railroad Rail, Mater. Eval., Vol 43 (No. 7), June 1985, p
854858
105. C.G. Gardner, Barkhausen Noise Analysis,
Proc. Workshop on Nondestructive Evaluation of Residual Stress, NTIAC-762,
Nondestructive Testing Information Analysis Center, 1975, p 211217
106. R. Chait, Residual Stress Pattern in a High
Hardness Laminar Composite Steel Weldment, Proc. Workshop on Nondestructive
Evaluation of Residual Stress, NTIAC-76
2, Nondestructive Testing Information
Analysis Center, 1975, p 237245
107. O. Sundstrom and K. Torronen, The Use
of Barkhausen Noise Analysis in Nondestructive Testing, Mater. Eval., Feb 1979,
p 5156
108. J.R. Barton, F.N. Kusenberger, R.E. Beissner, and G.A. Matskanin, Advanced
Quantitative Magnetic Nondestructive
Evaluation Methods, Theory and Experiment, Nondestructive Evaluation of Materials, J.J. Burke and V. Weiss, Ed., Plenum
Press, 1979, p 461463
109. K. Ono, M. Shibata, and M.M. Kwan, Determination of Residual Stress by Magnetomechanical Acoustic Emission, UCLA
Department of Materials, ONR Tech. Rep.
No. 8001, Conference on Residual Stress
for Designers and Engineers, ASM International, April 1980
110. R.R. King and V.D. Smith, Residual
Stress Measurements in Structural Steels,
Final Report SWRI Project 154600, Contract DOT-FH-119133, Federal Highway
Administration, April 1981
111. H.D. Allison and R.W. Hendricks, Correlation of Barkhausen Noise Signal and XRay Residual Stress Determinations in
Grinding-Burned 52100 Steel, Fifth Int.
Conf. Residual Stresses, Vol 2, Institute of
Technology, Linkopings University, Sweden, 1997, p 640645

Handbook of Residual Stress and Deformation of Steel


G. Totten, M. Howes, T. Inoue, editors, p118-124
DOI: 10.1361/hrsd2002p118

Copyright 2002 ASM International


All rights reserved.
www.asminternational.org

Stress Determination in Coatings*


COATINGS AND THIN FILMS can be produced by a large variety of deposition techniques. Typical processes are physical vapor
deposition (PVD), chemical vapor deposition
(CVD), electroplating, electroless deposition,
anodizing, thermal growth, and thermal spraying. Since the early 1980s, considerable progress
has been made in improving deposition processes for a wide range of high-technology applications. Consequently, many new ceramic
coatings and films have been introduced in various industries. Typical examples are metal-oxide
semiconductors for microelectronics; titanium
nitride, titanium carbide, aluminum oxide, and
silicon nitride for machining tools; and thermal
sprayed tungsten carbide/cobalt, M-chromiumaluminum-yttrium (where M stands for iron, cobalt, or nickel), and yttrium oxide/partially stabilized zirconia coatings for aerospace
applications.
Residual stresses, which are internal and
therefore locked in, are contained in materials
that are produced by nearly every mechanical,
chemical, and thermal process, either alone or in
combination. As a result, most coatings are in a
state of internal stress, including metallics and
ceramics. The stress can be either compressive
or tensile. It is generally recognized that compressive stresses in coatings are more favorable
than tensile stresses because they increase resistance to fatigue failure. However, extremely high
compressive stresses may cause either coating
separation from the base metal or intracoating
spallation. Generally, if a tensile stress causes
strain that exceeds the elastic limit of the coating,
then it will cause cracking in the coating perpendicular to the direction of the stress. Therefore,
understanding the formation of residual stress in
the coating is important to prevent the coating
from peeling or cracking during service. Furthermore, residual stresses have significant influences on the mechanical and physical properties
of the coatings, particularly electrical resistivity,
optical reflectance, fatigue, and corrosion.
There are three types of residual stresses:
Macrostresses, which are nearly homoge-

neous over macroscopic areas of the material


Microstresses, which are nearly homogeneous over microscopic areas, such as one
grain or subgrain

Inhomogeneous microstresses, which are in-

homogeneous even on a microscopic level


Residual macrostresses are the ones of most interest in engineering practice because they can
substantially affect component service performance. Both residual and inhomogeneous microstresses are of more interest in materials science.
This article intends to provide a useful guide
for measuring residual macrostress on a coating.
The most commonly used measurement methods
are mechanical deflection, x-ray diffraction, and
hole-drilling strain gage. After a discussion on
the origins of residual stress, the fundamental
principles, as well as examples of practical measurements, are described for each method.

