Vous êtes sur la page 1sur 477

2002 ASM International. All Rights Reserved. www.asminternational.

org
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
2002 ASM International. All Rights Reserved. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

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, 1939-

TA473 .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. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

Preface

Control of steel deformation is one of the most steels, influence of residual stresses on cyclic de- termination in Coatings provides a guide for
common concerns within the metals processing formation behavior, influence of residual stresses measuring residual macrostress in coatings, Spe-
industry. Numerous surveys have been con- on crack initiation and propagation, and effect of cific topics include origin of residual stresses in
ducted by various organizations in recent years residual stresses on S-N curves; an overview of coatings and residual stress measurement meth-
to assess the critical needs of the industry. In modeling of the effect of residual stresses on fa- ods including the deflection method, x-ray dif-
nearly every survey that has been conducted, dis- tigue behavior is provided. fraction, and hole-drilling. A comparison of
tortion is either the greatest or second greatest The next article provides an overview of the these methods is provided.
concern among the steel heat treating commu- stability and relaxation behavior of macro and The last article in this section provides a de-
nity. Steel distortion control will exhibit tremen- micro residual stresses in steel due to thermal tailed review of methods used to measure and
dous effects on the profitability of the commer- and mechanical treatments. This discussion in- subsequent data analysis of inhomogeneous re-
cial enterprise. Therefore, it is not surprising that cludes relaxation of residual stresses by anneal- sidual stress fields. This discussion includes re-
the ability to understand the overall distortion ing, residual stress relaxation by uniaxial defor- sidual stress as an inverse problem of experi-
process and to be able to design solutions to this mation, and relaxation by cyclic deformation. mental mechanics, indicator crack method of
problem typically rank very high on these same Hydrogen embrittlement of metals, as well as measuring residual stress, arbitrary cut-out in-
surveys. other types of brittle fracture, result from nucle- dicator method, and experimental methods and
In view of the enormous visibility and impor- ation and development of micro-cracks caused equipment including photoelastic coating
tance of steel deformation problems, the editors by internal stresses. The last article in this sec- method, and optical interferometry. Although
decided to put together an engineering handbook tion provides an overview of the effect of resid- this is a relatively rigorous numerical discussion,
on steel deformation. To address this subject ual stress on hydrogen embrittlement and stress practical examples also are provided.
properly, contributing factors to overall steel de- corrosion cracking (SCC) of steel. This discus- Residual Stress Formation in the Shaping of
formation problems, including material effects, sion includes the effect of hydrogen on structure Materials contains four articles. The first article
machining, heating and cooling, must be exam- and transformation of steel, types of hydrogen covers residual stress in the steel forming pro-
ined. embrittlement, delayed fracture in steel, crack cesses. The steel forming processes included are
This handbook contains 27 articles, divided initiation and growth, SCC of low alloy steels, cold forming such as wire drawing, and hot
into five sections: Effect of Materials and Pro- crack initiation and growth mechanism of SCC forming such as extrusion, rolling, and forging.
cessing, Measurement and Prediction of Resid- processes, methods of estimating sensitivity to The effects of residual stresses involved in these
ual Stress and Distortion, Residual Stress For- SCC, effect of alloying elements on resistance processes are reviewed, and specific topics in-
mation in the Shaping of Materials, Residual to SCC, and the role of structure and thermal clude residual stress in cold metal forming such
Stress During Hardening Processes, and Resid- processing in SCC. as bending of sheet, drawing of wire, rod, and
ual Stress Formation During Manufacturing Pro- In the section Measurement and Prediction of tube, and residual stresses in deep drawn cup,
cesses. Residual Stress and Distortion, the first article sunk tubes, and radial forging products.
There are five articles in the section Effect of describes a number of simple, inexpensive de- The effect of final shaping prior to heat treat-
Materials and Processing. Material Factors flection (dissection) methods used to estimate re- ment on residual stress formation is discussed in
discusses the effects of various material proper- sidual stress of various types of components. The the next article. The effects of shaping processes
ties such as thermal properties and the interac- methods include Almen strip; Navy C-Ring; including grinding, milling, turning, shot peen-
tions of residual stresses on the transformation plate or bar slitting and deflection; tube slitting ing, and straightening on residual stress are dis-
products formed and steel deformation during and opening; and bending of bars, H-beams, and cussed. Also discussed is distortion after final
fabrication. Transformation plasticity is dis- channels. part shaping and experimental and computa-
cussed in some detail along with the use of mod- The next article provides an overview of re- tional studies of these processes.
eling to better understand these processes. sidual stress measurement methods. Topics in- The next article provides a practical overview
Prestress Engineering of Structural Material clude residual stresses arising from various of the factors affecting residual stress and dis-
provides a global design approach to understand- manufacturing processes, measurement methods tortion during final part shaping. Included are
ing the effects of residual stress generated during including strain measuring technique, post-stress discussions of influence of component shape on
surface engineering manufacturing processes relaxation measurement, sectioning and material heat treatment distortion, the effect of cross-
such as PVD and CVD on the material properties removal methods. In addition, strain measure- section size and asymmetry, effect of heat treat-
obtained. Some of the topics discussed in this ment methods such as x-ray and neutron diffrac- ing procedure and machining process on final
chapter include developments in the measure- tion, ultrasonic, birefringent and laser, optical component shape, effect of sequence of heat
ments of residual stresses, advanced mechanical gages, brittle coatings, Barkhausen noise, and treating and machining, influence of machining
surface treatments, and modeling of fatigue be- chemical coatings are discussed. Semidestruc- allowance and stress relieving procedure, influ-
havior taking residual stresses into considera- tive methods such as blind hole drilling and ring ence of residual stresses caused by cutting, meth-
tion. coring are discussed. ods of manufacturing blanks and effect of origi-
The effect of residual stresses on fatigue be- Measurement of residual stresses in coatings nal structure, hot-rolled steels or forgings and
havior is discussed in detail in the next article. and thin films is important because their influ- effect of banded segregation and carbide segre-
Examples of topics discussed include stability of ence on mechanical and physical properties af- gation, influence of heat treating methods, the
residual stresses, some aspects of fatigue in fect component service performance. Stress De- effect of heating including the rules of heating,
vi
2002 ASM International. All Rights Reserved. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

G.M. Webster V.V. Zabilsky W. Zinn


Union Carbide Corporation Physical Technical Institute, Ural Branch of Institut fur Werkstofftechnik
Tarrytown, NY USA RAS Universitat Kassel, Germany
Ijevsk, Russia
W. Wubbenhorst
Metal Improvement Company, Inc.
Unna, Germany

iv
2002 ASM International. All Rights Reserved. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

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 Induction Hardening.......................................................... 220
J. Lu J. Grum

Residual Stresses and Fatigue Behavior.....................................27 Hardening by Reheating and Quenching .................................. 248
D. Lohe, K.-H. Lang, and O. Vohringer M. Narazaki, G.E. Totten, and G.M. Webster

Stability of Residual Stresses .................................................54 Metallo-Thermo-MechanicsApplication to Quenching ................ 296


D. Lohe and O. Vohringer T. Inoue

Control of Residual Stress Formation and Steel Deformation


Effect of Residual Stress on Hydrogen Embrittlement
during Rapid Heating and Cooling...................................... 312
and Stress Corrosion Cracking.............................................70
N.I. Kobasko, V.S. Morganyuk and V.V. Dobrivecher
A.I. Kovalev, V.P. Mishina, D.L. Wainstein, and V.V. Zabilsky
Effect of Cryogenic Cooling on Residual Stresses, Structure,
Measurement and Prediction of Residual Stress and Distortion and Substructure ........................................................... 331
Ioan Alexandru and Vasile Bulancea
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


Stress Determination in Coatings........................................... 118
R.W. Lewis, K.N. Seetharamu and A.Y. Hassan
J.Albert Sue and Gary S. Schajer
Residual Stress Formation during Casting:
Methods for Determination of Inhomogeneous Residual Continuous and Centrifugal Casting Processes........................ 372
Stress Fields ................................................................ 125 D.-Y. Ju
I.A. Razumovsky, M.V. Medvedev, and A.V. Fomin
Residual Stress Formation Processes during Welding and Joining.... 391
Residual Stress Formation in the Shaping of Materials W. Zinn and B. Scholtes

Residual Stress in the Forming of Materials.............................. 141 Residual Stresses in Powder-Metal Processing........................... 397
Z. Wang and B. Gong P. Ramakrishnan

Residual Stress Formation and Distortion of Rail Steel................. 424


The Effect of Final Shaping Prior to Heat Treatment ................... 150 F.D. Fischer and G. Schleinzer
T. Ericsson
Residual Stresses during Gear Manufacture .............................. 437
Factors Affecting Final Part Shaping ...................................... 159 K. Funatani
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. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

Preface

Control of steel deformation is one of the most steels, influence of residual stresses on cyclic de- termination in Coatings provides a guide for
common concerns within the metals processing formation behavior, influence of residual stresses measuring residual macrostress in coatings, Spe-
industry. Numerous surveys have been con- on crack initiation and propagation, and effect of cific topics include origin of residual stresses in
ducted by various organizations in recent years residual stresses on S-N curves; an overview of coatings and residual stress measurement meth-
to assess the critical needs of the industry. In modeling of the effect of residual stresses on fa- ods including the deflection method, x-ray dif-
nearly every survey that has been conducted, dis- tigue behavior is provided. fraction, and hole-drilling. A comparison of
tortion is either the greatest or second greatest The next article provides an overview of the these methods is provided.
concern among the steel heat treating commu- stability and relaxation behavior of macro and The last article in this section provides a de-
nity. Steel distortion control will exhibit tremen- micro residual stresses in steel due to thermal tailed review of methods used to measure and
dous effects on the profitability of the commer- and mechanical treatments. This discussion in- subsequent data analysis of inhomogeneous re-
cial enterprise. Therefore, it is not surprising that cludes relaxation of residual stresses by anneal- sidual stress fields. This discussion includes re-
the ability to understand the overall distortion ing, residual stress relaxation by uniaxial defor- sidual stress as an inverse problem of experi-
process and to be able to design solutions to this mation, and relaxation by cyclic deformation. mental mechanics, indicator crack method of
problem typically rank very high on these same Hydrogen embrittlement of metals, as well as measuring residual stress, arbitrary cut-out in-
surveys. other types of brittle fracture, result from nucle- dicator method, and experimental methods and
In view of the enormous visibility and impor- ation and development of micro-cracks caused equipment including photoelastic coating
tance of steel deformation problems, the editors by internal stresses. The last article in this sec- method, and optical interferometry. Although
decided to put together an engineering handbook tion provides an overview of the effect of resid- this is a relatively rigorous numerical discussion,
on steel deformation. To address this subject ual stress on hydrogen embrittlement and stress practical examples also are provided.
properly, contributing factors to overall steel de- corrosion cracking (SCC) of steel. This discus- Residual Stress Formation in the Shaping of
formation problems, including material effects, sion includes the effect of hydrogen on structure Materials contains four articles. The first article
machining, heating and cooling, must be exam- and transformation of steel, types of hydrogen covers residual stress in the steel forming pro-
ined. embrittlement, delayed fracture in steel, crack cesses. The steel forming processes included are
This handbook contains 27 articles, divided initiation and growth, SCC of low alloy steels, cold forming such as wire drawing, and hot
into five sections: Effect of Materials and Pro- crack initiation and growth mechanism of SCC forming such as extrusion, rolling, and forging.
cessing, Measurement and Prediction of Resid- processes, methods of estimating sensitivity to The effects of residual stresses involved in these
ual Stress and Distortion, Residual Stress For- SCC, effect of alloying elements on resistance processes are reviewed, and specific topics in-
mation in the Shaping of Materials, Residual to SCC, and the role of structure and thermal clude residual stress in cold metal forming such
Stress During Hardening Processes, and Resid- processing in SCC. as bending of sheet, drawing of wire, rod, and
ual Stress Formation During Manufacturing Pro- In the section Measurement and Prediction of tube, and residual stresses in deep drawn cup,
cesses. Residual Stress and Distortion, the first article sunk tubes, and radial forging products.
There are five articles in the section Effect of describes a number of simple, inexpensive de- The effect of final shaping prior to heat treat-
Materials and Processing. Material Factors flection (dissection) methods used to estimate re- ment on residual stress formation is discussed in
discusses the effects of various material proper- sidual stress of various types of components. The the next article. The effects of shaping processes
ties such as thermal properties and the interac- methods include Almen strip; Navy C-Ring; including grinding, milling, turning, shot peen-
tions of residual stresses on the transformation plate or bar slitting and deflection; tube slitting ing, and straightening on residual stress are dis-
products formed and steel deformation during and opening; and bending of bars, H-beams, and cussed. Also discussed is distortion after final
fabrication. Transformation plasticity is dis- channels. part shaping and experimental and computa-
cussed in some detail along with the use of mod- The next article provides an overview of re- tional studies of these processes.
eling to better understand these processes. sidual stress measurement methods. Topics in- The next article provides a practical overview
Prestress Engineering of Structural Material clude residual stresses arising from various of the factors affecting residual stress and dis-
provides a global design approach to understand- manufacturing processes, measurement methods tortion during final part shaping. Included are
ing the effects of residual stress generated during including strain measuring technique, post-stress discussions of influence of component shape on
surface engineering manufacturing processes relaxation measurement, sectioning and material heat treatment distortion, the effect of cross-
such as PVD and CVD on the material properties removal methods. In addition, strain measure- section size and asymmetry, effect of heat treat-
obtained. Some of the topics discussed in this ment methods such as x-ray and neutron diffrac- ing procedure and machining process on final
chapter include developments in the measure- tion, ultrasonic, birefringent and laser, optical component shape, effect of sequence of heat
ments of residual stresses, advanced mechanical gages, brittle coatings, Barkhausen noise, and treating and machining, influence of machining
surface treatments, and modeling of fatigue be- chemical coatings are discussed. Semidestruc- allowance and stress relieving procedure, influ-
havior taking residual stresses into considera- tive methods such as blind hole drilling and ring ence of residual stresses caused by cutting, meth-
tion. coring are discussed. ods of manufacturing blanks and effect of origi-
The effect of residual stresses on fatigue be- Measurement of residual stresses in coatings nal structure, hot-rolled steels or forgings and
havior is discussed in detail in the next article. and thin films is important because their influ- effect of banded segregation and carbide segre-
Examples of topics discussed include stability of ence on mechanical and physical properties af- gation, influence of heat treating methods, the
residual stresses, some aspects of fatigue in fect component service performance. Stress De- effect of heating including the rules of heating,
vi
2002 ASM International. All Rights Reserved. www.asminternational.org
Handbook of Residual Stress and Deformation of Steel (#06700G)

quenching and system design, tempering, and metallo-thermo-mechanics and numerical simu- first part describes manufacturing of ferrous P/
equipment and racking. lation methodology with practical examples. M parts including powder characteristics, com-
A more focused, but practical, discussion on In the article on the control of residual stress paction in rigid dies, isostatic compaction, sin-
the effect of process equipment design on dis- formation and steel deformation during rapid tering, heat treatment of P/M parts, hot pressing,
tortion follows. Subjects that are covered include heating and cooling, a particular emphasis is on roll compaction, powder forging, metal injection
distortion generating process equipment, meth- intensive quenching. This is the first detailed, molding, spray forming, warm compaction, and
ods that may be used to minimize equipment- article-length discussion of this old, but little- rapid prototyping. The second part discusses re-
related distortion, quench system design, and known technology in the western world. Topics sidual stresses in P/M processing including
press quenching. include mathematical model for calculation of powder production, compaction of metal pow-
Residual Stress During Hardening Processes thermal and stress-strain state, computation of ders, sintering of metal powders, pressure sin-
contains eight articles. The first article provides stress-strain state, possibility of predicting hard- tering and hot isostatic pressing, heat treatment
a detailed discussion on the residual stresses in ening cracks, predicting the deformation of bear- of P/M parts, and microstructural development
carburized, carbonitrided, and case-hardened ing rings during hardening, thermal stresses and properties.
components. Topics include process considera- formed in carburized steel products due to ex- Residual Stress Formation and Distortion of
tions for carburized and carbonitrided compo- cessive cooling rates, generalization of compu- Rail Steel covers the cooling process including
nents, transformations and stress evolution in tational and experimental results for heating and the cooling boundary conditions and heat trans-
carburized and case-hardened components, ef- cooling of parts with different geometries and fer, residual stress state analysis, weight and fric-
fect of heat treating operations on residual stress thermal and physical fundamentals of processing tionthe rail end problem, experimental results;
distribution, relationship between residual of high-strength materials. roller straightening including residual stresses in
stresses and properties of carburized parts and An often contradictory subject is the cryo- unused roller-straightened rails, behavior of rail
modeling and prediction of residual stress field. genic processing of steels, and the detailed over- steel under plastic deformation, simulation of
The article on residual stresses in steel nitrid- view of the effect of cryogenic cooling on resid- roller straightening; and rails in service includ-
ing includes a discussion of nitrided layer struc- ual stress is presented here. Specific topics ing residual stresses due to welding and residual
ture as a function of nitriding process, residual include role of residual stresses within marten- stress formation in rolling contact.
stresses in nitrided layers, influence of residual sitic transformation at cryogenic temperatures, The last article provides a detailed description
stresses on fatigue behavior of nitrided steel evaluation of residual stresses after cryogenic of residual stress formation during hypoid gear
components, and modeling and prediction of re- cooling, influence of cryogenic cooling on resid- manufacture. It includes an overview of residual
sidual stresses in nitrided steel components. ual stresses and dimensional stability of steels, stress formation in carburized and hardened
The article on residual stress formation in in- and influence of cryogenic cooling on the struc- work, profiles and peak magnitudes of residual
duction hardening processes include an over- ture and substructure of steels. stresses, measurement methods including the
view of the induction hardening process and The practical use of controlled shot peening Sachs hole-drilling method, x-ray and neutron
steels used for this process, magnetic flux con- to induce compressive residual stresses is de- diffraction, influence of steel properties on resid-
centrators, conditions in induction heating and scribed in detail next. This discussion includes a ual stresses, influence of carburizing process pa-
quenching of machine parts, residual stress sur- historic overview of shot peening, elementary rameters on residual stress formation, benefits of
face profiles after induction surface hardening, processes of shot peening, workpiece and ma- residual stresses on fatigue strength, and the ef-
stress profiles in the machine part in the loaded terial process parameters, process monitoring, fects of hardness, case depth, intergranular oxi-
state, workpiece distortion in induction surface process optimization, x-ray diffraction, and in- dation, influence of shot peening, change of re-
hardening, induction surface hardening of gear dustrial examples. sidual stresses during fatigue, and distortion of
wheels, fatigue strength of materials, and resid- In Residual Stress Formation During Manu- carburized and hardened steels.
ual stresses after induction surface hardening and facturing Processes, the first article includes an The preparation of a text of this scope was a
finish grinding. extensive discussion of residual stress and de- tremendous task. The editors are deeply indebted
The next article provides an overview of re- formation problems arising from the casting pro- to many colleagues for their patience, support,
sidual stresses and distortion resulting from re- cess, and modeling of residual stress formation and assistance; without them this text would not
heating and quenching. Topics include phase during casting. Discussion includes finite ele- have been possible. Special thanks go to the
transformation during heat treating including ment analysis of heat flow during casting, for- ASM staff who often labor in the background
steel transformations, TTT and CCT diagrams, mulation of the elasto-viscoplastic stress model, but who are vital members of the team. Particu-
metallurgical crystal structure, estimation of vol- and deformation of a solidifying material. larly, thanks go to Veronica Flint and Carol Ter-
umetric change due to steel transformation upon The next article describes residual stress for- man of ASM International for their help and en-
quenching, cooling of steel with and without mation during the casting process, and it in- couragement.
metallurgical transformation, tempering, basic cludes continuous and centrifugal casting. Top- Very special thanks go to our families for their
distortion mechanism, relief of residual stresses, ics discussed include inelastic behavior and seemingly unending support. Without their un-
material movement due to thermal gradients dur- unified constitutive theory of metallic material derstanding and encouragement, this project
ing heating and cooling, material, component in solidification, analytical method of the would never have been completed.
and process effects, retained austenite, quench thermal-mechanical problem for the casting pro- George E. Totten, Ph.D., FASM
severity and uniformity and process design ef- cess, residual stress formation during semicon- Editor
fects on distortion, quench distortion and crack- tinuous casting, residual stress formation during G.E. Totten & Associates Inc.
ing, quenchant selection, measurement and eval- centrifugal casting, and residual stress formation Stony Point, NY USA
uation of quenching power, estimation of heat during strip casting by the twin-roll method.
transfer coefficient, wetting behavior and non- The origin and assessment of residual stresses Prof. Maurice A.H. Howes, Ph.D. (Retired)
uniform quenching, surface conditions, and during welding or brazing is discussed next. Editor
quench process modeling and simulation of re- Welding residual stresses are discussed includ- Worcestershire, England
sidual stress and distortion after quenching. ing residual stresses due to shrinkage, quench- Prof. Tatsuo Inoue, Ph.D., FASM
A detailed approach to modeling and simula- ing, and phase transformations. Characteristic Editor
tion of residual stress and distortion applied to residual stress distributions in brazed compo- Department of Energy Conversion Science
quench processing follows. This discussion is nents is also discussed. Faculty of Energy Science
based on a metallo-thermo-mechanics approach, The article Residual Stresses in Powder Kyoto University
and topics discussed include an overview of Metal Processing is divided into two parts. The Kyoto, Japan
vii
Handbook of Residual Stress and Deformation of Steel Copyright 2002 ASM International
G. Totten, M. Howes, T. Inoue, editors, p3-10 All rights reserved.
DOI: 10.1361/hrsd2002p003 www.asminternational.org

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

RESIDUAL STRESSES are a consequence of tions that follow; others such as elastic modulus tion C1 corrects C LV{TD /T} to a specific heat at
interactions among time, temperature, deforma- and thermal conductivity still have to be mea- constant pressure. Ce is the electronic specific
tion, and microstructure (Fig. 1). Material or sured for individual alloys. heat coefficient, and C lP is the component of the
material-related characteristics that influence the specific heat capacity due to magnetism. Figure
development of residual stress include thermal 2 illustrates the data for ferrite and austenite in
conductivity, heat capacity, thermal expansivity, Heat Capacity pure iron. Whereas it is well known that ferrite
elastic modulus and Poissons ratio, plasticity, undergoes a paramagnetic to ferromagnetic tran-
thermodynamics and kinetics of transformations, The dominant contribution to specific heat ca- sition on cooling below 1042.15 K, the magnetic
mechanisms of transformations, and transfor- pacity comes from lattice vibrations (phonons), properties of austenite are seen from Fig. 2 to be
mation plasticity. since the majority of free electrons are prevented of some consequence in determining the heat ca-
Many general statements can be made about from participation in heat absorption by the Pauli pacity. There are two coexisting electron states
the role of material factors in the evolution of exclusion principle. However, for iron and its of austenite, one of which is ferromagnetic with
residual stress. Spatial variations in temperature alloys, a further important contribution comes a Curie temperature of 1800 K and the other of
give rise to nonuniform thermal strains, the ef- from magnetic changes. The net specific heat ca- which is antiferromagnetic with a Neel tempera-
fect of which becomes exaggerated when the pacity can therefore be factorized into three com- ture of 55 to 80 K (Ref 4). The balance between
material is elastically stiff and has a high yield ponents: these states changes with temperature, giving
strength. A large thermal conductivity helps re- rise to corresponding changes in heat capacity.
T C
duce residual stress by reducing temperature gra- TD The data in Fig. 2 are for pure iron, but there
dients (Ref 2). The dissipation or absorption of CP{T } C LV 1 CeT C Pl{T } (Eq 1)
is now sufficient understanding of the compo-
heat depends not only on the external environ- nents of heat capacity to enable similar estimates
ment of the component but also on internally where C LV{TD /T} is the Debye specific heat func- for iron alloys, using internationally available
generated heatfor example, during adiabatic tion and TD is the Debye temperature. The func- computer programs and thermodynamic data-
deformation or due to the latent heat of transfor-
mation. Similarly, the plastic strain distribution
in the component depends both on the constitu- Table 1 Physical properties that affect the development of residual stress in steels
tive properties and on how the shape deforma-
tions due to phase transformations compensate Temperature, C (F)

for the development of stress. Property Phase(a) 0 (32) 300 (570) 600 (1110) 800 (1470)
The fundamental material properties are, of Elastic modulus, GPa c 200 175 150 124
course, temperature dependent. Table 1 illus- P 210 193 165 120
b 210 193 165 120
trates how several key properties might vary 200 185 168 ...
with temperature (Ref 3). Some of these prop-
Poisson ratio c 0.291 0.309 0.327 0.345
erties, which can to some extent be estimated P 0.280 0.296 0.310 0.325
quantitatively, are discussed in detail in the sec- b 0.280 0.296 0.310 0.325
0.280 0.296 0.310 ...
Thermal expansivity, K1 c 2.1 105
P 1.4 105
b 1.4 105
1. Thermal stress 1.3 105
Time, Stress,
temperature strain Thermal conductivity, W/m K c 15.0 18.0 21.7 25.1
4. Heat of P 49.0 41.7 34.3 27.0
6. Latent deformation b 49.0 41.7 34.3 27.0
heat 5. Stress- 43.1 36.7 30.1 ...
induced Specific heat capacity, 106 J/m3 K c 4.15 4.40 4.67 4.90
transformation P 3.78 4.46 5.09 5.74
2. Temperature- b 3.78 4.46 5.09 5.74
dependent phase 3. Transformation 3.76 4.45 5.07 ...
transformations strain Yield strength, MPa c 190 110 30 20
P 360 230 140 30
Microstructure b 440 330 140 30
1600 1480 1260 ...

Fig. 1 The coupling of temperature, stress, and micro- (a) , P, b, and represent allotriomorphic ferrite, pearlite, bainite, and martensite, respectively. Source: Ref 3
structure. Source: Ref 1
4 / Effect of Materials and Processing

bases (Ref 6). After all, changes in fundamental a reflection of the two coexisting electronic deal with higher levels of complexity, are not yet
thermodynamic quantities such as enthalpy and states of austenite (c0 and c1), each with a ther- applicable to practical alloys.
entropy are derived from heat capacity data. It is mal expansion coefficient that is identical to that
surprising that this capability has not yet been of ferrite. The c0 component has the lower molar
exploited in any calculation of residual stress, volume and is the antiferromagnetic form, Plastic Deformation
even though the methodology is widely avail- whereas the denser c1 form is ferromagnetic. The
able. relative proportion of atoms in the c0 and c1
The familiar mechanisms of plastic deforma-
states changes with temperature, so that the ap-
tion are slip, mechanical twinning, and creep.
parent expansion coefficient of austenite as a
Phase transformations also cause permanent de-
Expansion Coefficient and Density whole, as detected experimentally, is much
formation (Ref 811). In steels, austenite can de-
larger than that of ferrite (Fig. 3).
compose into a large variety of microstructures
Table 1 shows that the expansion coefficient The molar volumes (in cm3 /mol) of c0, c1, c,
that are distinguished by the atomic mechanism
of austenite is larger than that of ferrite; this and over the temperature range of 300 to 1775
of transformation (Fig. 4). In a displacive trans-
might be considered surprising given the lower K are:
formation, the change in crystal structure is
density of ferrite. However, the behavior is again achieved by a deformation of the parent struc-
V cm0 6.695(1 2.043 105T 1.52 108T 2) ture. A reconstructive transformation is one in
which the change in structure is achieved by a
V cm1 7.216(1 2.043 105T 1.52 108T 2) flow of matter, which occurs in such a way that
40
strains are minimized.
30 V cm{T } (1 y)V m
c0
{T } y V m
c1
{T } All the transformations cause changes in
shape (Fig. 5a), which for reconstructive trans-
Heat capacity, CP, J mol 1K1

20
V am{T } formations simply reflects the change in density.
Austenite
10 7.061(1 2.043 105T 1.52 108T 2) For displacive transformations, the shape change
is an invariant-plane strain (IPS), that is, a com-
0 where y is the fraction of atoms of austenite in bination of a shear on the invariant plane and a
the c1 state, the details of which can be found dilatation normal to that plane. The strain energy
75 elsewhere (Ref 4, 5). associated with a constrained IPS is minimized
These data are for pure iron, but thermody- when the product phase has a thin-plate shape.
Ferrite
50 namic data can be used to assess how the expan- This is why Widmanstatten ferrite, bainite, acic-
sion coefficients would change with alloying, ular ferrite, and martensite in steels grow in the
since there are quite sophisticated treatments of form of plates. The distinguishing features of a
25
the effect of solute elements on the magnetic and variety of deformation modes are compared in
other components of the free energies of iron. Table 2, and Table 3 describes the shape defor-
0 Note that the two electronic states picture of mations.
0 300 600 900 1200 1500 1800
austenite is a simplification of the real scenario, The permanent strain caused by any transfor-
Temperature, K
but first-principles calculations (Ref 7), which mation is called transformation plasticity. A
Fig. 2 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 Cu-
rie, c, and c d transitions. d-ferrite is simply an Displacive
alternative historical name for high temperature . Source: Reconstructive Invariant-plane strain shape
Ref 5 Diffusion of all atoms during deformation with large shear
nucleation and growth. component. No iron or
Sluggish below about 850 K substitutional solute diffusion.
Thin plate shape
8.0


7.8
Allotriomorphic Widmansttten ferrite
ferrite Carbon diffusion during
7.6
paraequilibrium nucleation
Volume, cm3 mol1

V1 and growth
7.4 Idiomorphic
ferrite
V
Bainite and acicular
7.2
ferrite
V
Massive ferrite Carbon diffusion during
7.0 No change in paraequilibrium nucleation.
bulk composition No diffusion during growth
6.8
V0
Pearlite Martensite
6.6 Cooperative growth Diffusionless
0 300 600 900 1200 1500 1800
of ferrite and nucleation and
Temperature, K cementite growth

Fig. 3 Molar volumes of the various forms of iron.


Source: Ref 5
Fig. 4 Transformation products of austenite. Source: Ref 12
Material Factors / 5

phase change in a stress-free material is usually Phase transformation can also compensate for Deformation System
triggered by heat treatment, when the parent stress. Greenwood and Johnson (Ref 13, 14)
phase passes through an equilibrium transfor- showed that when a phase change is accompa- Displacive transformations can be regarded as
mation temperature. Alternatively, the applica- nied by a change in volume, the tensile strain modes of plastic deformation. Just as a combi-
tion of a stress in isothermal conditions can trig- expected when transformation occurs under the nation of a plane and a direction constitutes a
ger transformation in circumstances where it influence of a tensile stress r is given by: deformation system for slip or twinning, the
would not otherwise occur. Unusual effects can habit plane and displacement vector of the
occur when stress and temperature work to- 5 DV r invariant-plane strain accompanying displacive
gether. The transformation may occur at remark- e (Eq 2)
6 V rY transformation completely describe the defor-
ably low stresses or at very small deviations mation system responsible for transformation
from the equilibrium temperature. This is why where rY is the yield stress of the weaker phase plasticity. The displacement vector describes the
even minute stresses can greatly influence the and DV/V is the transformation volume strain. sense of the macroscopic displacements accom-
development of microstructure, and vice versa. The role of shear strains associated with trans- panying transformation and, along with the habit
It is not surprising that transformation plasticity formation has been emphasized in later work by plane indices, also contains information about
can be obtained at stresses that are much smaller Magee and Paxton (Ref 15, 16), and subse- the magnitude of the shear component and dil-
than the conventional yield stress of the parent quently by Fischer (Ref 17), Leblond et al. (Ref atational component of the displacements. Typ-
phase. 1822), Olson (Ref 23), and Bhadeshia et al. ical data for the deformation systems associated
(Ref 24). Not only does transformation affect with transformations are listed in Table 4. Note
Transformations, Residual Stresses, stress, but the latter modifies the development of that reconstructive transformations involve only
microstructure. The microstructure tends to be a volume change together with diffusional mass
and Related Phenomena more organized when transformation occurs in a flow, so it is not appropriate to regard them as
stresss parent phase, because the stress favors deformation systems in the present context.
The strains due to phase transformations can the formation of certain orientations relative to Given the cubic crystal structure, and the fact
alter the state of residual stress or strain. It is well others. This is illustrated schematically in Fig. that habit planes tend to be irrational, there will
known that the martensitic transformation of the 5(b) to (d). These aspects will now be discussed in general be 24 of these systems per austenite
carburized surface of a steel component puts the in more detail, because transformation plasticity grain, and they may operate simultaneously to
surface under compression. It is argued that this can radically alter the state of residual stress. varying extents. Of course, unlike ordinary slip,
is because of the expansion at the surface due to
formation of the lower-density martensite from
austenite.
Table 2 Characteristics of different modes of deformation
Slip Mechanical Displacive Reconstructive
Characteristic deformation twinning transformation transformation
Causes permanent change in shape Yes Yes Yes Yes
Reconstructive Invariant-plane strain shape changewith a large shear component Yes Yes Yes No
Changes crystallographic orientation No Yes Yes Yes

Changes lattice type No No Yes Yes
Single Can lead to a density change No No Yes Yes
crystal

Displacive
Table 3 Shape change due to transformation
Transformation Shape change (a) s(b) d(b) Morphology

(a) Allotriomorphic ferrite Volume change 0.00 0.02 Irregular


Idiomorphic ferrite Volume change 0.00 0.02 Equiaxed, faceted
Pearlite Volume change 0.00 0.03 Spherical colonies
Widmanstatten ferrite Invariant-plane strain 0.36 0.03 Thin plates
Bainite Invariant-plane strain 0.22 0.03 Thin plates
Acicular ferrite Invariant-plane strain 0.22 0.03 Thin plates
Martensite Invariant-plane strain 0.24 0.03 Thin plates
 Cementite plates Invariant-plane strain? 0.21? 0.16? Thin plates
Mechanical twins () Invariant-plane strain 1/2 0.00 Thin plates
Annealing twins (c) 0.00 0.00 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)