Origins of Residual Stress


Residual macrostress in a coating combines
the intrinsic stress and the thermal stress acting
in the coating plane parallel to the coating/substrate interface:
rt ri rth

(Eq 1)

where rt is the total macrostress, and ri and rth


are intrinsic stress and thermal stress, respectively. Intrinsic stress results from the growth
processes, depending primarily on deposition
parameters, whereas thermal stress arises from a
mismatch in coefficients of thermal expansion
between the coating and the substrate.
Many phenomenological models have been
proposed to explain the occurrence of intrinsic
stresses by correlating them with a variety of
coating microstructure and process features. To
varying degrees, the intrinsic stress of a coating
is associated with these deposition conditions
and coating features:
Incorporation of residual gas atoms in the

coating

Grain size, microvoid, and dislocation density

in the coating

Combined effect of surface tension and

growth process at grain boundaries

Deposition temperature relative to the melt-

ing temperature of the coating material

Annealing and shrinkage of disordered ma-

terial buried behind the advancing surface of


a growing coating
Although many studies have described the intrinsic stresses, information on the corresponding structural details is limited. It seems unlikely
that one can formulate a generalized model of
intrinsic stress for various coating materials and
deposition processes.
Any coating that is prepared at elevated temperatures (T2) and then cooled to room temperature (stress measurement temperature, T1) will be
thermally stressed because of the difference in
the coefficients of thermal expansion between
the coating and the substrate. Assuming no deformation of the substrate, the magnitude of the
thermal stress in the coating is:
rth (c s)(T2 T1)Ec /(1 mc)

(Eq 2)

where c and s are the thermal expansion coefficients for the coating and the substrate, respectively, and Ec and mc are the Youngs modulus and Poissons ratio of the coating,
respectively. A coating deposited at an elevated
temperature exhibits compressive stress if s 
c, but tensile stress if s c. In the case of s
c, the substrate shrinks more than the coating
does during cooling from the deposition temperature and compresses the coating to maintain
dimensional compatibility.
In some cases, thermal stress is the primary
residual stress of the coating. For example, a titanium nitride coating can be deposited on a cemented carbide substrate (tungsten carbide-1
wt% Ta-10 wt% Co) via a CVD process at 1000
C (1830 F). With the values of s 5
106 /K, c 9.54 106 /K, ETiN 411 GPa
(60 106 psi), and mTiN 0.24, the stress in
the coating, as calculated from Eq 2, is 2.39 GPa
(0.35 106 psi) in tension at 25 C (77 F).

Energetic particle bombardment during coat-

Deflection Method

ing growth
Lattice misfit between the substrate and the
growing coating

The deflection method is the most widely used


technique for determining the residual stress in

*This article is adapted from Stress Determination for Coatings by J. Albert Sue and Gary S. Schajer in Surface Engineering, Volume 5, ASM Handbook, ASM International, 1994, p 647
653.

Stress Determination in Coatings / 119


a coating. In terms of basic principles, it involves
measuring the amount of bending in a strip that
is due to the deposition of the coating. A formula
for calculating the residual stress in an electrodeposited coating was derived first by Stoney
(Ref 1) in 1909. Subsequent researchers have derived more complex formulas to improve the accuracy of the stress evaluation. Comparative
studies on those stress-evaluation formulas are
reviewed and discussed elsewhere (Ref 24).
Nearly all formulas are variants of Stoneys formula.
Consider a coating deposited on one side of a
strip substrate. Both coating and substrate are assumed to be homogeneous. A mismatched force
at the coating/substrate interface results in residual stress in the coating, which bends the strip
either upward (concave) or downward (convex),
depending on whether the stress is tensile or
compressive. For overall force and moment
equilibrium of the coating/substrate composite,
it can be shown that the residual stress in the
coating is (Ref 5):
rc (Es d 2s )/[6(1 ms)Rdc](ds  dc)

rc {Es d s2/[6(1 ms)Rdc]}


{(1 CH3)/(1 H)}

(Eq 4)

where
C [Ec(1 ms)]/[Es(1 mc)]
H ds /dc
R  L2 /8f  L2 /2d

(Eq 5)

where E is Youngs modulus; m is Poissons ratio;


d is thickness, with the subscript c denoting coating
and s denoting substrate; R is the radius of curvature of the bent strip, L is the length of the strip, f
is the deflection from the free end of the strip, and
d is the deflection at the center of the strip
(Fig. 1).