Table 4 Deformation systems associated with transformations
Fig. 5 Shape changes accompanying unconstrained
transformations. Note that the horizontal scale Phase Habit plane indices Displacement vector m
bars are all the same length. (a) The two kinds of shape
Martensite (0.363 0.854 0.373) [0.195 0.607 0.771] 0.185
changes that occur when a single crystal of austenite trans-
forms to a single crystal of ferrite, as a function of the mech- Bainite (0.325 0.778 0.537) [0.159 0.510 0.845] 0.27
anism of transformation. (b) Polycrystalline sample of aus- Widmanstatten ferrite (0.506 0.452 0.735) [0.867 0.414 0.277] 0.36
tenite. (c) Polycrystalline sample of austenite that has
partially transformed by a displacive transformation mech-
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
anism into a random set of ferrite plates. (d) Polycrystalline
approximate, since the habit plane and displacement direction are usually irrational. The displacement vector does not quite lie in the habit plane
sample of austenite that has partially transformed by a dis- 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
placive transformation mechanism into an organized set of including the shear and the dilatational components.
ferrite plates.
6 / Effect of Materials and Processing

the different deformation systems within an aus- nary notion of work hardening does not apply. the volume change. The corresponding interac-
tenite grain cannot intersect, except in special Work hardening nevertheless manifests itself via tion with displacive transformations is much
circumstances where intervariant transforma- a different mechanism, in which the stability of larger because of the shear component of the
tions are possible, as is the case with some the austenite increases as it becomes ever more IPS.
shape-memory alloys. It follows that the ordi- finely divided. For displacive transformations, the influence
The Taylor/von Mises criterion (Ref 25, 26) of stress on the transformation can be expressed
states that in any given crystal, a minimum of as a mechanical driving force (DGmech), which is
N five independent slip systems is necessary to the work done by the external stress in producing
A (applied tensile stress) produce an arbitrary shape change. A crystal in the macroscopic shape deformation (Ref 30, 31):
a polycrystalline aggregate has to accommodate
the arbitrary deformations of neighboring grains.
DGmech rNd ss (Eq 3)
Therefore, a polycrystalline material is brittle
max unless each grain contains at least five indepen-
dent slip systems. Similar logic can be applied where rN is the normal stress on the habit plane
d to the crystallographic variants of a phase gen- and s is the component of the shear stress on the
erated by displacive transformation. The habit habit plane that is parallel to the direction along
plane is predicted theoretically (Ref 27, 28) and which the shear displacements of the shape de-
Resolution of the applied stress, rA. The normal found experimentally (Ref 29) to have irrational formation occur (Fig. 6). The strains d and s are
Fig. 6
stress, rN, and the shear stress, s, both act on the indices. This means that there exist, in principle, the dilatational and shear components, respec-
habit plane. The vector d is the direction along which lie 24 possible variants of the habit plane per grain
the shear displacements of the shape deformation. smax is
tively, of the shape deformation. Some typical
the maximum shear stress on the habit plane, but s is given
of austenite (that is, 24 independent deformation values of the mechanical driving force terms are
by resolving smax along d. Note that d differs slightly from systems). Given this large number of transfor- given in Table 5. Given a free choice of some
the displacement vector of the IPS, which includes a dila- mation variants available per grain, the Taylor 12 to 24 crystallographic variants of the trans-
tational component in addition to the shear. criterion leads to the conclusion that transfor- formation product in each grain of austenite, the
mation plasticity can cause, or accommodate, work done by the shear stress is always expected
any externally imposed, arbitrary shape to be positive, whereas that due to the dilata-
Table 5 Typical values of the mechanical changeassuming that a sufficient quantity of tional component depends on the sign of rN. For
driving force coefficients parent phase is available. It follows that poly- steels, this latter component is relatively small.
Nature of stress DG/r, J/(mol MPa)
crystalline samples can remain intact at grain Any observed consequences of stress must there-
boundaries when transformation plasticity is the fore reflect the dominant role of the shear com-
Uniaxial tension 0.86
Uniaxial compression 0.58 sole mode of deformation. ponent unless the stress is purely hydrostatic.
Elastic crack tip (a) 1.42 Since the shear stress remains positive irre-
spective of whether the sample is pulled in ten-
(a) The stress state for the crack tip is multiaxial, but the coefficient is Mechanical Driving Force sion or uniaxially compressed, and since the
calculated by expressing the stress in terms of the von Mises equivalent
tensile stress. Source: Ref 32 shear component of the shape change is large, a
The interaction of an applied elastic stress uniaxial stress will always cause a temperature
with a phase change can occur in two ways: increase for displacive transformations in steels.
1. The stress can alter the driving force for the Hydrostatic stress, on the other hand, has no de-
transformation. viatoric components and consequently interacts
Change in bainite-start temperature

2. The stress can change the appearance of the only with the dilatational component of the
Tensile
microstructure by favoring the formation of shape change. Thus, hydrostatic compression is
+ expected and found to lead to a decrease in the
those variants which best comply with the ap-
Compression transformation temperature (Fig. 7); some data
plied stress.
0 (Ref 32) on the sensitivity of the transformation
For reconstructive transformations, only the temperature to applied stress are presented in Ta-
hydrostatic component of stress can interact with ble 6.

Hydrostatic
compression 1000
Chemical
Limits to Stress-Assisted
Free energy, J mol1

Stress (below austenite yield strength)


Transformation
Indication of how the transformation-start tem- 0 Mechanical
Fig. 7
perature (for Widmanstatten ferrite, bainite, acic-
ular ferrite, or martensite) should vary as a function of the At temperatures close to that at which the
nature and magnitude of an applied stress whose magni-
1000 equilibrium transformation occurs, an applied
tude is less than that of the yield stress.
Total stress can assist reaction when the chemical driv-
ing force is insufficient to achieve the change on
2000 its own. There must exist a point, however, when
Table 6 Sensitivity of transformation-start 300 400 500 600 700 800 900
the applied stress simply cannot provide enough
temperatures in steels to applied stress Temperature, C mechanical driving force to complement the
Phase Nature of stress Sensitivity, K/MPa
Typical magnitudes of the chemical and me-
chemical term to give a driving force large
Martensite Pressure 0.06
Fig. 8 enough to induce transformation. After all, the
chanical driving forces for stress-affected trans-
Bainite Pressure 0.09 formation. The mechanical driving force is estimated for an magnitude of the stress that can be applied is
Eutectoid Pressure 0.011 applied stress that is equal to the yield stress of austenite. limited by the yield point of the parent phase.
Martensite Tensile 0.06 Since this yield stress becomes small at high temperatures,
the contribution of the mechanical driving force also de-
Thus, there are limits to what can be achieved
Source: Ref 32 creases. Therefore, transformation becomes impossible as by the application of stress as a stimulus to trans-
the temperature exceeds about 700 C (1290 F). formation (Fig. 8).
22 / Effect of Materials and Processing

100 100
0 Axial stresses Tangential stresses
Residual stresses, MPa
under tensile loading 0 under tensile loading

Residual stresses, MPa


100 Measurement Measurement
100
Calculated Calculated
200
200
300
300
400
400
500
500
600
600
0 0.05 0.10 0.15 0.20 0.25 0.30 0 0.05 0.10 0.15 0.20 0.25 0.30
Depth, mm Depth, mm

100 100
0 Axial stresses 0 Tangential stresses
under torsion loading
Residual stresses, MPa

Residual stresses, MPa


under torsion loading
100 Measurement 100 Measurement
Calculated Calculated
200 200
300 300
400 400
500
500
600
600
700
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Depth, mm Depth, mm
Fig. 25 Initial distribution of the residual stresses

Stress. The initial residual stress distribution is elastic-plastic properties of material, because the der the cyclic loading can be obtained. Part of
calculated from the residual stresses measured plastic deformation is the main cause of the re- the results and their comparison with the test val-
by the x-ray diffraction method in the upper laxation. So, the cyclic behaviors of material are ues have been shown in Fig. 27. In the case of
layer where initial stresses are introduced by sur- very important for the prediction of the residual the traction-compression loading, two loading
face treatment. In depth, the stresses are calcu- stresses. However, they can be measured by ex- levels were calculated. The same level for tor-
lated for the structural equilibrium. Figure 25 periment. sion loading with two different cycle numbers
shows the comparison of initial residual stresses Figure 26 shows the cyclic behaviors of was also analyzed. Figure 28 shows the stabi-
between the calculation and the experiment. 35NCD16 grade steel, a cyclic softening mate- lized residual stresses for different traction-com-
Prediction of the Residual Stress Relaxa- rial. pression load levels that are used in the predic-
tion. In order to predict the relaxation of the re- When the stabilized residual stresses are cal- tion of the fatigue life. Unfortunately, it is
sidual stresses, a simplified method proposed by culated, the behavior corresponding to NR /2 is impossible to compare with the experiment re-
Lu et al. has been used (Ref 30). It supposes that used. After an elastic-plastic calculation with sults.
the relaxation of the residual stresses depends on FEM, the relaxation of the residual stresses un- Prediction of the Fatigue Life. For predict-
ing the fatigue life under a complex stress state,
it is necessary to have two S-N curves. One is
Rec(MPa) = 920 55,14 log N
for simple traction or traction compression; an-
other is for alternated torsion. These curves are
1 cycle basic data of a material, and they can be obtained
2 N, Re(Y.S.), by experimental method.
1000
3 cycles MPa
15 Using the program FATIGUE3D, a design
900 50 1 940 tool on fatigue developed in the LASMIS labo-
/2, MPa

150 ratory, a distribution of fatigue life and a distri-


2 920
800
NR/2 3 910 bution of safety factor for a structure can be ob-
15 870 tained. In this example, the fatigue life under
50 830 traction-compression cyclic loading and torsion
700
150 800 loading has been calculated. The stabilized re-
35 NCD 16 sidual stresses play a role as static load. Figure
600 (Rm (UTS) = 110 MPa) NR/2 660
29 shows the results of fatigue life and their com-
parison with the experimental results. It is very
0.5 0.75 1 clear that the proposed method is available.
t/2, %
Prediction of Admissible Residual Stress.
In this approach, a calculation method has been
The cyclic behavior of 35NCD16 grade steel. Re, yield stress; R ec, cyclic yield stress as a function of number developed that can predict the admissible resid-
Fig. 26 of cycles; Det, axial plastic strain range during cyclic loading ual stress for a given fatigue life. This tool can
Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 25

ity control tool. For industrial applications, fu- grant Brite-EuRam, BRRT-CT985090, EN- Austenitic Stainless Steel, Residual Stress,
ture developments are necessary: SPED project), and the National Science Foun- V. Hauk, H.P. Hougardy, E. Macherauch,
dation of China (two bases project) is acknowl- and H.D. Tietz, Ed., DGM, Verlag, 1992, p
Measurement techniques edged. 891900
Improvement of ultrasonic and magnetic 12. J.B. Roelens, F. Maltrud, and J. Lu, Deter-
measurement methods mination of Residual Stresses in Submerged
Integration of the portable optical method for REFERENCES Arc Multi-Pass Welds by Means of Numer-
strain measurement in destructive techniques ical Simulation and Comparison with Ex-
Processing and materials perimental Measurements, Welding in the
1. J. Lu, Ed., Handbook of Measurement of Re- World, Vol 33 (No. 3), 1994, p 152159
Development for the industrialization of new sidual Stresses, Society for Experimental 13. Proceedings of the National Conference on
prestress processes: ultrasonic shot peening, Mechanics, USA, 1996 the Residual Stress and Design Office, CE-
laser shock, surface nanocrystallization, and 2. J. Lu, A. Niku-Lari, and J.F. Flavenot, Lat- TIM, (Senlis), 1991
so on. est Development of Residual Stresses Mea- 14. D. Deslaef, E. Rouhaud, and J. Lu, Finite
Optimization of residual stress in advanced surement by the Hole Drilling Method, Ma- Element Modeling of Shot-Peening Pro-
materials: functional gradient coating system, teriaux et Techniques, Dec 1985, p 709718 cesses, Proceedings of the SEM Spring Con-
for example, plasma spray coating, PVD, 3. Z. Wu, J. Lu, and B. Han, Study of Residual
ference on Experimental and Applied Me-
CVD, or diamond-like carbon (DLC) coating; Stress Distribution by a Combined Method
chanics, 13 June 1998, (Houston, Texas),
metal-matrix composite; electronic packag- of Moire Interferometry and Incremental
p 419422
ing; intermetallic materials Hole Drilling. Part I: Theory, J. Appl.
Mech., ASME, Sept 1998 15. A. Chabenat, and J.F. Flavenot, Measure-
Modeling 4. Z. Wu, J. Lu, and B. Han, Study of Residual ment of Residual Stress on the Quenched
Axles after Heating by Induction, Traite-
Development of advanced tools for the pre- Stress Distribution by a Combined Method
of Moire Interferometry and Incremental ment thermique/Revue de metallurgie,
diction of initial residual stress in various pro-
Hole Drilling. Part II: Implementation of (Heat treatment/Journal of Metallurgy), No.
cesses: forging, casting, machining, welding,
plasma spray coating, composite processing, Applied Mechanics, J. Appl. Mech., ASME, 124, 1978, p 233239 (in French)
electronic packaging manufacturing, mechan- Sept 1998 16. N. Skalli, and J.F. Flavenot, Prediction of
ical surface treatment 5. Z. Wu and J. Lu, Residual Stresses by Moire Fatigue Life by Simulation with the Resid-
Integration of the residual stress modeling Interferometry and Incremental Hole Drill- ual Stress Considerations, Conference pro-
method into a global life design (3-D) and the ing, Experimental Mechanics, Advances in ceedings, Spring Conference, SF2M
creation of optimization codes for the design Design, Testing and Analysis, Vol 2, I.M. (French Society for Metal and Materials),
of mechanical components using the concur- Allison, Ed., Conf. ICEM 98, 2428 Aug 2223 May 1984, p 98117 (in French)
rent engineering approach 1998, (Oxford), p 13191324 17. H.P. Lieurade, P. Castelucci, J.F. Flavenot,
6. Z. Wu, X.K. Niu, J. Lu, and P.G. Ifju, Study and J. Lu, Efficiency of Improvement Tech-
With the recent progress in the field of re- of Process-Induced Stress in Orthotropic niques on the Fatigue Strength as a Function
search and development, it is now possible to Composite LaminateCarbon/Epoxy [O2 / of the Type of Welded Joint, Welding in the
introduce a new area of research: prestress en- 902]2s, Proceedings of the SEM Spring World, Vol 31 (No. 4), 1991, p 268271 (in
gineering of structural materials (PESM or Conference on Experimental and Applied French)
PRESMA). In the near future, the mechanical Mechanics, 13 June 1998, (Houston, 18. H.P. Lieurade, C. Maillard-Salin, and M.
engineer will be able to use these tools in a pre- Texas), p 179181 Truchon, Fatigue Cracking of Welded Joints
stressed concrete approach to civil engineering. 7. Z. Wu, J. Lu, and Y. Guo, A New Method of High Strength Steel, Proceedings of
The ENSPED program in Europe is an excellent of Residual Stress Measurement on Elec- IABSE, Colloquium, Lausanne, 1982, p
example of the cooperation of industrial and re- tronic Packaging, Experimental Mechanics,
search partners. Using this new approach, the fa- 137144 (in French)
Advances in Design, Testing and Analysis, 19. D. Francois, Role of the Residual Stress on
tigue behavior and mechanical behavior of new Vol 2, I.M. Allison, Ed., Conf. ICEM 98,
materials can be considerably improved, increas- the Fracture, Conference Proceedings, Re-
2428 Aug 1998, (Oxford), p 987991 sidual Stress in Welding Structure, CETIM,
ing two- or threefold in some cases. In fact, many 8. J. Lu, J.F. Flavenot, and S. Thery, Study on
failures occur in the area close to the surface. SENLIS, Dec 1987, p 8797 (in French)
the Effect of the Finishing Treatment on the 20. W.H. Friske, Shot Peening to Prevent the
Improving the mechanical behavior is often only Residual Stress Gradient in SiC Reinforced
necessary on the surface (fatigue, fretting fa- Corrosion of Austenitic Stainless Steels,
Aluminum Metal Matrix Composite, J. Rockwell International Cooperation, 1975
tigue, wear, stress corrosion). As a result, this Compos. Technol. Res., ASTM, Vol 12 (No.
type of approach can be expected to develop 21. A. Niku-Lan`, M. Menier, and G. Beranger,
4), 1990, p 232238
very quickly in order to introduce adequate pre- Effect of the Surface State on the Stress
9. J. Li, J. Lu, M. Perrin, M. Ceretti, and A.
stressing treatment in the area indicated by the Corrosion Behaviour of a Stainless Steel,
Lodini, Study on the Residual Stress in
advanced design tool. Cold-Rolled 7075 Al-SiC Whisker Rein- CETIM Informations, No. 78, Dec 1982
forced Composites by X-ray and Neutron 22. C. Richard, J. Lu, J.F. Flavenot, and G. Be-
Diffraction, J. Compos. Technol. Res., July ranger, Study of Residual Stress in the
ACKNOWLEDGMENTS
1995, p 194198 Plasma Sprayed NiCrAlY Coating and
The author is grateful to all the people at LAS- 10. J. Lu, P. Peyre, C. Omam Nonga, A. Ben- Characterization of Interface Toughness,
MIS working in the field of residual stress for amar, and J.F. Flavenot, Mechanical Surface Memoires et Etudes Scientifiques, Revue de
their help during the preparation of this paper: Treatments, Current Trends, and Future metallurgie, May 1991, p 295306 (in
Prof. J.L. Chaboche, Dr. D. Retraint, Dr. E. Rou- Prospects, Surface Modification Technolo- French)
haud, Dr. X.L. Gong, Dr. L. Couturier, Dr. R. gies VIII, T. S. Sudarshan and M. Jeandin, 23. C.S. Richard, G. Beranger, J. Lu, and J.F.
Akrache, Dr. Z. Wu, Dr. S. Rasouli Yazdi, Mr. Ed., TMS, Sept 1994, p 589601 Flavenot, The Influence of Heat Treatments
A. Milley, Mr. D. Deslaf, Mr. B. Guelorget, Mr. 11. A. Benamar, J. Lu, J.F. Flavenot, P. Bar- on the Adhesion of Plasma Sprayed Ni-
A. Voinier, Mr. F. Belahcene, Mr. K. Gong. Fi- barin, G. Chalant, and G. Inglebert, Mod- CrAlY Coatings, Surf. Coat. Technol., Vol
nancial support from the European Union (under eling Residual Stresses by Cold Rolling an 82, 1996, p 99109
Material Factors / 9

lowed by those that do not. This also carries the not yet calculable in the same manner. It may be Kang, and M.E. Glicksman, Ed., Minerals,
implication that the interaction of the stress is the case that they are insensitive to alloying, but Metals and Materials Society, 1992, p 21
with the growth process (that is, the IPS shape that remains to be demonstrated in the context 30
deformation) rather than the strain field of the of residual stress analysis. 3. R. Schroder, Mater. Sci. Technol., Vol 1,
nucleus, which is likely to be different. It is There is little doubt that transformations in 1985, p 754764
worth noting that there are similar results for steel play a major role in the development of 4. L. Kaufman, E.V. Clougherty, and R.J.
martensite: most favored variants grow first in residual stresses. For reconstructive transforma- Weiss, Acta Metall., Vol 11, 1963, p 323
the sequence of transformation under stress (Ref tions (for example, pearlite), it is the difference 335
15, 24). in density between the parent and product phases 5. L. Kaufman, in Energetics in Metallurgical
that contributes to transformation plasticity. The Phenomenon, Vol III, W.M. Mueller, Ed.,
plasticity can be much larger for displacive Gordon and Breach, 1967, p 5584
Summary transformations (Widmanstatten ferrite, bainite, 6. K. Hack, Ed., The SGTE Casebook: Ther-
martensite) because of the large shear compo- modynamics at Work, Institute of Materials,
Many of the thermal properties of steelsfor nent of the shape deformation when these trans- 1996
example, heat capacity, thermal expansion co- formation products form. These are quite so- 7. D.G. Pettifor and A.H. Cottrell, Ed., Elec-
efficients, and latent heats of transformation phisticated effects which, with few exceptions, tron Theory in Alloy Design, Institute of
are remarkably well understood. Indeed, com- are not incorporated in most residual stress anal- Materials, 1992
mercially available thermodynamic databases yses. 8. J.W. Christian, in Decomposition of Austen-
and programs can be used to estimate these ite by Diffusional Processes, V.F. Zackay
quantities as a function of temperature and and H.I. Aaronson, Ed., Interscience, 1962,
chemical composition. This capability has not REFERENCES p 371386
been exploited in the analysis of residual 9. H.M. Clark and C.M. Wayman, Phase
stresses, even though phase diagram calculations 1. T. Inoue and Z. Wang, Mater. Sci. Technol., Transformations, American Society for
using the same software are now routine in in- Vol 1, 1985, p 845850 Metals, 1970, p 59114
dustry and academia. 2. P.W. Fuerschbach, in The Metal Science of 10. J.D. Watson and P.G. McDougall, Acta Me-
Other properties, such as elastic modulus, are Joining, M.J. Cieslak, J.H. Perepezko, S. tall., Vol 21, 1973, p 961973
11. J.W. Christian, Physical Properties of Mar-
tensite and Bainite, Special Report 93, Iron
and Steel Institute, 1965, p 119
12. H.K.D.H. Bhadeshia, Bainite in Steels, 2nd
+ L + ed., Institute of Materials, 2001, p 1453
T L, T
13. G.W. Greenwood and R.H. Johnson, Proc.
Strain