(Eq 3)

or, in general:

Although Eq 3 is used most often in practice


and represents a generalized Stoneys formula
for a planar state of stress, it tends to overestimate the value of measured stress. However, it
does not require knowledge of the elastic properties of the coating. Equation 4 provides a much
better approximation than Eq 3, but it does require knowledge of elastic properties. Equation
4 differs from Eq 3 in terms of a correction factor
[(1 CH3)/(1 H)]. Table 1 compares
stresses as a function of ds /dc, calculated from
both Eq 3 and 4. In the calculation, it was assumed that Ec 400 GPa (60 106 psi), Es
200 GPa (30 106 psi), and mc ms. It is clear
that both equations are in good agreement when
ds /dc  50. By choosing a large value of ds /dc,
the error can be minimized if Eq 3 is used.
Cantilever Beams. To measure the deflection
of cantilever beams, various techniques have
been developed (Ref 7), including optical, capacitance, mechanical, electromechanical, interferometric, and electromechanical or magnetic
restoration. Figure 1(a) shows the setup for the
deflection measurement when an optical system
is used. In this example, a titanium nitride coating was deposited on a quartz beam via a PVD
process, and the average residual stress in the
coating was determined by measuring the
amount of deflection at the free end, f, of the bent
beam, according to the relation:
rc [4Es d s2 f ]/[3(1 rs)L2 dc]

(Eq 6)

With the values of Es 71.7 GPa (10.4 106


psi), ms 0.16, ds 3 mm (0.12 in.), dc 3
106 m (1.2 104 in.), L 50 mm (2
in.), and f 2 106 m (8.0 105 in.), the
stress in the coating is determined to be 273 MPa
(40 ksi) in compression.
Disks. The average stress in the coating on a
disk substrate can be determined from the
amount of deflection, d, at the center of the disk
caused by the deposition of a coating on one
side. This value can be measured optically by
interferometry or microstylus profilometry. Figure 1(b) shows a typical interferometry apparatus
setup. Either technique measures deflection at
the same position, across a diameter of the disk,
both before and after coating deposition. The
stress in the coating is then calculated:
rc {Es d s2[(Ddx Ddy)/2]}/[3(1 ms)r2dc]

Fig. 1

Stress measurement techniques. (a) Bending of


cantilever beam. (b) Disk deflection. Source:

(Eq 7)

Ref 6

Table 1 Comparison of average residual


stress calculated using Eq 3 and 4
Calculated stress
Eq 3

Eq 4

ds /dc

MPa

ksi

MPa

ksi

200
100
50
20
10

1143
571
286
114
57

165.7
82.8
41.5
16.5
8.3

1137
565
280
109
52

164.9
81.9
40.6
15.8
7.5

Typical patterns of interference fringes. (a) rx


ry. (b) rx ry, where both components are in
compression or tension. (c) rx ry, where one component
is in compression and the other is in tension. Source:
Ref 8

Fig. 2

where Ddx and Ddy are deflection changes measured before and after coating deposition, and r
is the radius of the disk. Whether the stress is
tensile or compressive, it is determined by the
curvature of the disk before and after coating
deposition, as determined by a depth microscope. When the change in curvature is upward
(concave), viewed from the coating side, the
stress in the coating is tensile. The opposite
change in curvature indicates a compressive
stress in the coating.
Typical interference fringe patterns are shown
in Fig. 2. Figures 2(a) and (b) correspond to

120 / Measurement and Prediction of Residual Stress and Distortion


equal and unequal principal stresses of the same
sign, whereas Fig. 2(c) corresponds to a biaxial
stress state with principal stresses of opposite
signs. Using Fig. 2(a) as an example, the measured deflection, d, is equal to the wavelength of
monochromatic light multiplied by the number
of light fringes (d 632.8 nm 17).
Figure 3 shows typical microstylus traces on
the titanium nitride coated surface of an AISI
304 stainless steel disk, both before and after
coating deposition. The curvature of the coated
surface was downward (convex), viewed from
the coated side. With the values of Es 193
GPa (28 106 psi), ms 0.28, ds 4.74 mm
(0.186 in.), r 11.94 mm (0.4700 in.), dc 20
lm (800 lin.), and the deflection Dd 6.3
106 m (2.5 104 in.), the stress in the titanium nitride coating is 4.44 GPa (0.644 106
psi) in compression. The measurement errors
from both the optical and the microstylus trace
setups are within one-half light band and 0.1
lm (4 lin.), respectively.
Practical Considerations. First, either strip
or disk-shape substrate specimens should be parallel within 0.02 mm (0.001 in.) over their length
or diameter. Specimen edges should be free from
visible flaws and chips. In addition, the surfaces
on which the deflection measurements (interferometer or profilometer) will be performed should
be ground and lapped to a finish better than 0.08
lm (3 lin.) Ra.
Second, the dimensions of a strip of length L,
width w, and substrate thickness ds should obey
L  10w  10 ds. The substrate thickness of a
strip or disk depends on the coating thickness to
be deposited. The thickness ratio of the substrate
to the coating, ds /dc, should be greater than 50
to ensure the accuracy of the stress calculation
from Eq 3.
Third, substrate specimens should be stress relieved before coating deposition. They should be
placed between two stainless steel surface plates
under at least an 8.8 kPa (1.3 psi) normal load,
at an annealing temperature, in a vacuum furnace

for at least 1 h. The annealing temperature is


dependent on the substrate material.
Fourth, for coatings with an inherently smooth
surface, such as those produced by PVD and
CVD, the deflection can be determined either by
interference fringe or microstylus profilometer
measurements on the coated surface. For a coating with a relatively rough surface, such as those
produced by thermal spraying and electroplating, the measurements can be made on the surface that is opposite the coated surface.
Significance and Use. The deflection measurement method is recommended for determining the average stress in the cross section of a
coating with a thickness ranging from several
hundred angstroms to several hundred micrometers. Typically, the elastic constants of a thin
coating are much different from those of a bulk
material. Equation 3 provides a means for stress
measurement on a thin coating without any
knowledge of its elastic constants. The measurement normally applies only to a test sample. The
disk deflection method is particularly useful for
the direct inspection of silicon wafers used in
solar cells or integrated circuits.

X-Ray Diffraction Method


Basic Principles. Stress measurement, using
the x-ray diffraction method, is based on the
change in the interplanar spacing (strain) close
to the surface of the specimen material. The details of the theory and interpretation of residual
stress measurements are well described in the article X-Ray Diffraction Residual Stress Techniques in Volume 10 of the ASM Handbook, as
well as in Ref 9 to 11. Consider an isotropic
material with a lattice parameter d hkl
for un0
stressed material in the sample plane normal.
The strain in a direction inclined by an angle y
to the surface normal of the coating and the stress
acting in the surface plane of the coating at an
angle u with the principal axis of the specimen
are related by:
hkl
hkl
hkl
ehkl
uw (d uw d 0 )/d 0

[(1 m hkl)/E hkl]


(rx cos2 u sxy sin2 u ry sin2 u rz)
sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)
(rx ry rz) [(1 m hkl)/E hkl]
(sxz cos u syz sin u) sin w)
2

Typical microstylus trace on titanium nitridecoated AISI 304 stainless steel disk before and
after coating deposition

Fig. 3

coating in various stress states can be determined


using Eq 8.
Biaxial Stress. At a free plane, the out-ofplane stress components rz, sxz, and syz are all
zero, at a free surface. Because the penetration
depth of x-rays is very small, the resulting measurements refer specifically to near-surface material. Plane stress conditions, therefore, often
apply to x-ray measurements, and Eq 8 is simplified to:
hkl
hkl hkl
hkl
hkl
ehkl
uw (d uw d 0 /d 0 [(1 m )/E ]

ru sin2 w (m hkl/E hkl)(rx ry) (Eq 9)

where ru rx cos2 u ry sin2 u is the macrostress in the coating parallel to its surface at
an angle u with the principal axis of the sample.
For a biaxial stress state, rx ry ru, and
at w 0:
hkl
hkl
hkl
hkl
(d hkl
u,w0 d 0 )/d 0 2m ru/E

(Eq 10)

From Eq 9 and 10:


hkl
hkl
hkl
hkl
2
d hkl
uw d u,w0 rud 0 [(1 m )/E ] sin w
(Eq 11)

In practice, high-angle diffraction peaks of an


(hkl) reflection are obtained from w to w at
a given angle u. Lorentz polarization, absorption, and background corrections are applied to
the diffraction peak profile. The peak positions
are determined by profile fitting or other methods
and are subsequently converted to interplanar
spacing d hkl
uw for stress analysis.
2
In the linear plot of d hkl
uw versus sin w, the
intercept is I d hkl
,
and
the
slope
is M
u,w0
hkl
hkl
rud hkl
]. The stress of the coating
0 [(1 m )/E
can then be determined by:
ru M/{d 0hkl[(1 m hkl)/E hkl]}

(Eq 12)

Triaxial Stress State without Shear Stress. For


a material in a three-dimensional (triaxial) stress
state without shear stress, but with the stress
component rz having a finite value within the xray penetration volume, Eq 8 becomes (Ref 9,
11, 12):
hkl
hkl
hkl
ehkl
uw (d uw d 0 )/d 0

[(1 m hkl)/E hkl](rx cos2 u ry sin2 u rz)


sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)