Strain

0 0 Roy. Soc., Vol 238, 1965, p 403422


14. R.H. Johnson and G.W. Greenwood, Na-
 
ture, Vol 195, 1962, p 138139
15. C.L. Magee, Ph.D. thesis, Carnegie Mellon
University, 1966
16. C.L. Magee and H.W. Paxton, Trans. Met.
Transformation time Equivalent time
(a) Soc. AIME, Vol 242, 1968, p 17411749
17. F.D. Fischer, Acta Metall. Mater., Vol 38,
1990, p 15351546
+ L + 18. J.B. Leblond, G. Mottet, and J.C. Devaux,
L J. Mech. Phys. Solids, Vol 34, 1986, p 395
T T 409, 411432
Strain

Strain

0 0 19. J.B. Leblond, J. Devaux, and J.C. Devaux,


Int. J. Plasticity, Vol 5, 1989, p 551572
  20. J.B. Leblond, Int. J. Plasticity, Vol 5, 1989,
p 573591
21. J.B. Leblond and J. Devaux, Residual
Transformation time Equivalent time Stresses, Elseviers, 1989, p 16
(b) 22. J.B. Leblond, Internal Report CSS/L/NT/
90/4022, FRAMASOFT, 1990, p 112
L L 23. G.B. Olson, Deformation, Processing and
+ + Structure, American Society for Metals,
1982, p 391424
Strain

Strain

0 0 24. H.K.D.H. Bhadeshia, S.A. David, J.M. Vi-


tek, and R.W. Reed, Mater. Sci. Technol.,
  T Vol 7, 1991, p 686698
T
25. R. Von Mises, Z. Angew. Math. Mech., Vol
8, 1928, p 161
Transformation time Equivalent time 26. G.I. Taylor, J. Inst. Met., Vol 62, 1928, p
(c) 307
27. J.K. MacKenzie and J.S. Bowles, Acta Me-
Fig. 13 Schematic of the reported variations (Ref 24) in longitudinal and radial strains during the isothermal formation tall., Vol 2, 1954, p 138147
of bainite under the influence of a tensile load, presented alongside predictions (Ref 36) from the crystallo-
graphic/thermodynamic model. The stresses are all intended to be below the austenite yield strength, and the data in this
28. M.S. Wechsler, D.S. Lieberman, and T.A.
case refer to uniaxial tension. (a) Zero stress, any temperature. (b) Small stress, low temperature. (c) Small stress, high Read, Trans. Amer. Inst. Min. Metall. Eng.,
temperature; or large stress, low temperature. Vol 197, 1953, p 15031515
10 / Effect of Materials and Processing

29. A.B. Greninger and A.R. Troiano, Trans. 32. S. Denis, E. Gautier, A. Simon, and G. Stresses in Welded Constructions, Welding
Amer. Inst. Min. Metall. Eng., Vol 140, Beck, Mater. Sci. Technol., Vol 1, 1985, p Institute, 1977, Paper 2
1940, p 307336 805814 35. H.K.D.H. Bhadeshia, Worked Examples in
30. J.R. Patel and M. Cohen, Acta Metall., Vol 33. W.K.C. Jones and P.J. Alberry, Ferritic the Geometry of Crystals, Institute of Met-
1, 1953, p 531538 Steels for Fast Reactor Steam Generators, als, 1987
31. L. Delaey and H. Warlimont, in Shape British Nuclear Engineering Society, 1977, 36. A. Matsuzaki, H.K.D.H. Bhadeshia, and H.
Memory Effects in Alloys, J. Perkins, Ed., p 14 Harada, Acta Metall. Mater., Vol 42, 1994,
TMS-AIME, Plenum Press, 1975, p 89114 34. W.K.C. Jones and P.J. Alberry, Residual p 10811090
Handbook of Residual Stress and Deformation of Steel Copyright 2002 ASM International
G. Totten, M. Howes, T. Inoue, editors, p11-26 All rights reserved.
DOI: 10.1361/hrsd2002p011 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 in- Residual stress is usually defined as the stress an objective mechanical behavior of the material
troduce residual stress into mechanical parts, that remains in mechanical parts that are not sub- and structure. Under this approach, the residual
which influences its fatigue behavior, fracture jected to any outside stresses. Residual stress ex- stress must not be considered as a parameter that
strength, and even its corrosion resistance. Few ists in practically all rigid parts, whether metallic only depends on the material processing condi-
metalworking methods exist that do not produce or not (wood, polymer, glass, ceramic, etc). It is tions, but must also be considered as a parameter
new stresses. The role of residual stress is, there- the result of the metallurgical and mechanical that can be optimized.
fore, very important when designing mechanical history of each point in the part and the part as There is an increasing interest in how the state
parts. Over the last few years, an increasing num- a whole during its manufacture. of residual stress affects the mechanical proper-
ber of studies have been carried out to under- In the case of structural materials, surface en- ties of a material and its structure. The failure of
stand the effects of residual stress on mechanical gineering can lead indirectly to the innovation of a structure or a mechanical component is not
performance. This article attempts to present a genuinely new materials based on conventional only due to external loads. Residual stress is an
global approach to including residual stress in materials. The frustrating slowness of some of important parameter in this respect. All manu-
expected fatigue life calculations, and the pos- these developments reflects not only difficulties facturing processes, for example, introduce a
sibility of introducing it into mechanical engi- in scaling up laboratory techniques, but also gen- new state of residual stress. This can have a posi-
neering design offices. The definitions and ori- eral conservation in the engineering industries tive effect, such as increasing the fatigue limit in
gins of residual stress according to production and a reluctance to change established habits. the case of surface compressive stress, or it can
methods are first presented. Then, shown are the One of the factors that contributes to slowing the have a negative effect, such as decreasing the
problems involved in correctly adapting these pace is that the current approach to surface en- stress corrosion behavior of a material in the case
modeling techniques for use in design offices gineering directly correlates performance to pro- of tensile residual stress.
and the industrial consequences of taking resid- cessing parameters; designers are left without the Basic and applied research in the field of re-
ual stress into account on quality assurance con- tools needed to optimize and take into account sidual stress has been stepped up in the last few
trol procedures. This article deals mainly with modified or prestressed surfaces for a particular years. Residual stress is taken into account in
application, and the processors are left without advanced design in the aerospace, automotive,
residual stress measurement techniques and the
appropriate goals for surface properties. and nuclear industries. Even the microelectron-
overall necessity to combine destructive (incre-
The recent European Network of Surface and ics industry is starting to take residual stress into
mental hole-drilling method) and nondestructive
Prestress Engineering and Design (ENSPED) account for the dimensional stability of elec-
(x-ray and neutron diffraction) methods in order
project led by the University of Technology of tronic packaging.
to precisely evaluate the residual stress distri-
Troyes and funded by the European Union is one The introduction of advanced materials has
bution. Some new results concerning optical also contributed to the development of knowl-
of the contributions currently being made to the
methods are also discussed. Shown, too, are the development of such tools. The aim is to develop edge in the field of residual stress. In fact, many
beneficial and harmful effects of residual stress a project with a hard core of about 18 partners new materials are multimaterials, for example,
on the resistance of structures or industrial com- representing a balanced selection of Europes metal-matrix composites, plasma-sprayed coat-
ponents, depending on whether they are tensile major industries, such as SNECMA; European ing, physical vapor deposition (PVD), and chem-
or compressive. The different models used to Aeronautic, Defense and Space (EADS); ABB; ical vapor deposition (CVD) coatings, which
predict residual stress induced by different types Siemens; Volvo; Fiat; British Aerospace; Robert contain residual stress as a result of the thermal
of processing are then presented. The last section Bosch; Hydro Aluminum; Wartisla; and so on. and mechanical incompatibilities of the different
shows the effect of residual stress on fatigue be- It is based on building a bridge between the phases of the material or structure.
havior. A model based on the finite element fields of surface modification and prestress pro- Figure 1 shows the different fields of research
method (FEM) for predicting the relaxation of cessing in materials engineering and computer- in which residual stress is taken into account and
residual stresses is presented. Finally, prediction aided design in mechanical engineering in order its relevance for industrial applications. Three
of fatigue life, taking residual stress into account, to find an appropriate way of approaching this main fields must be developed for a global ap-
using FEM is discussed. A new approach to con- interdisciplinary area. The members of the pro- proach of prestress engineering: measurement
current engineering applied to the design of me- ject will focus on the field of prestress process- techniques for the quality control and processing
chanical components with residual and applied ing. The main goal of the prestress engineering analysis, processing parameter optimization and
stress consideration is presented. approach is the optimization of residual stress for processing modeling, and modern design tool for
12 / Effect of Materials and Processing

the life cycle simulation with residual stress con- The first series of methods is based on de- ment method. The in-plane surface displace-
sideration. stroying the state of equilibrium in the mechan- ments Ux and Uy and out-of-plane surface dis-
ical component. The residual stress is then eval- placement Uz produced by the relaxation of
uated from its relaxation. However, it is only residual stresses are obtained using moire inter-
Origins of Residual Stress possible to measure the consequences of stress ferometry and Twyman-Green interferometry,
relaxation and not the relaxation itself (displace- respectively. Figure 2 shows three-dimensional
In general, macroscopic residual stress can be ment, fracture, strain). In most cases, the change displacement as a function of the drilling depth.
induced due to: in strain is selected as the parameter to be stud- The main advantage of this technique is the pos-
ied. The following procedure is used: sibility of studying the residual stress with an in-
Nonhomogeneous plastic flow under the ac- plane stress gradient (Ref 5). It is also possible
tion of external treatment (shot peening, au- 1. Creation of a new stress state by machining to study composite materials (Ref 6). In this
tofretting, roller burnishing, hammer peening, or layer removal case, the through thickness residual stress distri-
shock laser treatment) 2. Detection of the local change in stress by bution was evaluated ply by ply in a carbon fi-
Nonhomogeneous plastic deformation during measuring the strain or displacement ber/epoxy composite. Figure 3 shows an exam-
nonuniform heating or cooling (ordinary 3. Calculation of the residual stress as a function ple of residual stress determined by using this
quenching, molding of plastics) of the strain measured using the elastic theory technique.
Structural deformation from metalworking (analytical approach or numerical calcula- Recently, this method was used (Ref 7) to
(heat treatment) tions such as FEM) study the Plastic Ball Grid Array (PBGA) pack-
Heterogeneity of a chemical or crystallo- age, which is a cost-effective surface-mounting
During the recent years, the incremental hole-
graphic order (nitriding or case hardening) package with a high-density interconnection,

drilling method is extensively used. It is sensi-


Various surface treatments (enameling, low profiles, and light weight. It is currently used
tive to the first kind of residual stress, that is, the
nickel plating, chrome plating, PVD and in many electronic products, including portable
macroscopic residual stress. The principle of this
CVD coating) telecommunication and computing products. A

technique is simple. It involves monitoring the


Differences in expansion coefficients and me- typical structure in a PBGA package consists of
change in strain when a hole is drilled in a com-
chanical incompatibility of the different com- four layers: a plastic molding compound, a sili-
ponent with residual stress. These strain mea-
ponents of composites (composites with a con chip, a chip-attach adhesive layer, and an
surements can be related to the original residual
metallic and organic matrix, ceramic coat- organic chip carrier. Due to the coefficient of
stress distribution in the analyzed sample at the
ings) thermal expansion mismatch between the silicon
hole location. The relationship between the
Table 1 shows the different origins of residual strain and the residual stress can be calculated chip, the plastic compound, and the organic chip
stress for metalworking operations usually car- with the calibration coefficients Ain and Bin. The carrier, considerable residual stresses are devel-
ried out in the industry. To produce an industrial general approach used to determine the Ain and oped in the package during the assembly pro-
part, one or several of the techniques listed in the Bin FEM is detailed in Ref 2. Recently, the high cess. The process-induced residual stress can
table can be used. To calculate the residual stress sensitivity moire interferometer and incremental play a significant role in the reliability of elec-
existing in a part, the source of the stress must hole-drilling method for residual stress measure- tronic components and packages. Since a PBGA
be identified first. ment has been developed. The theoretical devel- package is small and the surface layer is made
Residual Stress Measurement Techniques opment of a combined method is introduced in of a plastic material, it has proved very difficult
in Global Approach and Quality Assurance. Ref 3 and 4. The relationship between the three- to use other existing methods of residual stress
Over the last few decades, various quantitative dimensional surface displacements produced by measurement. In this research work, a practical
and qualitative techniques have been developed introducing a blind hole and the corresponding method has been developed to determine resid-
(Ref 1). These techniques are used for the pro- residual stress is established by employing the ual stress for electronic packaging. In this
cessing optimization and quality control of ma- existing theoretical solution containing a set of method, blind holes are drilled into the speci-
terial. In general, a distinction must be made be- undetermined coefficients. The coefficients are mens, and relationships are established between
tween destructive and nondestructive methods. calibrated by the three-dimensional finite ele- the released surface displacement and the cor-
responding residual stress by introducing a set
of calibration coefficients. A multilayer three-di-
mensional FEM is established to determine the
relevant calibration coefficients. The surface dis-
placements are measured accurately in a small
Quality control Manufacturing
region around the hole. For a practical PBGA
package, the tensile residual stress is determined
in both the plastic molding compound and the
Measurement Prestress process glass/epoxy laminate chip carrier. The method
technique development and is accurate, simple, convenient, and practical.
modeling More applications in the field of electronic prod-
ucts are anticipated.
Take residual stress into The x-ray diffraction and neutron diffraction
account in integrated design methods are based on the measurement of lattice
strains by studying variations in the lattice spac-
ing of the polycrystalline material. The first
Relaxation of Effect of residual method measures the residual strain on the sur-
residual stress stress on behaviors face of the material, and the second measures the
residual strain within a volume of the sample.
Diffraction techniques can be used to study all
Dimensional New design Mechanical three types. The peak shift method is sensitive
stability design to the first two, whereas line broadening is sen-
sitive to the second and third types. These tech-
Fig. 1 Main research fields and industrial application fields in which residual stress is taken into account niques 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 widespread. Rapid ways of checking the residual the reference techniques. Figure 4 proposes a re-
materials, such as metal-matrix composites (in stress must therefore be developed. The methods sidual stress inspection plan. Other new tech-
matrix and in reinforcement) (Ref 8, 9). used industrially (x-ray diffraction and the incre- niques, such as neutron diffraction and optical
Residual stress can be incorporated into the mental hole method) will not be sufficient in the method, will be introduced in industry. A de-
design of mechanical components. Although this future. Other nondestructive testing (NDT) tech- tailed analysis of different techniques of mea-
leads to a better knowledge of the fatigue life of niques (ultrasound, magnetic methods, acoustic surement is out of the scope of this article.
parts and reduces the safety coefficient at the de- emission) are presently being developed. But as
sign stage, it also poses a host of new problems they currently stand, these techniques use physi-
on a quality assurance level. All statistical con- cal parameters that depend not only on the resid- Modeling of Process
trols are only applied today to a few critical com- ual stress present in the parts, but also on mi-
ponents in the aeronautical and nuclear indus- crostructural changes. In the near future, NDT The experimental techniques developed pre-
tries; this practice could easily become techniques will be applied at the same time as viously can contribute to the development of a
residual stress prediction model. The results cal-
culated by the model make it easier to take the
residual stress into account during the mechan-
ical design. The residual stress induced by the
thermal processing is extensively treated in the
other articles of this Handbook. Only some ex-
amples concerning the mechanical surface treat-
ments and the comparison between the numeri-
cal simulation and experimental validation are
presented briefly.
Two kinds of models can be mentioned: ana-
lytical 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 ther-
mal 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 nec-
essary to obtain good results in all directions. If
a two-dimensional calculation is used, the resid-
ual stress evaluation correlates well in one di-
rection 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-dimen-
sional finite element dynamic model was devel-
oped to obtain a better description of the shot-
Three-dimensional displacements measured for different hole depth by a combination of optical methods and peening process and to introduce this approach
Fig. 2 into the component design (Ref 14). With the
the incremental hole-drilling method in a shot peened sample. Champ, field