(Eq 8)

where h, k, and l are the indices of the Bragg


hkl
reflection; ehkl
uw and d uw are the strain and interplanar spacing of (hkl) in the direction of (u, w),
respectively; m hkl and Ehkl are Poissons ratio and
Youngs modulus in (hkl) in the coating; rx, ry,
and rz are normal stresses; sxy, syz, and sxz are
shear stresses; rx is the normal stress acting in
the x direction on a plane perpendicular to the x
axis; and syz is the shear stress on a plane normal
to the y axis (the first subscript) in the z direction
(the second subscript). Residual stresses of the

(rx ry rz)

(Eq 13)

Two data sets, u 0 and u 90, are needed


to obtain rx, ry, and rz. The slopes and intercepts
2
of a linear function of d hkl
uw versus sin w at u
0 and u 90 are given by:
hkl
hkl
Mu0 d hkl
0 [(1 m )/E ](rx rz)

Mu90 d 0hkl[(1 m hkl)/E hkl](ry rz)


hkl
hkl
hkl
hkl
I d hkl
0 {[(1 m )/E ]rz (m /E )

(rx ry rz)}

(Eq 14)

Stress Determination in Coatings / 121


Stresses rx, ry, and rz can be determined from
the sum of the slopes and the intercept in Eq 14.
Triaxial Stress State with Shear Stress. A
coating with a three-dimensional (triaxial) stress
state, including shear stresses, is fully described
by Eq 8. The shear stresses, sxz and syz, have a
2
sin2 w dependence. The d hkl
uw versus sin w distribution is no longer linear and has two branches
of an ellipse for w  0 and w  0. This effect
is termed w splitting, which is an indication of
the presence of shear stress.
To obtain these stress-tensor components,
three data sets (u 0, u 45, and u 90)
are obtained for both w and w. The average
strain a1 and the deviation a2 from the strains of
w splitting are determined to be (Ref 9, 11,
12):
hkl
a1 (euw
ehkl
uw)/2
hkl
hkl
[(d uw
d uw
)/2d 0hkl] 1

[(1 m hkl)/E hkl]


(rx cos2 u sxy sin2 u ry sin2 u rz)
sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)
(rx ry rz)

(Eq 15)

hkl
a2 (euw
ehkl
uw)/2
hkl
hkl
(d uw
d uw
)/2d 0hkl

[(1 m )/E ](sxz cosu syz sinu) sin w


(Eq 16)
hkl

hkl

The stress-tensor components can be calculated from the slopes of linear plots of a1 versus
sin2 w and a2 versus sin2 w. For a1 versus sin2
w, (rx rz) is obtained at u 0, (ry rz)
at u 90, and sxy at u 45, whereas rz is
evaluated from the intercept if d hkl
0 is known.
Similarly, sxz and syz are obtained when u 0
and u 90, respectively, from the slope of a2
versus sin2 w.

Stress Measurement. Modern diffractometers are fully automated and equipped with computer software for performing numerically intensive analyses. Diffractometers are capable of
measuring residual stresses efficiently and economically. Typically, the measurement can be
completed in several hours. The following examples illustrate some stress measurements using Cu K radiation on cathodic arc PVD titanium nitride coatings on a substrate with various
stress states.
Biaxial. An approximately 10 lm (400 lin.)
thick coating of highly (111) oriented titanium
nitride was deposited on AISI 304 stainless steel
at 500 C (930 F). The x-ray diffraction sin2 w
technique was applied to determine the residual
stress in the (333)/(511) reflection of the coating.
Figure 4 shows the linear distribution of d hkl
uw versus sin2 w, indicating typical biaxial stress in the
coating. As shown, the slope M 1.1933
103 nm (4.698 1011 in.) and the intercept
I d hkl
uw0 0.082123 nm (0.003233 lin.).
Youngs modulus and Poissons ratio for titanium nitride in (333)/(511) are 364 GPa (52.8
106 psi) and 0.245, respectively (Ref 13). The
calculated residual stress for titanium coating is
4248 MPa (615 ksi) in compression. For
highly (111) oriented titanium nitride film, (511)
contribution is negligible. This is addressed in
Ref 13. The (422) reflection has 2h  130,
which is not desirable.
Triaxial without Shear Stress. To exemplify a
triaxial stress distribution, without shear stresses,
triaxial stress analysis was applied to a titanium
nitride coating deposited on Inconel 718 sub2
strate at 550 C (1020 F). The d hkl
uw versus sin
w distributions were obtained from (333)/(511)
for u 0 and u 90 (Fig. 5). The slopes
Mu0 and Mu90 are 0.96782 and 1.01415
103 nm (3.8103 and 3.9927 1011
in.), respectively, and the intercept I is 0.082083
nm (0.003231 lin.). The strain-free interplanar
spacing for (333)/(511) is 0.08160 nm

(0.003213 lin.). Based on Eq 14, the stress tensor (rij) from this analysis, in units of MPa, is:

rx sxy sxz
rij syx ry syz
szx szy rz

2653 151
0
0
0
2819 150
0
0
0
815 140

The result shows that the planar stresses are


equal biaxial within experimental error and that
the stress perpendicular to the coating surface is
in tension.
Triaxial with Shear Stress. A triaxial stress
distribution, including shear stresses sxy, syz, and
sxz, was studied in a cathodic arc PVD titanium
nitride coating on AM-355 stainless steel. The
2
d hkl
uw versus sin w distributions were obtained
from (333)/(511) at u 0, u 45, and u
90 using Cu K radiation. Figure 6 shows the
2
typical ellipse distribution of d hkl
uw versus sin w
at u 0. Based on analysis discussed at the
beginning of this section, the stress tensor in the
titanium nitride coating, in units of MPa, is:
rij

2699 148
1 181
110 32
1 181
2776 148 34 34
110 32
34 34 833 148

Practical Considerations. First, a W-diffractometer is preferable for conducting stress measurements. The X- and W-diffractometers are defined on the basis of the sample axis for w tilt
perpendicular or parallel to the plane of the incident and detected x-ray beam, respectively. A
W-diffractometer gives symmetric irradiated areas in w and w tilted angles at a given (hkl)
reflection line and a greater range of sin2 w (from
0 to 0.95 for a W-diffractometer and 0 to 0.5 for
an X-diffractometer).
Second, the diffractometer should be mechanically aligned and calibrated using a stress-free
standard sample (NIST SRM 660 lanthanum
hexaboride powder or SRM 640 silicon powder)
to obtain the peak position of the (hkl) reflection
line within 0.01 at Bragg angles 2h in the
range of w  0 and w  0.
Third, an appropriate x-ray wavelength should
be selected to achieve the desired sampling volume (penetration depth) for a particular set of
(hkl) planes. The penetration depth, which is defined as the distance from the surface to the depth
with 63% or 1/e of the intensity of the reflection
line, is calculated for X- and W-diffractometers,
respectively:
sX (sin2 h sin2 w)/[2l(sinh cosw)]
sW (sinh cosw)/2l

Fig. 4
flection

hkl
Linear distribution of d uw
as a function of sin2 w
of titanium nitride coating from (333)/(511) re-

hkl
Linear distribution of d uw
as a function of sin2 w
of titanium nitride coating from (333)/(511) reflection at u 0 and u 90

Fig. 5

where l is the linear absorption coefficient,


which can be obtained from Ref 14 for various
materials and radiations.
Fourth, to achieve high accuracy in residual
stress, the measurement should only be carried
out on an (hkl) reflection line with Bragg angles
2h  130 and with sufficient intensity for peakposition determination.

122 / Measurement and Prediction of Residual Stress and Distortion


Fifth, an appropriately sized collimator should
be selected. It should have an irradiated area
large enough to ensure that a statistically relevant number of grains or subgrains in the coating
are included in the measurement. The angular
resolution is essential in stress measurement. If
the spot mode of x-ray beam cannot sample a
sufficient number of grains, one should use another method.
Sixth, appropriate methods (Ref 9, 10, 15
17), such as modified Lorentzian, Gaussian,
Cauchy, Pearson VII, parabola, center of gravity,
gravity line, and cross correlation should be applied to determine the peak positions of the reflection line (hkl).
Seventh, a sufficient number of measurements
in the w and w directions should be conducted to obtain an accurately linear distribution
2
or w-splitting of d hkl
uw versus sin w. The significance of this distribution should be verified by
repeating measurements at different u angles. If
a nonlinear relation other than w-splitting is ob2
tained in the d hkl
uw versus sin w distribution, then
the coating being measured is inhomogeneous
and, therefore, the x-ray diffraction sin2 w
method is no longer applicable.
Eighth, the x-ray elastic constants of the coathkl
hkl
ing for an (hkl) plane, S hkl
and
1 m /E
hkl
hkl
S hkl
/2