Table 1 Main origins of residual stress resulting from different manufacturing processes
Process Mechanical Thermal Structural
Casting No Temperature gradient during cooling Phase transformation
Shot peening, hammer peening, roller burnishing, laser Heterogeneous plastic deformation No Depends on the material
shock treatment, bending, rolling, chasing, forging, between the core and surface of the
straightening, extrusion part
Grinding, turning, milling, drilling, boring Plastic deformation due to the removal Temperature gradient due to heating Phase transformation during machining if
of chips during machining the temperature is sufficiently high
Quenching without a phase transformation No Temperature gradient No
Surface quenching with a phase change (induction, EB, No Temperature gradient Change of volume due to a phase change
laser, plasma, classical methods)
Case hardening, nitriding No Thermal incompatibility New chemical component with volume
modification
Welding Shrinkage Temperature gradient Microstructural change (HAZ)
Brazing Mechanical incompatibility Thermal incompatibility New phase at interface
Electroplating Mechanical incompatibility Mechanical incompatibility Composition of plating depending on bath
used
Thermal spraying (plasma, laser, HVOF) Mechanical incompatibility, Thermal incompatibility, temperature Change of phase in plating
microcracking gradient
PVD, CVD Mechanical incompatibility Mechanical incompatibility Change of phase
Composite Mechanical incompatibility Mechanical incompatibility No

EB, electron beam; HVOF, high-velocity oxygen fuel; HAZ, heat-affected zone
14 / Effect of Materials and Processing

help of three-dimensional modeling, the enor- of analysis seems promising for studying the re- Effect of Residual Stress on the
mous influence of shot interaction is verified by percussions of a large number of parameters. Mechanical Strength of Materials
simulating simultaneous impacts. This simula- However, these methods are only efficient when
tion is very similar to the industrial process. Re- a maximum number of experimentally measured Generality. When a part is subjected to a field
sults of the residual stress obtained by the simu- parameters can be introduced. They also need to of elastic residual stresses characterized by a ten-
lation are closer to the experimental results, if a be improved if they are to take most phenomena sor rR, on which is superposed a field of service
three-dimensional FEM is used (Fig. 5). The into account. The proposed models could be stresses defined by the tensor rS, the real stress
higher the coverage rate, the lower the intensity completed by analyzing the influence of the type to which the part is subjected is characterized by
of the stress. This study shows how FEM can be of contact, the direction of the shot during the the tensor rR rS (Fig. 6). If the residual
used to model shot peening and determine the impact, behavior of the material, and so on. It stresses are added to the service stresses (resid-
associated residual stress field. could then be possible to introduce the residual ual tensile stress, for example), the part is locally
The results obtained for both a single impact stress field into metal parts in order to study their overloaded due to residual stress. If, on the con-
and several impacts follow the general distribu- behavior, using numerical simulation, for several trary, an appropriate finishing operation (shot
tion of measured residual stress fields. This type types of mechanical and fatigue life tests. 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
70 is increased as a result.
Average per lamina fibers Average per lamina of epoxy Figure 7 shows the properties of materials that
60 Average in two laminas of fibers Average in two laminas of epoxy are influenced by residual stress. In the subse-
50 quent section, several quantitative examples of
40 the effect of residual stress are given.
Influence on the Fatigue Strength (Initia-
30
tion Crack Phase). Residual stress plays an ex-
Residual stress, MPa

20 tremely important role with respect to the fatigue


10 strength of materials. It can be considered to be
a mean or static stress superimposed on the cy-
0
112.5 225 337.5 450 562.5 675 787.5 900 clic stress. As the mean stress rm increases, the
10 fatigue strength decreases. This is demonstrated
20 in the Haigh and Goodman diagrams.
30
y
40
50
Hole depth, m S
R
Fig. 3 Residual stress distribution in the half-thickness of a carbon fiber/epoxy composite

R + S
M
I M
Mechanical components
manufacturing lines

Continuous control using


NDE techniques (US, Fig. 6 Superposing of residual stress and service stress
magnetic, AE) 100
Good
quality 50
Defect
0 Fatigue
0.10 0.20 0.30 0.40
Measurement of the residual Assembly
50
stress distribution through of parts
Residual stress, MPa

the thickness of the part 100 Corrosion


using conventional techniques under tensile Breaking
(x-ray, hole drilling) 150 stress
200 Residual
stress
250
Statistical analysis Tensile
300 srr L = 0.3 mm Friction
of results strength
szz L = 0.3 mm
350

400
Improvement of Interfacial
process 450 Dimensional bond strength
Depth, mm stability of coating

Fig. 4 Residual stress inspection plan for the purposes Fig. 5 Modeling of the residual stress distribution in-
of quality assurance. NDE, nondestructive eval- duced by shot peening using the 3-D finite ele- Fig. 7 Effect of residual stress on the performance of
uation; US, ultrasonic; AE, acoustic emission ment method materials
Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 15

Quenching treatment, after induction heating, leads to a tangential residual stress equal to or lindrical bars, quenched after induction heating,
introduces very high residual compressive stress slightly greater than the longitudinal stress. The and subjected to repeated bending stress (Ref
into the hardened layer, which results from the thickness of the material subjected to residual 16). The results obtained are presented in Table
increase in volume of the martensitic structure compressive stress is in the same order of mag- 2. It can be seen that the higher the residual com-
with respect to the ferrito-pearlitic structure (this nitude as the layer transformed during treatment pressive stress, the greater the fatigue strength.
applied to the treatment of annealed steel, for (Ref 15). Fatigue tests were carried out by the The resulting gain in fatigue strength produced
example). In induction-quenched cylindrical French Technical Center for Mechanical Indus- by the residual stress can be as much as 50% of
bars, the residual stress on the surface usually try (CETIM) on 36 mm diameter XC42 steel cy- 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
Table 2 Effect of quenching conditions and residual stress on fatigue strength joints (Ref 17). Three cases are show: as-welded
(residual tensile stress), stress relieved (no resid-
Fatigue limit Residual stress stabilized
ual stress), and shot peened (residual compres-
after 5 106 at the fatigue limit, MPa (ksi) sive stress). A marked increase in the fatigue
Type and depth of treatment Surface cycles, MPa (ksi) Longitudinal Transverse strength was observed in the case of shot peen-
at 45 HRC, mm (in.) hardness, HRC rm ra stress stress ing.
A, induction 2.7 (0.11) 5556 596 (87) 584 (85) 128 (19) 468 (68) Influence on Fatigue Failure (Propagation
243 (35) 571 (83) Phase) and Brittle Fracture (Ref 13, 18, 19).
B, induction 4.2 (0.17) 5556 623 (90) 610 (88) 273 (40) 583 (85)
341 (49) 676 (98)
In the case of welded assemblies, the presence
C, induction 4.7 (0.19) 5459 670 (97) 660 (96) 655 (95) 603 (87) of welding defects at the weld toe and the geo-
D, water quenched after through 6061 780 (113) 750 (109) 863 (125) 1132 (164) metric profile of the latter generally lead to a
heating without stress-relieving 777 (113) 1156 (168) limited period of crack initiation. The cracking
annealing 3.5 (0.14)
phase must be considered by taking into account
the residual stress field induced by the welding
Source: Ref 16
operation.
The decisive influence of the residual stress
field on the crack propagation speed has been
demonstrated (Ref 18). Figure 9 shows the re-
Sample type State
6
2.10, MPa BX, MPa
Crack sults of cracking as a function of the residual
initiation site stress. Relieving residual stress by heat treatment
changes the crack propagation speed consider-
As welded 207 40
ably when the stress is high.
In the case of a brittle fracture, cleavage starts
Post weld heat treated 207 37 Weld toe in a grain when the local stress reaches a critical
1 value of rf* and it generally propagates without
Shot peened 392 519 difficulty in the adjacent grains by producing a
Multipass brittle fracture. The tensile residual stress rR, in
addition to the applied stress r, initiates this type
Fig. 8 Effect of residual stress on the fatigue strength of E690 welded joints. rRX, residual stress perpendicular to the of failure for low loads, such that:
fillet (rRX); rRI, principal maximum residual stress evaluated at a depth of 0.1 mm below the surface. Source:
Ref 17
r rR rf*

Once cracking has been initiated, the applied


106 stress alone can be enough to allow propagation
to continue at a high speed. Failure is therefore
ZAT D.T.
(40-75 mm) very sudden. Residual stresses that facilitate the
ZAT B.S.
initiation of brittle fracture by cleavage are there-
Nontreated metal fore very dangerous for steels under load at low
(40 mm)
temperature. This is why the stress relieving of
107 welded joints is also recommended.
Grain slips come up against inclusions and
da/dN, m/s

create concentrated stresses at their interface that


lead to fracture of either the interface or the in-
clusion. Cavities then appear for a critical initi-
ation stress and grow by plastic deformation of
108 the matrix until their coalescence leads to ductile
Rs = 0.1
fracture at least on a microscopic level. The
E36Z speed at which the cavity grows is not only pro-
portional 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
109 stress. Coalescence is a plastic instability phe-
5 10 20 30 40 50 60 70 80
nomenon that no doubt occurs for a critical cav-
K, MPa m ity size. Tensile residual stress not only facili-
Effect of a residual stress-relieving treatment on the cracking speed in the HAZ (butt-welded assembly of an
tates the initiation of cavities but, by increasing
Fig. 9 the mean stress, also accelerates growth. These
E36Z steel). ZAT D.T., HAZ of the heat treated sample; ZAT D.T., HAZ of the as-welded sample. Source:
Ref 18 two effects combine to decrease the critical elon-
16 / Effect of Materials and Processing

gation of ductile fracture. However, this is only used, can remain in the coatings and in the sub- defect. Reference 27 gives a very methodical ap-
important if the ductility is already very low in strates. They are of several types: microstresses proach to defining the criteria and processes re-
the absence of residual stresses, since plastic de- in the grain, produced during cooling, and lating to relieving stress in welded structures.
formation can eliminate them. macrostresses affecting the entire coating. The same type of reflection can be applied to
Effect on Stress Corrosion (Ref 20, 21). Macrostresses are created not only by cooling other types of parts.
Stress corrosion is a mechanical and chemical but also by the difference in temperature be-
cracking phenomenon that can lead to failure un- tween the substrate, the sprayed layer, and the
der the combined effect of tensile stress and a outside surface. The differential contraction thus Taking Residual Stress into Account
corrosive environment. Cracking is generally produced between the various materials, due to when Calculating the Expected
transcrystalline and can appear on all types of the difference in physical and mechanical prop- Fatigue Life
materials, such as aluminum alloys, steels, cop- erties, determines the stresses in the coating and
per, titanium and magnesium. The introduction the coating-substrate interface. These stresses In the previous section, the different effects of
of residual compressive stress can considerably therefore influence the mechanical and thermo- residual stress on the mechanical strength of
increase the fatigue life of parts subjected to mechanical behavior of the coated parts. structures and materials were mentioned. Al-
stress corrosion. Tests carried out on magnesium In order to appreciate the quality of a coating, though the ability to quantitatively estimate the
test specimens placed under stress in a salt so- three types of damage to parts in service can be fatigue life taking residual stress into account is
lution gave the following results: considered: just beginning, it is still too early to extend these
Ground test specimen: failure after two min- The coating deteriorates rapidly. predictions to other types of stress that are far
utes The properties of the substrate are modified more complex and involve physical and chemi-
Shot peened test specimen: no cracking after by the coating. cal phenomena. Statistics show that failures of a
12 days under the same conditions The damage is common to both materials. It purely mechanical origin are mainly due to fa-
is located at the interface and jeopardizes both tigue. It is for the reasons indicated previously
The tests conducted by W.H. Friske show that that this article only addresses problems con-
the fatigue life is 1000 times greater for a shot- the adhesion and the fatigue life.
cerning the prediction of fatigue life. Two arti-
peened 304 grade stainless steel part than it is C. Richard et al. have shown that decreasing cles (H. P. Lieurade and A. Pellissier-Tanon) in
for a non-shot-peened part (Ref 20). Tests car- the residual stress by thermal treatment of the Ref 13 deal with the question of predicting the
ried out by CETIM on Z6CN18.9 stainless steel coating considerably improves adhesion at the effect of residual stress on crack propagation
produced similar results (Ref 21). interface (Ref 23). Figure 10 illustrates the effect phase. Although, they concern welded struc-
Effect on Adhesion of Coatings (Ref 22 of residual stress. It can be seen that the apparent tures, the concepts developed in these two ref-
25). Most coatings are produced for a specific toughness of the interface is improved by 100% erences can be applied to other types of struc-
reason, particularly to improve the corrosion and when heat treatment is applied. There is a high tures. By limiting the approach to prediction of
wear resistance of the base material, or to pro- level of residual tensile stress in the test speci- the fatigue life to the fatigue cracking initiation
vide a thermal barrier for use at high tempera- men without heat treatment. When the level of stage, the problem of predicting the fatigue life
ture. But this is only achieved if the coating ad- residual tensile stress increases, the true tough- of mechanical components subjected to a high
heres to the substrate correctly. Adhesion ness of the coating decreases. An increase in the cycle fatigue can be analyzed.
therefore indicates correct preparation of the sur- residual compressive stress produces the oppo- Calculating the Effect of Residual Stress on
faces to be coated and the quality of the coating site effect. the Fatigue Strength. Based on the experimen-
operation. The last few years have seen the ap- Influence of Residual Stress on the Tensile tal results mentioned previously, it would seem
pearance of plasma-spraying techniques, both at Strength, Friction, Wear and Dimensional that a linear relationship of the Goodman type
atmospheric pressure and at reduced pressure. Stability. The effect of residual stress on the ten- can be used to take residual stress into account:
These processes offer a high degree of flexibility sile strength is obvious, particularly in structures
for coatings in critical areas. However, high re- made of composite materials or when the pre- rD
sidual stress, inherent to the coating method stressed layer is very thick compared with the ra rD (rm rR) (Eq 1)
Rm
thickness of the parts. In composites, residual
stress is produced as a result of the thermal and
where ra is the amplitude of admissible stress,
mechanical incompatibility of the reinforce-
R
rm is the mean fatigue stress, rD is the purely
7 ments and matrix. This can influence the mac-
reverse tensile fatigue limit, Rm is the true rup-
Kc roscopic properties of composites under tensile
6 300 ture strength, and rR is the residual stress mea-
or compressive stress (Ref 26).
sured in the direction of the applied service
Little research has been carried out on the ef-
5 stress. The numerous studies mentioned in Ref
fect of residual stress on friction and wear prop-
28 show that the effect of residual stress is
Residual stress, MPa

4 200 erties. Their role is often masked by other param-


greater when the properties of the materials are
Kc, MPa m

eters. The increase of hardness during treatment


high.
3 and changes in the toughness and adhesion of
If we try to represent the development of ra
antiwear coatings due to residual stress can con-
2 100 according to the residual stress rR by an equation
siderably affect the resistance to friction. Up un-
of the following type:
1 til the present, this effect has been integrated into
the global parameter of adhesion. In the future,
ra rD rR (Eq 2)
0 0 work will be carried out to try to determine the
real effect of residual stress.
1 the experimental results generally show that
The problem of dimensional stability has been
increases with the strength of the material; for
2
known for a long time. When a part is machined
100 example, in the case of machining stresses in an
TT STT that contains residual stress produced by heat
XC38 grade steel, Syren et al. found:
treatment or welding, the shape of the part can
Influence of heat treatment on the residual
Fig. 10 change after operation due to the relaxation of 0 in the annealed state
stress and the toughness of the interface: case
of plasma-sprayed coatings at atmospheric pressure. TT, residual stress. This is why stress-relieving treat- 0.27 when quenched and tempered
heat treated; STT, as-sprayed ments are frequently used to avoid this type of 0.4 when quenched
Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 17