(1

l
)/E
,
can
be
calculated
from
2
single-crystal compliance, according to an appropriate model (Ref 11, 12), or measured experimentally in uniaxial tension or bending tests
with a series of loads (Ref 13, 18, 19). X-ray
elastic constants and applied loads (ra) obey
these equations:

practice. For biaxial stress analysis, the lattice


spacing measured at w 0, d hkl
uw0, can be substituted for d hkl
0 . In this case, the contribution to
the total error is less than 0.1%. For a triaxial

stress analysis, the difference between d hkl


uw0
and d hkl
0 is included in the calculation. Consequently, a small error in d hkl
0 can lead to a relatively large error in the measured stress. To obtain acceptable stress results, d hkl
0 must be within
0.01% of its true value. The stress-free interplanar spacing d hkl
0 in the strain-free direction w*
is given by (Ref 20):
hkl
sin2 w* [S hkl
1 /(S 2 /2)]

{1 [(ry rz)/(rx rz)]


hkl
[3 (S hkl
2 /2S 1 )rz]/(rx rz)}; u 0

sin2 w* [S 1hkl/(S 2hkl/2)]


{1 [(ry rz)/(rx rz)]}
when |rx ry 2rz|  2|rz|
d hkl
0

hkl
d uw0
/[1 S hkl
1 (rx ry rz)

(S hkl
2 /2)rz]

Significance and Use. The advantage of the


x-ray diffraction method is its capability for
measuring residual stress directly and nondestructively on a product component. Portable diffractometers are commercially available and can
be used for on-site measurement. The method
can apply to a coating with a thickness ranging
from 0.5 to 350 lm (2 105 to 1.4 102
in.). However, it is difficult to measure residual
stresses in extremely thin coatings and, in some
cases, highly textured coatings. Furthermore, xray diffraction is inapplicable to amorphous
coatings, and a large scatter in stress measurement is often encountered in coatings with large
grain size.

hkl
hkl
S hkl
2 /2 (1 m )/E
hkl
2
(1/d hkl
0 )[/ra(d uw/sin w)]

Hole-Drilling Method

S 1hkl m hkl/E hkl [1/(2d 0hkl)](d hkl


uw0/2ra)
(a) Typical three-element strain-gage rosette. (b)
In-plane strain components caused by release of
residual stress through introduction of a hole. Source:
Ref 21

Fig. 7

Ninth, a stress-free interplanar spacing, d hkl


0 ,
for a coating may not be readily available in

Fig. 6

Basic Principles. The hole-drilling method


for measuring residual stresses involves drilling
a shallow hole in the test specimen to a depth
approximately equal to the hole diameter. Typ-

hkl
Typical ellipse distribution of d uw
as a function of sin2 w of titanium nitride coating from (333)/(511) reflection at varying angles. (a) u 0. (b) u 45. (c) u 90

Stress Determination in Coatings / 123


ical hole diameters range from 0.8 to 5.0 mm
(0.0300.200 in.). The creation of the hole redistributes the stresses in the material surrounding the hole. A specially designed three-element
strain-gage rosette, such as that shown in Fig.
7(a), measures the associated partial strain relief.
The in-plane residual stresses that originally existed at the hole location can then be calculated
from the measured strain reliefs using the
method described in ASTM E 83792 (Ref 21).
The ASTM standard also gives details of practical drilling procedures.
The partial strain relief measured by one of
the three strain gages in the rosette in Fig. 7(b)
is related to the principal in-plane residual
stresses by:
er (rmax rmin)A
(rmax rmin)B cos
2

(Eq 17)

where rmax and rmin are maximum and minimum


principal residual stresses, and is the angle
from the gage axis to the maximum principal
stress direction. A and B are calibration constants, the values of which depend on the specimen material properties, the rosette geometry,
the hole diameter, and the hole depth. ASTM E
837-92 tabulates the calibration constants for the
standard rosette pattern shown in Fig. 7.
Equation 17 can be inverted to determine the
principal residual stresses from the measured
strain reliefs. The result is:
rmax, rmin [(e3 e1)/4A]
{[(e3 e1)2 (e3 e1 2e2)2]1 / 2 /4B}
b 12 arctan [(e3 e1 2e2)/(e3 e1)]
(Eq 18)

Table 2 Numerical values of dimensionless


calibration coefficients a and b

D0 /D

0.30
0.31
0.32
0.33
0.34
0.35
0.36
0.37
0.38
0.39
0.40
0.41
0.42
0.43
0.44
0.45
0.46
0.47
0.48
0.49
0.50

Through-thethickness hole(a)
a
b

0.089
0.095
0.101
0.108
0.114
0.121
0.128
0.135
0.143
0.150
0.158
0.166
0.174
0.183
0.191
0.200
0.209
0.218
0.228
0.237
0.247

0.278
0.295
0.312
0.329
0.347
0.364
0.382
0.400
0.418
0.436
0.454
0.472
0.490
0.508
0.526
0.544
0.562
0.579
0.596
0.613
0.629