Unfortunately, these results are in contradic- the cyclic behavior of the materials. However, invent new criteria, but it leads to complica-
tion with an equation of the Goodman type. In Syrens results show that relaxation is much tions because of the increasing number of param-
Eq 2, the coefficient is none other than what greater when the mechanical properties are eters that need to be determined. Even with a
is usually called the endurance ratio: lower. linear relationship of the Dang Van type, two
When these experimental results are used with Wohler curves have to be determined to obtain
rD
the residual stress measured after carrying out a at least the two points needed to produce the di-
fatigue test, and therefore stabilized, it is some- agram. If additional constants are added, the test
Rm
times possible to use an equation such as Eq 2. plane will be even greater, which means that the
In the case of the fatigue bending test on cylin- criterion cannot be used in industry. As a result,
This parameter decreases as the rupture strength drical XC42 steel bars quenched after induction the criterion to be used must be simplified as
of steels increases. heating (Table 2), the fatigue test results for the much as possible. This case deals with radial
This apparent contradiction is probably ex- different treatments correspond perfectly to the loading problems (r1 K1 r2 K2 r3), and a
plained by the fact that the residual stress relax- Haigh diagram, provided any possible influence relationship of Eq 3 is sufficient. To simplify
ation phenomenon has not been taken into ac- of transverse residual stress on the fatigue stress matters further, the Crossland or Dang Van cri-
count. The value of the residual stress rR to be is ignored (Fig. 11). terion can be used. In the case of combined and
introduced into equation of type 1 or 2 men- It is not possible, however, to extend these out-of-phase loading, new criteria have been de-
tioned previously, must correspond to the stabi- results to all materials and to the different manu- veloped to take the out-of-phase effect into ac-
lized fatigue stress, or the coefficient of influence facturing processes that introduce residual stress. count (Ref 3639).
will include the relaxation process. The article Also, preliminary tests are needed to validate the But as yet, these criteria have not been vali-
by D. Lohe and O. Voehinger in this Handbook methodology. The use of residual stress in cal- dated in a study in which combined and out-of-
presents a large number of results concerning the culations based on endurance diagrams of the phase residual stresses have been taken into ac-
relaxation of residual stress under mechanical Haigh or Goodman type usually only allows for count. When the fatigue stress is complex, it is
and thermal loading. References 29 and 30 pro- an estimation of the increase in fatigue strength also very difficult to calculate the expected re-
vide the information on the relaxation mecha- as a function of the residual stress. Secondly, this sidual stress relaxation.
nism of residual fatigue stress as a function of approach only allows for the combination of uni- When fatigue cracks are initiated on the sur-
axial stresses. Yet the residual stresses produced face, the stresses to be taken into account are
by the various manufacturing methods used to biaxial; this gives the following for the Cross-
800 make the part are always multiaxial. The stresses land or Dang Van criterion:
on the surface are biaxial while those inside the
700 A 2
r 1a r 2a r r
part are triaxial. Depending on the area in which 2 2
600
B Repeated
the fatigue crack is initiated (on or below the socta 1a 2a
C bending tests 3
D surface), the biaxial or triaxial stresses need to
500 Repeated tension test r1a
a, MPa

be included when calculating the fatigue life. sa


400 This raises the problem of choosing a multiaxial 2
300 fatigue stress criterion. A simplified approach 1
based on an endurance diagram can therefore Pmax (r1a r2a r1m r2m r1R r2R)
200 3
only be an approximation.
100 The test carried out by the CETIM (Ref 16) where r1a, r2a represent the amplitude of the
0 shows that the traditional Mises and Tresca cri- main reversed fatigue stresses (r1a r2a); r1m,
0 1000 2000 teria can only be used in the presence of higher r2m represent the average value of the main fa-
m, MPa
mean or residual stress. In this case, it is pref- tigue stresses; and r1R, r2R are the residual stress
Use of Haigh diagrams to take longitudinal re-
erable to use criteria (Ref 31) that include the values measured in the two main directions (sta-
Fig. 11 amplitude of octahedral shearing (socta) or the bilized values).
sidual stress into account (XC42 steel
quenched after induction heating). A through D represent maximum shearing (sa) and maximum hydro- To use the multiaxial fatigue criteria, the ref-
results of repeated bending tests; X, repeated tension test. static pressure (Pmax), as indicated here: erence curve for the material being considered is
Table 2 presents fatigue test results that correspond to A to
D.
needed, just as it is when using the Goodman or
salt f (A, B Pmhp, C Palt ) Haigh diagram. Reference 16 shows that the use
of Crossland or Dang Van criteria takes the in-
where Pmhp is the mean hydrostatic pressure, salt crease in the bending fatigue strength into ac-
a, MPa
is the amplitude of octahedral shearing or am- count perfectly as a function of the residual
A plitude of the maximum shearing, and Palt is the stress introduced by the various treatment con-
B Repeated
C bending tests amplitude of hydrostatic pressure. ditions (Fig. 12).
400 D An example can be given as follows: When the multiaxial aspect is brought into the
Repeated tension test picture, the method that consists in introducing
300 salt A B PD
mhp C Palt
E
(Eq 3) residual stress into the calculation in the same
way as a mean stress, therefore, seems to give
200 where A, B, C, D, and E are material constants. satisfaction. The whole problem lies in defining
If Dsalt is taken on the maximum shearing the residual stresses to be included in the cal-
100 plane, D E l and B C, the result is the culation.
Dang Van criterion (Ref 32). Taking residual stress into account is essential
If Dsalt is taken on the octahedral shearing for correct prediction of the fatigue limit. Figure
200 100 0 100 200 300 400 plane, when D E 1 and B C, the Cross- 13 shows the important role played by compres-
Pmax, MPa land criterion results (Ref 33); when D E sive residual stress. If it is not taken into account,
1 and C 0, the Sines criterion (Ref 34); and the fatigue strength is underestimated (Fig. 13a).
Use of the Dang Van criterion to take residual
Fig. 12
stress into account (XC42 grade steel quenched
when D E 1 and B C, the Kakuno cri- If the residual stress measured or calculated is
after induction heating). A through D represent repeated terion (Ref 35). used without taking relaxation of the residual
bending test; X, repeated tension test. This type of development can be continued to stress into account, the fatigue strength is over-
Residual Stresses and Fatigue Behavior / 37

kt 2.5, g 2 mm1. However, with regard machining procedures. After correction of the with rather small and with compressive surface
to the residual stress state, there is no significant data points given in Fig. 32 to the same hardness, residual stresses yield identical fatigue behavior.
influence on the bending fatigue strength, even it turns out that the bending fatigue strength is It should be noted that the surface layer bearing
though the range of residual stresses covered hardly changed or slightly diminished at most, compressive residual stresses is rather small.
comes to more than 1000 MPa regarding speci- if the residual stresses change from compressive Contrarily, the generation of tensile residual
mens with kt 4.4, g 15 mm1. Careful to tensile (Ref 49). stresses causes a significant decrease of bending
inspection of the hardness of the specimens S-N curves for push-pull loading of smooth fatigue strength and a rather small decrease of
tested shows that a positive slope of the lines in specimens made from normalized SAE 1015 finite fatigue life, which obviously vanishes at
Fig. 32 is not related to the changing (macro) steel (German grade Ck 15) in the as-heat-treated high stress amplitudes.
residual stress state, but to different hardness of state and after an additional deep rolling are The alternating bending fatigue strength of
the specimen and, hence, differences in the mi- shown in Fig. 33(a) (Ref 53). Again, the fatigue ground smooth and notched specimen with dif-
cro residual stress state produced by different behavior in the range of finite fatigue life and the ferent geometries made of the same material
fatigue strength do not differ much. Figure 33(b) state is plotted in Fig. 36 as a function of the
shows a plot of S-N curves for rotating bending surface residual stresses (Ref 4752). Similar to
of the same material states. Now, by deep rolling Fig. 32, the influences of the stress-concentration
Nominal stress amplitude (n,a), MPa

320 finite fatigue life is increased by one order of factor and the normalized stress gradient are
300 magnitude or more, and the bending fatigue clearly discernible. However, contrary to the
280 strength is increased significantly. normalized steel there is distinct reduction of the
Medium-Strength Steel. Figure 34 shows S- bending fatigue strength with increasing tensile
260
Deep rolled N curves that were determined in alternating residual stress, this being more pronounced in
240 bending on notched specimens of quenched-and- the comparison of smooth specimens with
tempered (600 C/2 h) SAE 1045 steel (Ref 47 notched specimens. In the range of compressive
220
49). All data are valid for a failure probability residual stresses covered, the influence of resid-
200 of 50% and Nu 107. Three batches of speci- ual stress on bending fatigue strength is rather
mens were manufactured by grinding. The grind- small.
180 ing parameters (final feed, cutting speed) and the The S-N curves of smooth specimens in the
104 105 106 107
resulting depth distributions of residual stresses ground state and after an additional shot peening
Number of cycles to failure, Nf
are given in Fig. 35 (Ref 4749). Specimens are compared in Fig. 37(a) (Ref 4749). There
(a)
is a distinct increase of the bending fatigue
strength by shot peening, but a rather small in-
Nominal stress amplitude (n,a), MPa

320
1000
300 Deep rolled
Residual stress (l ), MPa

280

Nominal stress amplitude (n,a), MPa


2 20 m, 30 m/s 600
260
500
rs

240 2 6 m, 15 m/s 500 Shot peened

220 400
0
200 300
2 3 m, 15 m/s Ground
180 200
104 105 106 107 500
0 0.1 0.2 100
Number of cycles to failure, Nf
Distance from surface, mm
(b)
0
Fig. 35 Depth distribution of the residual stress in 104 105 106 107
Fig. 33 S-N curves of specimens made from normal- notched specimens made from quenched-and-
ized plain carbon SAE 1015 steel in the as- Number of cycles to failure, Nf
tempered (600 C/2 h) plain carbon SAE 1045 steel by dif-
heat-treated state and after an additional deep rolling for ferent grinding processes with the indicated two steps of (a)
(a) push-pull loading and (b) rotating bending. After Ref 53 final feed and cutting speed. Source: Ref 4749
Nominal stress amplitude (n,a), MPa

600

500
Bending fatigue strength (Rf), MPa
Nominal stress amplitude (n,a), MPa

600 400 Shot peened


Smooth = 0.4 400
500 Milled
300 = 0.1
kt = 1.7 300
400
Notched Ground
200
300 rs = 221 MPa
200 =2
100
200 =5
19 100 kt = 2.5 0
kt = 1.7 =2
100 602 104 105 106 107
=2
0 0 Number of cycles to failure, Nf
104 105 106 107 500 250 0 250 500 750 (b)
rs
Number of cycles to failure, Nf Surface residual stress (l ), MPa Alternating bending S-N curves of specimens
Fig. 37
made from quenched-and-tempered (600 C/2
Fig. 34 Alternating bending S-N curves of notched Fig. 36 Alternating bending fatigue strength of ground h) plain carbon SAE 1045 steel. (a) Smooth specimens after
specimens made from quenched-and-tem- smooth and notched specimens made from grinding and after additional shot peening. (b) Notched
pered (60 C/2 h) plain carbon SAE 1045 steel after differ- quenched-and-tempered plain carbon SAE 1045 steel ver- specimens after milling, grinding, and grinding with addi-
ent grinding processes. Source: Ref 4749 sus surface residual stress. Source: Ref 4752 tional shot peening. Source: Ref 4749
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 occurs between 2 and 5 million cycles. The tests actly show what the author is trying to demon-
model was presented using FEM to determine give an average fatigue life for the previously strate, since in the two types of tests for the same
the stabilized residual stress after fatigue load- mentioned loading in the order of 3.5 million material, the ratio of R(rmin /rmax) is modified.
ing. This estimation of the residual stress can cycles. This example shows that, if the cyclic As has been shown, the residual stress relaxation
then be used to calculate the fatigue life of a part, properties of the material are correctly known, it changes when the R ratio is varied (Ref 29). The
taking residual stress into account. Despite the is possible to predict the fatigue strength of the real contribution of each factor will therefore be
initial definition of the Dang Van criterion, material in the presence of residual stress. different from that indicated in Fig. 18.
which proposes that it only be used in cases of However, it should not be forgotten that other It can thus be seen that taking residual stress
fatigue strength with an unlimited number of cy- factors must also be taken into account in cal- into account in the calculation requires a serious
cles, an attempt was made to extend this criterion culating the fatigue lifethe introduction of re- examination of the different parameters in-
to include a limited fatigue life with a very large sidual stress is often accompanied by other volved. When reliable results are needed, fatigue
number of cycles (more than 2 106) (Ref 40). changes in parameters that have an influence on tests will no doubt need to be carried out on the
Figure 16 shows an example of the fatigue life the fatigue strength. In particular, these include: part or the structure concerned. However, mod-
estimated by calculation. First, a fatigue strength eling enables the variation in the different pa-
A new roughness that changes the local stress
diagram of the Dang Van type was defined ac- rameters to be rapidly simulated in order to find
concentration
cording to the fatigue life obtained from a series an optimal solution.
Additional strain hardening of the surface due
of fatigue life contours. Then the residual stress Partial Summary. The previous results show
to plasticizing
relaxation was calculated using FEM (Ref 29). that it is now possible to take residual stress into
A new metallurgical structure of the surface
Finally, the stabilized residual stress was intro- account in calculations designed to predict the
layers
duced into the diagram. In this example, in the fatigue life using a global approach. This must
case of a loading of 550 MPa, the point cor- Figure 17 shows the effect of the surface finish take the relaxation of residual fatigue stress into
responding to loading is inside the limit of the and strain hardening on the fatigue strength of account, as well as the other effects (strain hard-
fatigue life at 107 cycles. No failure occurs. For materials. It can be seen that an increase in the ening, hardness) introduced by the manufactur-
a loading of 600 MPa, the point corresponding roughness decreases the safety area, and strain ing method used. A multiaxial fatigue criterion
to loading including the residual stress is be- hardening increases the safety area, provided it that can integrate both the problem of residual
tween the line corresponding to 5 106 cycles does not damage the material. stress and the effect of the stress gradient applied
and that of 2 106 cycles. Failure therefore In the case of thermal or thermochemical sur- to a zone in the presence of stress concentration
face treatments (induction quenching, case hard- has been developed, that is, the Crossland or
ening, etc.), for example, it is necessary to take Dang Van criterion. It is used for a stabilized
alt the new fatigue strength of the treated layer into state of residual stress, averaged out for a basic
account in the calculation. volume of damage (thickness of critical layer),
1 2 Hardening effect The problem is more complex in the case of and applied within a network of contours that
2 1 3 Roughness increase residual stress introduced by plastic deformation represents the fatigue life. In the future, tests will
1 (pre-straining, machining, shot peening, roller be carried out to validate this type of criterion in
burnishing), since it is more difficult to distin- the case of combined stresses on notched parts
3
guish between the influence of residual stresses in the presence of residual stress.
and residual microstresses present in the grains
of the deformed material and that of strain hard-
ening of the material. Incorporating the Notion of
Evans (Ref 41) made this distinction in the Residual Stress into the Design
case of shot peening; to do so, he carried out
Pmax three types of fatigue tests on materials with
Office
various mechanical properties:
Fig. 17 Illustration of the effect of the surface finish and
Incorporating the notion of residual stress into
strain hardening on the fatigue strength Fatigue tests on a non-shot-peened material
the design office must be gradual and can be di-
Fatigue tests on a shot-peened material
vided up into several phases.
Fatigue tests on a shot-peened material, but
Today, very few industrial sectors consider
40 with a mean test stress rm, which compen-
the residual stress parameter directly. In techni-
sated for the surface residual stress. In this
cal specifications, requirements are included that
Role of hardening type of test, the effect of the macroscopic re-
are often closely related to residual stress with-
30 and residual stress sidual stress is cancelled out and the fatigue
out actually naming it. An Almen intensity must
Fatigue life increasing, %

strength obtained only depends on the in-


be guaranteed in the case of shot peening, for
crease in the mechanical properties of the
20 example, a roller-burnishing load, a machining
plastically deformed material and the residual
Role of procedure or a minimum treated thickness in the
microstresses distributed throughout the ma-
residual stress case of thermal or thermochemical treatment,
terial. The results obtained are presented in
10 and a maximum dimensioning tolerance in the
Fig. 18.
case of a machined or welded part.
It can be observed that, for materials with low In the first phase of incorporation, a semi-
0 resistance, the increase in the fatigue strength is quantitative notion can be used to evaluate the
Role of superficial hardening mainly due to surface strain hardening. On the increase in performance in terms of fatigue life
other hand, for highly resistant materials, it is or fatigue strength. A few examples can be pre-
10
20 30 40 50 mainly the influence of the residual stress that sented. Table 3 gives an example of the effec-
HRC governs the fatigue strength. When materials tiveness of shot peening in increasing the fatigue
have low elastic limits, the stresses introduced life of different types of mechanical parts, and
Fig. 18 Effect of the resistance of the base metal on the
increase in the fatigue strength after shot-
by shot peening relax much more easily than Fig. 19 shows the beneficial role played by roller
peening treatment, distinguishing between the effect of they do when the elastic limit is high. This test burnishing on the fatigue strength of spherical
strain hardening and that of residual stress only shows a general tendency and does not ex- graphite cast iron crankshafts. Figure 20 shows
20 / Effect of Materials and Processing

a horizontal comparison of gains to be expected According to a mechanical approach, we can For a cyclic softening material, the residual
in terms of fatigue strength from various surface make the following general predictions (Ref 30): stresses relax with the increase of the number
treatments. The results presented here are not at of cycles up to a stabilized state of cyclic
For a cyclic hardening material, the relaxation
all exhaustive and are taken from a limited bib- properties.
of residual stresses is realized in the first se-
liography. However, this figure should not be
ries of fatigue cycles. The cyclic behaviors of material are very im-
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 im-
portant effect on the gain achieved. Each indus- Table 3 Increase in the fatigue life of various mechanical components as a result of shot
trial sector must carry out this type of compari- peening
son for the treatments and materials used in order
Type of part Type of stress Increase in the fatigue life (%)
to help engineers design their products more ef-
fectively. Spindles Reverse bending 4001900
Shafts Torsional 700
The second phase consists of predicting the Gear box Fatigue life tests in service 80
fatigue life using the notions developed in the Crankshafts Fatigue life tests in service 3000, but highly variable
previous paragraph. Aircraft coupling rods Tensile compression 105
The third phase is the development of inte- Driving rods Tensile compression 45
Cam springs Dynamic stress 100340
grated tools for taking the residual stress into Helical springs Fatigue life in service 3500
account. The following paragraph presents an Torque rods Dynamic stress 140600
example of such a design tool on fatigue for Universal joint shaft Reverse bending 350
three-dimensional components. Gear wheel Fatigue life tests 130
Tank chain Fatigue life tests 1100
Weld Fatigue life tests 200
Valve Fatigue life tests 700
Example of Integrated Design Tool Rocker arm Fatigue life tests 320

By using the cyclic behaviors of material, a


simplified method to calculate residual stress re- 250
laxation has been proposed in the first section.
Maximum admissible stress, MPa

Secondly, a method to predict fatigue life by tak-


200
ing the stabilized residual stresses into account
Alternated
is presented. This design tool is based on the stress
FEM. It has been applied to shot-peened 150
35NCD16 grade steel. The different fatigue pa- 0
rameters often used in material research are stud- 100
ied. On the other hand, an experimental inves-
tigation about this material had been done by
Bignonnet (Ref 42). The results of the study 50 0
show that this design tool on fatigue developed
by the Laboratory of Mechanical Systems En- 0 Repeated
stress
G

gineering (LASMIS) (Ref 43) is able to take into


S-

S-

S-

S-

S-

S-
70

40

40

70

70

70

account different loading parameters.


w

io

te

sh

in

su
ith

ni

du
ni

The residual stresses, however, can be relaxed


ot

rfa
c

fe
ou

ct
pe

ce
ni

io
tt

by the deliberate application of thermal or me-


tri

en

ro
re

di

in

llin
at

ng

chanical energy. They will be especially relaxed


g
m

g
en

when the structure is subject to cyclic loading.


t

The relaxation phenomenon depends on a com- Fig. 19 Effectiveness of roller burnishing in increasing the fatigue strength
plex interaction of a number of factors, such as
the applied stress amplitude, the number of cy-
cles of loading, the state of initial residual
stresses, and the nature, origin, and mechanical Thermal treatment by electron beam, R = 0.1
properties of the material. In this article, only
TT by CO2 laser, R = 0.1
relaxation during cyclic loading and the influ- TT by CO2 laser, R = 1
Treatment or process

ence of the stable residual stresses on the fatigue TT by laser shock, R = 0.1
life are studied.
ST by nitriding R = 0.1
In only a few cases are the residual stresses
systematically analyzed using measurement of TT by induction R = 0.1
the residual stress state during and after fatigue TS carburizing R = 0.1
testing. This is usually a difficult, time-consum-
ST by surface rolling R = 0.1
ing task. Several fatigue tests under tensile and
torsion loading with different stress amplitudes 200%
ST by surface rolling R = 0.1
have been done.
A numerical method for the prediction of the ST by shot peening R = 0.1
residual stress distribution during and after fa-
0 25 50 75
tigue has been developed. Finite element soft-
Increase, %
ware is used for incorporating cyclic plasticity
into the calculation. A simplified method to cal- Beneficial effect of various surface treatments on the fatigue strength (maximum gain reported in the literature).
culate stabilized residual stresses was proposed.
Fig. 20
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 relax- (Ref 43). It can take the residual stresses into can be obtained by the measurement method or
ation. They make it possible to calculate the re- account when calculating the fatigue life of a by the simulation of surface treatment. Then, this
sidual stress state for a corresponded number of structure. It is based on the FEM. With this tool, initial field of residual stresses can be introduced
cycles. an iso-colored image of distribution of fatigue by FEM using ABAQUS (Ref 44, 45). Further-
Then, a method to predict fatigue life while life or safety factor can be obtained. more, a stabilized field of residual stresses can
taking these stabilized residual stresses into ac- These methods mentioned above are applied be predicted after an elastic-plastic calculation,
count is presented. The fatigue criterion under to shot-peened 35NCD16 grade steel. The dif- in which the cyclic behaviors of material are
multiaxial stresses is used. After the relaxation, ferent fatigue parameters often used in material used. This stabilized field of residual stresses af-
the residual stresses can be combined directly research are studied. A comparison with the ex- ter relaxation must be taken into account when
with the cyclic loading. They change the value perimental results has been carried out (Ref 43). calculating the fatigue life. At last, with the help
of the mean stresses and influence the fatigue Basic Steps for the Calculation. Figure 21 of a design tool on fatigue, FATIGUE3D, the
life. shows the elementary steps of the calculation of distribution of fatigue life and that of the safety
As a result, a design tool on fatigue for three- fatigue life while taking into account the residual factor on fatigue can be obtained.
dimensional components, FATIGUE3D, has stresses. Principle and Basic Function. According to
been developed in the laboratory of LASMIS First, the initial state of the residual stresses the criteria of fatigue under multiaxial loading,
the fatigue life changes along with two param-
eters. One of them is maximum static pressure;
another is equivalent octahedral stress. There-
Calculation of Introduction of initial field Measurement of fore, two stress-number of cycles (S-N) curves
residual stresses of residual stresses residual stresses 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
Prediction of Cyclic behaviors
(life) in the stress plane P s (Fig. 22).
residual stress of material
A different fatigue life is represented by a dif-
ferent line. If there is a complex stress state, it
Outside solicitation, Mechanical properties gives a point in the plane P s. Respectively,
Fatigue, 3D the line to which this point is nearest represents
cyclic loading of material
the fatigue life of that stress state. In this way,
the fatigue life for any complex stress state and
Distribution of fatigue life, for any part of a component can be predicted. A
safety factor, and distribution of the fatigue life is helpful for the
admissible residual stress designer.
Fig. 21 Basic steps of calculation while taking into account the residual stresses
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 de-
signed life and M is the point given by the stress
S state. The safety factor on fatigue is defined as S
OM/OM.

N1
Example and Results

N1 Material and Geometry. The shot-peening


process is a widely used technique because it can
produce a field of residual stress on the surface
N P of a part. These stresses are compressive and will
Fig. 22 Principle of calculating fatigue life improve the properties on fatigue. As an exam-
ple, a shot-peened part made of 35NCD16 grade
steel (0.35% carbon, 3.7% chromium, 0.3% mo-
lybdenum 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 ex-
s = OM/OM periments are available regarding tension com-
pression and torsion loading (Ref 44).
The geometry of the part is defined from the
M real geometry so that the testing results can be
compared. It is similar to the fatigue test sample,
N-cycles only the numerical values (radius, length) vary
for tension compression or torsion loading. The
M difficulty is to define a thin mesh near the surface
(0.20.4 mm, or 0.0080.016 in.) where the re-
P sidual stresses are introduced while maintaining
O
a normal mesh in other zones. Figure 24 shows
Principle of calculating the safety factor on fa- the three-dimensional mesh of the part.
Fig. 23 Fig. 24 Introduction of the Initial Field of Residual
tigue Three-dimensional mesh
22 / Effect of Materials and Processing

100 100
0 Axial stresses Tangential stresses
Residual stresses, MPa
under tensile loading 0 under tensile loading

Residual stresses, MPa


100 Measurement Measurement
100
Calculated Calculated
200
200
300
300
400
400
500
500
600
600
0 0.05 0.10 0.15 0.20 0.25 0.30 0 0.05 0.10 0.15 0.20 0.25 0.30
Depth, mm Depth, mm

100 100
0 Axial stresses 0 Tangential stresses
under torsion loading
Residual stresses, MPa

Residual stresses, MPa


under torsion loading
100 Measurement 100 Measurement
Calculated Calculated
200 200
300 300
400 400
500
500
600
600
700
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Depth, mm Depth, mm
Fig. 25 Initial distribution of the residual stresses

Stress. The initial residual stress distribution is elastic-plastic properties of material, because the der the cyclic loading can be obtained. Part of
calculated from the residual stresses measured plastic deformation is the main cause of the re- the results and their comparison with the test val-
by the x-ray diffraction method in the upper laxation. So, the cyclic behaviors of material are ues have been shown in Fig. 27. In the case of
layer where initial stresses are introduced by sur- very important for the prediction of the residual the traction-compression loading, two loading
face treatment. In depth, the stresses are calcu- stresses. However, they can be measured by ex- levels were calculated. The same level for tor-
lated for the structural equilibrium. Figure 25 periment. sion loading with two different cycle numbers
shows the comparison of initial residual stresses Figure 26 shows the cyclic behaviors of was also analyzed. Figure 28 shows the stabi-
between the calculation and the experiment. 35NCD16 grade steel, a cyclic softening mate- lized residual stresses for different traction-com-
Prediction of the Residual Stress Relaxa- rial. pression load levels that are used in the predic-
tion. In order to predict the relaxation of the re- When the stabilized residual stresses are cal- tion of the fatigue life. Unfortunately, it is
sidual stresses, a simplified method proposed by culated, the behavior corresponding to NR /2 is impossible to compare with the experiment re-
Lu et al. has been used (Ref 30). It supposes that used. After an elastic-plastic calculation with sults.
the relaxation of the residual stresses depends on FEM, the relaxation of the residual stresses un- Prediction of the Fatigue Life. For predict-
ing the fatigue life under a complex stress state,
it is necessary to have two S-N curves. One is
Rec(MPa) = 920 55,14 log N
for simple traction or traction compression; an-
other is for alternated torsion. These curves are
1 cycle basic data of a material, and they can be obtained
2 N, Re(Y.S.), by experimental method.
1000
3 cycles MPa
15 Using the program FATIGUE3D, a design
900 50 1 940 tool on fatigue developed in the LASMIS labo-
/2, MPa

150 ratory, a distribution of fatigue life and a distri-


2 920
800
NR/2 3 910 bution of safety factor for a structure can be ob-
15 870 tained. In this example, the fatigue life under
50 830 traction-compression cyclic loading and torsion
700
150 800 loading has been calculated. The stabilized re-
35 NCD 16 sidual stresses play a role as static load. Figure
600 (Rm (UTS) = 110 MPa) NR/2 660
29 shows the results of fatigue life and their com-
parison with the experimental results. It is very
0.5 0.75 1 clear that the proposed method is available.
t/2, %
Prediction of Admissible Residual Stress.
In this approach, a calculation method has been
The cyclic behavior of 35NCD16 grade steel. Re, yield stress; R ec, cyclic yield stress as a function of number developed that can predict the admissible resid-
Fig. 26 of cycles; Det, axial plastic strain range during cyclic loading ual 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 102 cycles, traction compression 700 MPa
0 0
Axial measurement Axial measurement
Residual stresses, MPa

Residual stresses, MPa


Axial calculated Axial calculated
100 Tangential measurement 100 Tangential measurement
Tangential calculated Tangential calculated

200 200

300 300

400 400

500 500
0 0.05 0.10 0.15 0.20 0.25 0.30 0 0.05 0.10 0.15 0.20 0.25 0.30
Depth, mm Depth, mm

100 200
105 cycles, torsion 380 MPa 102 cycles, torsion 380 MPa
100
0 Axial measurement Axial measurement
Axial calculated Axial calculated
0
Residual stresses, MPa

Residual stresses, MPa


Tangential measurement Tangential measurement
100 Tangential calculated Tangential calculated
100

200 200

300
300
400
400
500

500 600
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Depth, mm Depth, mm
Fig. 27 The relaxation results of the residual stresses

be used during the design phase to evaluate life. In other zones, tensile residual stress is ad- ing under different conditions, the physical and
whether or not residual compressive stress needs missible. mechanical data of the material need to be
to be introduced. The computer code also indi- The fatigue life can then be used to deduce a known. Generally speaking, these parameters
cates the level of the residual stress and the zone stabilized residual stress. An example of defining depend on the temperature, but the surface heat-
in which it needs to be introduced. In this way, the quenching conditions according to the resid- transfer coefficient depends not only on the tem-
the treatment condition during the mechanical ual stress field obtained before relaxation is an- perature, but also the sample geometry, the quen-
design phase can be defined. Figure 30 shows a alyzed. chant, and the quenching temperature. In order
map of the prestress zone indicated for a notch An inverse technique designed to obtain the to solve the problem, it is considered that there
sample. It can be seen that in the high stress con- quenching conditions is then developed (Ref 46, is no change in any of the quenching parameters
centration area, residual compressive stress must 47). In this technique, it is necessary to simulate or the sample geometry. Only the quenching
be introduced to obtain a predetermined fatigue quenching using the FEM. To simulate quench- temperature changes. Quenching at different
temperatures was carried out. This gives the vari-
ation of the surface heat-transfer coefficient as a
100 100
function of the temperature for each quenching
temperature. It should be mentioned that numer-
Tangential residual stress, MPa

Traction-compression Traction-compression
Axial residual stresses, MPa

550 MPa 0 550 MPa ical methods alone are not sufficient to obtain
0 600 MPa 600 MPa the heat-transfer coefficient. The temperature
620 MPa 100 620 MPa must be measured during quenching. Since the
650 MPa 650 MPa heat-transfer-coefficient curve is known, the pa-
100 700 MPa 700 MPa
200 rameters of these curves as a function of the
quenching temperature can be defined. Once this
200 300 coefficient is known, quenching can be modeled
at different temperatures. The quenching tem-
400 perature in the case of water quenching varies
300
between 20 and 80 C (68 and 176 F). The lower
500 the quenching temperature, the higher the level
0 0.05 0.10 0.15 0.20 0.25 0.30 0 0.05 0.10 0.15 0.20 0.25 0.30
of the residual stress field introduced. Initially,
Depth, mm Depth, mm
the residual stress field obtained was found to be
Fig. 28 The stabilized residual stresses under different traction-compression loading cases lower than at 20 C (68 F) and higher than at
24 / Effect of Materials and Processing

750 80 C (176 F). After this condition has been


Traction-compressionn load, MPa

440 verified, the quenching temperature can be de-


Without residual stresses Without residual stresses
700 termined by varying the quenching temperature
Experimental 420 Experimental

Torsion load, MPa


Numerical Numerical during simulation. After each simulation, the
650 400
calculated residual stress field is compared to
380 that obtained from the fatigue life. A limit con-
600
360 dition can be defined to improve the precision of
550 the results. When comparing the residual stress
340
fields, the hydrostatic pressure was used at each
500 320 point.
450 300
105 106 107 105 106 107
Fatigue life N cycles Fatigue life N cycles
Summary
Fig. 29 Results of fatigue life
Residual stress plays a very important role
with respect to the different properties of mate-
rials. The gain obtained from the presence of re-
sidual stress can be enormous. This article at-
tempts to show the effects of residual stress
through the example of fatigue strength. Here, it
S11 Value has been shown that it is now possible to predict
2.24E+02 the fatigue strength of materials, taking residual
4.24E+01 stress into account. The results of this study
+1.39E+02 show that it is possible to predict the residual
+3.20E+02 stress relaxation and fatigue life, with consider-
+5.01E+02 ation of the influence of residual stress by the
+6.82E+02 FEM. It has been found that the calculated re-
sults of fatigue life at surface agree very well
with experimental results.
Although the author is not in a position to pro-
vide the same type of calculation tools for other
properties, such as corrosion resistance and the
1
adhesion of coatings, it is now reasonable to ex-
3 pect that the notion of residual stress will be
gradually introduced into the design stage of me-
2 chanical parts. Numerical modeling of the be-
havior beforehand saves a considerable amount
Fig. 30 Prediction of admissible residual stress (or the prestressed area) for a notched sample under bending loading, of time because of the reduction in the number
using the finite element method
of experimental tests required. These tests are
often very long and costly, but they have proved
to be indispensable. The problem of taking re-
sidual stress into account at the design stage will
become more and more critical with the devel-
opment of new materials (multimaterials, etc.)
and new treatments (combined treatments, etc.)
Mechanical component With the development of different experimen-
development tal and numerical techniques, it is now possible
to introduce residual stress into the design office
for the integrated design of mechanical compo-
Introduction of a surface nents, thus offering a new concurrent engineer-
Optimization of geometry and prestress treatment
according to the choice ing approach applied to the design of mechanical
of manufacturing process components taking residual and applied stress
into consideration. Figure 31 shows the different
No Is fatigue life connections between residual stress-integrated
in accordance Yes
Stress analysis by Stop design and other sectors that use the concurrent
a finite element method with
specifications?
engineering approach. A mechanical component
Adjustment of a designer can simulate dynamic characteristics,
finite element model material processing, and product life.
Experimental checking of Consideration of residual stress is becoming
stress: rapid prototyping
increasingly important for two reasons: the in-
troduction of multimaterials that induce residual
stress, and the need for the designer to reduce
No Are results Yes Fatigue life the weight of components in order to remain
coherent? estimation competitive. Basic research has brought a better
understanding of the phenomena relating to re-
Various connections between residual stress-integrated design and other sectors that use the concurrent en- sidual stress.
Fig. 31 The main aim is to develop an integrated qual-
gineering approach
Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 25