Blind hole(b),
depth 0.4D
a

0.111
0.118
0.126
0.134
0.142
0.150
0.158
0.166
0.174
0.182
0.190
0.199
0.208
0.217
0.226
0.236
0.246
0.255
0.265
0.275
0.285

0.288
0.305
0.322
0.340
0.358
0.376
0.394
0.412
0.430
0.448
0.466
0.484
0.503
0.521
0.540
0.558
0.576
0.594
0.612
0.630
0.648

(a) In a thin sheet. (b) In a thick material. Source: ASTM E 83792

where b is the angle measured clockwise from


the location of gage 1 to the direction of rmax.
The above equations are valid for a homogeneous isotropic material that is wide, when compared with the hole diameter, and thick, when
compared with the hole depth. The equations
also apply to a through-the-thickness hole in a
material in the form of a thin sheet, provided that
the sheet thickness is uniform near the hole. The
associated calibration constants have slightly
different values than those of the thick-material
case. Table 2 (from ASTM E 837-92) lists the
dimensionless calibration coefficients for both
the thin-sheet and thick-material cases. A and B
can be determined from the dimensionless coefficients using:
A (1 m)a/2E

B b/2E

(Eq 19)

Residual stress measurements in coated materials create an additional complication because


the combination of coating and substrate is no
longer a homogeneous material. The calibration
coefficients provided in ASTM E 837-92 no
longer accurately apply. However, the above two
equations are still valid for coated materials, if
the coating has a uniform thickness. If the substrate is less than several times greater than the
hole depth, then it is also necessary for the substrate thickness to be uniform.
The calibration coefficients A and B for hole
drilling in a coated material differ from the standard values given in ASTM E 837-92. The actual
values depend on the elastic properties of the
coating and the substrate, the coating thickness,
and the hole diameter and depth. These coefficients can be determined by either experimental
calibrations (Ref 22) using known externally applied stresses or finite-element calculations (Ref
23).
Approximate values of A and B for thick coatings can be estimated from the values given in
ASTM E 837-92. For this purpose, a thick
coating is one that is at least 0.25 times the mean
radius of the strain-gage rosette. For the smallest
commercially available hole-drilling rosette, the
mean radius is about 1.25 mm (0.050 in.). Therefore, the minimum acceptable coating thickness
is about 0.3 mm (0.012 in.).
An approximate estimation of A and B for a
coated material is based on the observation that
the hole-drilling method is most sensitive to the
stresses closest to the specimen surface. Almost
all of the measured strain relief is due to the
stresses in the material within a depth of about
0.25 times the mean radius of the hole-drilling
rosette. Thus, a specimen coated to at least this
depth is likely to behave similarly to a homogeneous thick specimen consisting only of coating material. Thus, the A and B calibration coefficients for a thick coating are approximately
equal to the ASTM tabulated values for a homogeneous material with the elastic properties
of the coating.
The A and B coefficients for coatings that are
thinner than 0.25 times the mean radius of the

hole-drilling rosette will deviate significantly


from the ASTM tabulated values. The coefficients must be determined on an individual basis,
either by experimental or computational means.
The use of the hole-drilling method with such
thin coatings is not generally recommended
because the sensitivity of the resulting strain
measurements is rather low. As a result, small
absolute errors in the strain measurements can
cause large relative errors in the computed residual stresses.
Stress Measurement. The following example
illustrates a hole-drilling measurement on a detonation-gun type of tungsten carbide-cobalt
(WC-Co) coating deposited on an AISI 1018
steel substrate. The coating was approximately
0.75 mm (0.03 in.) thick and had a macroscopically homogeneous structure. The elastic properties of the coating were Ec 172 GPa (25
106 psi) and mc 0.3.
A 062-RE hole-drilling strain-gage rosette
(5.13 mm, or 0.202 in., strain-gage mean diameter) was attached to the coated specimen. A 2.44
mm (0.096 in.) diam hole was cut in the WC-Co
coating by abrasive-jet drilling using 27 lm
(1080 lin.) alumina particles. Drilling proceeded
in four approximately equal depth increments,
up to a final depth of 0.356 mm (0.014 in.). The
strain measurements listed in Table 3 were made
after each hole-depth increment. Using the A and
B calibration coefficients from ASTM E 837-92,
adjusted for the elastic properties of the coating
material, the principal residual stresses in the
coating were found to be 260 MPa (38 ksi)
and 286 MPa (41 ksi), respectively. As
might be anticipated, the residual stresses in the
coating are approximately isotropic.
Practical considerations for the use of the
hole-drilling method are:
A high-speed drilling technique using carbide

drills is recom