ity control tool. For industrial applications, fu- grant Brite-EuRam, BRRT-CT985090, EN- Austenitic Stainless Steel, Residual Stress,
ture developments are necessary: SPED project), and the National Science Foun- V. Hauk, H.P. Hougardy, E. Macherauch,
dation of China (two bases project) is acknowl- and H.D. Tietz, Ed., DGM, Verlag, 1992, p
Measurement techniques edged. 891900
Improvement of ultrasonic and magnetic 12. J.B. Roelens, F. Maltrud, and J. Lu, Deter-
measurement methods mination of Residual Stresses in Submerged
Integration of the portable optical method for REFERENCES Arc Multi-Pass Welds by Means of Numer-
strain measurement in destructive techniques ical Simulation and Comparison with Ex-
Processing and materials perimental Measurements, Welding in the
1. J. Lu, Ed., Handbook of Measurement of Re- World, Vol 33 (No. 3), 1994, p 152159
Development for the industrialization of new sidual Stresses, Society for Experimental 13. Proceedings of the National Conference on
prestress processes: ultrasonic shot peening, Mechanics, USA, 1996 the Residual Stress and Design Office, CE-
laser shock, surface nanocrystallization, and 2. J. Lu, A. Niku-Lari, and J.F. Flavenot, Lat- TIM, (Senlis), 1991
so on. est Development of Residual Stresses Mea- 14. D. Deslaef, E. Rouhaud, and J. Lu, Finite
Optimization of residual stress in advanced surement by the Hole Drilling Method, Ma- Element Modeling of Shot-Peening Pro-
materials: functional gradient coating system, teriaux et Techniques, Dec 1985, p 709718 cesses, Proceedings of the SEM Spring Con-
for example, plasma spray coating, PVD, 3. Z. Wu, J. Lu, and B. Han, Study of Residual
ference on Experimental and Applied Me-
CVD, or diamond-like carbon (DLC) coating; Stress Distribution by a Combined Method
chanics, 13 June 1998, (Houston, Texas),
metal-matrix composite; electronic packag- of Moire Interferometry and Incremental
p 419422
ing; intermetallic materials Hole Drilling. Part I: Theory, J. Appl.
Mech., ASME, Sept 1998 15. A. Chabenat, and J.F. Flavenot, Measure-
Modeling 4. Z. Wu, J. Lu, and B. Han, Study of Residual ment of Residual Stress on the Quenched
Axles after Heating by Induction, Traite-
Development of advanced tools for the pre- Stress Distribution by a Combined Method
of Moire Interferometry and Incremental ment thermique/Revue de metallurgie,
diction of initial residual stress in various pro-
Hole Drilling. Part II: Implementation of (Heat treatment/Journal of Metallurgy), No.
cesses: forging, casting, machining, welding,
plasma spray coating, composite processing, Applied Mechanics, J. Appl. Mech., ASME, 124, 1978, p 233239 (in French)
electronic packaging manufacturing, mechan- Sept 1998 16. N. Skalli, and J.F. Flavenot, Prediction of
ical surface treatment 5. Z. Wu and J. Lu, Residual Stresses by Moire Fatigue Life by Simulation with the Resid-
Integration of the residual stress modeling Interferometry and Incremental Hole Drill- ual Stress Considerations, Conference pro-
method into a global life design (3-D) and the ing, Experimental Mechanics, Advances in ceedings, Spring Conference, SF2M
creation of optimization codes for the design Design, Testing and Analysis, Vol 2, I.M. (French Society for Metal and Materials),
of mechanical components using the concur- Allison, Ed., Conf. ICEM 98, 2428 Aug 2223 May 1984, p 98117 (in French)
rent engineering approach 1998, (Oxford), p 13191324 17. H.P. Lieurade, P. Castelucci, J.F. Flavenot,
6. Z. Wu, X.K. Niu, J. Lu, and P.G. Ifju, Study and J. Lu, Efficiency of Improvement Tech-
With the recent progress in the field of re- of Process-Induced Stress in Orthotropic niques on the Fatigue Strength as a Function
search and development, it is now possible to Composite LaminateCarbon/Epoxy [O2 / of the Type of Welded Joint, Welding in the
introduce a new area of research: prestress en- 902]2s, Proceedings of the SEM Spring World, Vol 31 (No. 4), 1991, p 268271 (in
gineering of structural materials (PESM or Conference on Experimental and Applied French)
PRESMA). In the near future, the mechanical Mechanics, 13 June 1998, (Houston, 18. H.P. Lieurade, C. Maillard-Salin, and M.
engineer will be able to use these tools in a pre- Texas), p 179181 Truchon, Fatigue Cracking of Welded Joints
stressed concrete approach to civil engineering. 7. Z. Wu, J. Lu, and Y. Guo, A New Method of High Strength Steel, Proceedings of
The ENSPED program in Europe is an excellent of Residual Stress Measurement on Elec- IABSE, Colloquium, Lausanne, 1982, p
example of the cooperation of industrial and re- tronic Packaging, Experimental Mechanics,
search partners. Using this new approach, the fa- 137144 (in French)
Advances in Design, Testing and Analysis, 19. D. Francois, Role of the Residual Stress on
tigue behavior and mechanical behavior of new Vol 2, I.M. Allison, Ed., Conf. ICEM 98,
materials can be considerably improved, increas- the Fracture, Conference Proceedings, Re-
2428 Aug 1998, (Oxford), p 987991 sidual Stress in Welding Structure, CETIM,
ing two- or threefold in some cases. In fact, many 8. J. Lu, J.F. Flavenot, and S. Thery, Study on
failures occur in the area close to the surface. SENLIS, Dec 1987, p 8797 (in French)
the Effect of the Finishing Treatment on the 20. W.H. Friske, Shot Peening to Prevent the
Improving the mechanical behavior is often only Residual Stress Gradient in SiC Reinforced
necessary on the surface (fatigue, fretting fa- Corrosion of Austenitic Stainless Steels,
Aluminum Metal Matrix Composite, J. Rockwell International Cooperation, 1975
tigue, wear, stress corrosion). As a result, this Compos. Technol. Res., ASTM, Vol 12 (No.
type of approach can be expected to develop 21. A. Niku-Lan`, M. Menier, and G. Beranger,
4), 1990, p 232238
very quickly in order to introduce adequate pre- Effect of the Surface State on the Stress
9. J. Li, J. Lu, M. Perrin, M. Ceretti, and A.
stressing treatment in the area indicated by the Corrosion Behaviour of a Stainless Steel,
Lodini, Study on the Residual Stress in
advanced design tool. Cold-Rolled 7075 Al-SiC Whisker Rein- CETIM Informations, No. 78, Dec 1982
forced Composites by X-ray and Neutron 22. C. Richard, J. Lu, J.F. Flavenot, and G. Be-
Diffraction, J. Compos. Technol. Res., July ranger, Study of Residual Stress in the
ACKNOWLEDGMENTS
1995, p 194198 Plasma Sprayed NiCrAlY Coating and
The author is grateful to all the people at LAS- 10. J. Lu, P. Peyre, C. Omam Nonga, A. Ben- Characterization of Interface Toughness,
MIS working in the field of residual stress for amar, and J.F. Flavenot, Mechanical Surface Memoires et Etudes Scientifiques, Revue de
their help during the preparation of this paper: Treatments, Current Trends, and Future metallurgie, May 1991, p 295306 (in
Prof. J.L. Chaboche, Dr. D. Retraint, Dr. E. Rou- Prospects, Surface Modification Technolo- French)
haud, Dr. X.L. Gong, Dr. L. Couturier, Dr. R. gies VIII, T. S. Sudarshan and M. Jeandin, 23. C.S. Richard, G. Beranger, J. Lu, and J.F.
Akrache, Dr. Z. Wu, Dr. S. Rasouli Yazdi, Mr. Ed., TMS, Sept 1994, p 589601 Flavenot, The Influence of Heat Treatments
A. Milley, Mr. D. Deslaf, Mr. B. Guelorget, Mr. 11. A. Benamar, J. Lu, J.F. Flavenot, P. Bar- on the Adhesion of Plasma Sprayed Ni-
A. Voinier, Mr. F. Belahcene, Mr. K. Gong. Fi- barin, G. Chalant, and G. Inglebert, Mod- CrAlY Coatings, Surf. Coat. Technol., Vol
nancial support from the European Union (under eling Residual Stresses by Cold Rolling an 82, 1996, p 99109
26 / Effect of Materials and Processing

24. C.S. Richard, G. Beranger, J. Lu, J.F. Flav- letaud, and T.C. Lindley, Ed., Mechanical Multiaxial Fatigue Criteria into Account,
enot, and T. Gregoire, Four-Point Bending Engineering Publications, London, 1996, p Proceedings of 2nd International Confer-
Tests of Thermally Produced WC-Co Coat- 477487 ence on Residual Stress, (Nancy, France),
ing, Surface and Coatings Technology, (No. 32. K. Dang Van, On the Fatigue of Metals, Sci- Elsevier Applied Science, Beck et al., Ed.,
78), 1996, p 284294 ence Technique Armement, Memorial de Nov 1988, p 784790
25. C.S. Richard, J. Lu, G. Beranger, and F. De- lartillerie francaise, 3ie`me Fascicule, 1973, 41. W.P. Evans, R.E. Ricklefs, and J.F. Millian,
comps, Study of Cr2O3 Coating Materials, p 647722 (in French) Local Atomic Arrangements Studied by X-
Part II: Adhesion to a Cast Iron Substrate, 33. B. Crossland, Effect of Large Hydrostatic ray Diffraction, J.B. Cohen and J.E. Hillard,
J. Therm. Spray Technol., ASM Interna- Pressure on the Torsional Fatigue Strength Ed., Proc. of a Symposium, (Chicago), Feb
tional, Dec 1995, Vol 5 (No. 4), p 347352 of an Alloy Steel, Int. Conf. on Fatigue of 1965, Chap. 11
26. R.J. Arsenault, Tensile and Compressive Metals, IME/ASME, (London), 1956, p 42. A. Bignonnet, Fatigue Behavior of a Shot
Properties of Metal Matrix Composites, 138149 Peened 35NCD16 Grade Steel, Evolution of
Metal Matrix Composites: Mechanisms and 34. G. Sines, Fatigue Criteria under Combined Residual Stress Due to the Shot Peening as
Properties, R.K. Evertt and R.J. Arsenault, Stress or Strains, Transactions ASME, J. a Function of the Type of the Loading, Pro-
Ed., p 133167 Eng. Mater. Technol., Vol 13, April 1981, ceedings of Ninth Conference on Shot Peen-
27. P. Balladon, Stress Relieve After Welding, p 8290 ing, CETIM, Senlis, France, Nov 1985, p
Criterion and Process, Conference Proceed- 35. H. Kakuno and Y. Kawada, A New Crite- 107116 (in French)
ings, Residual Stress in Welding Structure, rion of Fatigue Strength of a Round Bar 43. FATIGUE3D, User Manual, UTT-LAS-
CETIM, Senlis, Dec 1987, p 4569 (in Subjected to Combined Static and Repeated MIS, Jan 2000
French) Bending and Torsion, Fatigue of Engineer- 44. ABAQUS Theory Manual, Hibitt, Karlsson,
28. R. Lematre, J.L. Lebrun, and J. Maeder, ing Materials and Structures, Vol 2, 1979, and Sorensen, Inc., 1996
Residual Stress and Fatigue, Materiaux et p 229236 45. A. Milley, E. Rouhaud, and J. Lu, Prediction
Techniques, Sept 1982 (in French) 36. I.V. Papadopoulos, Cyclic Fatigue of of Residual Stress Relaxation in Three-Di-
29. J. Lu, J.F. Flavenot, A. Turbat, and D. Fran- Metal, a New Approach, Ph.D. disserta- mensional Structures under Tensile and
cois, Modeling of the Relaxation of Resid- tion, Ecole Nationale des Ponts et Chaus- Torsional Loadings, Fifth International
ual Stress under Fatigue Loading, Confer- se`es, France, Dec 1987 (in French) Conference on Residual Stress, Vol 1, T. Er-
ence proceedings, Residual Stress at Design 37. A. Galtier and J. Seguret, Multiaxial Fatigue icsson, Ed., (Linkoping), 1618 June 1997,
Office, CETIM, Senlis, Nov 1991, p 7198 Criterion, Revue Francaise de Mecanique, p 424429
(in French) No. 4, 1990, p 291299 (in French) 46. S. Rasouli Yazdi and J. Lu, Simulation of
30. J. Lu, J.F. Flavenot, and A. Turbat, Predic- 38. A. Deperrois, On the Fatigue Limit Cal- Quenching and Fatigue Relaxation of Re-
tion of the Residual Stress Relaxation Dur- culation, Ph.D. dissertation, Ecole Poly- sidual Stresses in Aluminium Parts, Fifth In-
ing Fatigue, Mechanical Relaxation of Re- technique, France, June 1991 (in French) ternational Conference on Residual Stress,
sidual Stresses, ASTM STP993, 1998, p 39. R. Akrache and J. Lu, 3D Calculation of Vol 1, T. Ericsson, Ed., (Linkoping), 1618
7590 High Cycle Fatigue Life under Out-of-Phase June 1997, p 490495
31. J. Lu, J.F. Flavenot, A. Diboine, S. Lasserre, Multiaxial Loading, Fatigue and Fracture 47. S. Rasouli Yazdi and J. Lu, Inverse Method
C. Froustey, M. Bennebach, and T. Palin- of Engineering Materials and Structures, for Determining the Quenching Conditions
Luc, Development of a General Multiaxial Vol 22, July 1999, p 527534 Using a Residual Stress Field, Proceedings
Fatigue Criterion for High Cycles of Fatigue 40. J. Lu and J.F. Flavenot, Prediction of the of the Third Int. Conf. on Quenching Con-
Behavior Prediction, Multiaxial Fatigue Residual Stress Relaxation During Fatigue trol and Distorsion, ASM, (Prague), March
and Design, ESIS 21, A. Pineau, G. Cail- Loading and Taking the Residual Stress in 1999
Handbook of Residual Stress and Deformation of Steel Copyright 2002 ASM International
G. Totten, M. Howes, T. Inoue, editors, p27-53 All rights reserved.
DOI: 10.1361/hrsd2002p027 www.asminternational.org

Residual Stresses and Fatigue Behavior


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

COMPONENTS MAY FAIL DURING SER- mechanisms of residual stress formation are in- with work-hardening and/or work-softening
VICE in very different ways. Important aspects homogeneous plastic deformation caused by me- processes occur. In both cases, the amount of
are: chanical, thermomechanical, or thermal attack cyclic plastic deformation is the most important
and the development of constraint between dif- parameter. Since during fatigue loadings that re-
Onset of plastic deformation or exceeding of ferent constituents of a component such as sult in technically relevant lifetimes or in infinite
allowable plastic deformation phases or coatings during the formation of the life, the cyclic plastic deformations decrease
Crack initiation or exceeding of allowable constituents (e.g., during the coating process) with increasing hardness, it is generally expected
crack length and/or during cooling from a higher process that the influence of macro residual stresses is
Unstable crack propagation that results in temperature (e.g., the sintering temperature). lower in low-strength steels than in high-strength
fracture of the component Manufacturing processes not only result in the steels.
Instability such as buckling, which may result formation of definite macro residual stress states, Comparison between Loading Stresses and
in elastic or plastic collapse and, hence, cat- but may also effect other changes of the com- Residual Stresses. Since the alteration of the
astrophic failure ponent state close to its surface: micro residual stress state caused by manufac-
In practice, failure of a component is almost Formation of a specific topography turing processes changes the local hardness
always determined by the interaction of several Work-hardening or (in very hard material within a component and hence the local strength,
external parameters, such as unexpected high states) work-softening processes and hence it is convenient to comprise the corresponding
static and/or cyclic loads, friction, thermal en- alteration of the micro residual stress state influence on the fatigue behavior of the compo-
ergy, oxidation, corrosion, and so forth. On the Phase transformation nent as a change of the local fatigue resistance
other hand, the component itself may contribute Crack initiation of the steel used for the component.
to premature failure, if there is confusion of the On the other hand, stable macro residual
component material, wrong heat treatment, sur- All these processes may change the fatigue be- stresses are static stresses and may be regarded
face decarburization, and so forth. havior of a given component. It is hardly possi- as locally varying mean stresses. However, it is
The residual stress state in a component may ble to completely separate these influences, even important to realize that there are definite differ-
be one of the most important parameters influ- though this is very important for the understand- ences between loading mean stresses and resid-
encing its service behavior, particularly regard- ing of the existing relationships. ual stresses:
ing high-strength material states. Residual
stresses may reduce yield or elastic collapse Loading mean stresses exist as a consequence
loads and may promote corrosion cracking (Ref Residual Stress of acting external forces, moments, pressure
1, 2). With respect to fatigue, which is the topic and/or internal pressure, and eventually tem-
of this article, residual stresses may alter the cy- Stability of Residual Stresses. For the as- perature gradients. Residual stresses exist as
clic deformation behavior, promote or retard sessment of the influence of residual stresses on a consequence of inhomogeneous plastic de-
crack initiation, accelerate or decelerate crack fatigue behavior, the stability of the residual formation and/or as a consequence of con-
propagation, and may be beneficial or detrimen- stress state is of utmost importance. The reader straints between different constituents of a
tal to finite fatigue life and the endurance limit. is referred to the article Stability of Residual component.
It is important to realize that in a component lo- Stresses in this Handbook, which deals with the Distribution and sign of loading mean
cally very high amounts of residual stresses may relaxation of residual stresses due to monotonic stresses depend on the external loading, the
exist. For example, in a spring steel up to 2000 or cycling loading or due to thermal energy. It geometry of the component, the elastic-plas-
MPa compressive residual stresses may be pro- is shown there that the micro residual stress state tic deformation behavior of the material (if
duced by a combined warm and stress peening is more stable against mechanically and ther- maximum loading stresses exceed the yield
process (Ref 3). mally induced relaxation than the macro residual strength or the cyclic yield strength), and on
In practice, no component is free of micro re- stress state. On the other hand, it is also proved thermophysical material properties (if tem-
sidual stresses. Almost all components have that during fatigue the micro residual stress state perature gradients exist). Distribution and
macro residual stresses to a certain extent. Only of a given steel may be changed by cyclic hard- sign of residual stresses depend on the
a small number of components exist where ening and/or cyclic softening processes, which amount and extension of inhomogeneous
macro residual stresses are negligible, for ex- are closely related to cyclic plastic deformations. plastic deformation in relation to the whole
ample after a suitable stress relaxation heat treat- Hence, during fatigue loading of a component component volume, on the interaction of me-
ment. The reasons for residual stress formation with locally varying macro as well as micro re- chanically and/or thermally driven deforma-
in components are treated in other articles in this sidual stress states, complex interactions of the tion during manufacturing processes, on the
Handbook. The most important ones are the vari- macro residual stress state with the cyclic load- strength of the material state, and/or on the
ous manufacturing processes. The fundamental ing stresses and the micro residual stress state volume fraction, the elastic as well as the
28 / Effect of Materials and Processing

thermophysical properties of the constituents if existing temperature gradients are changed Stage D: Unstable crack propagation and fail-
of the component that impose constraint on and finally removed. Residual stresses may be ure
each other. changed and eventually disappear by ther-
Loading mean stresses are in equilibrium mally and/or mechanically induced relaxa- An example of the consequences of these fatigue
with external forces and moments. Residual tion. stages is shown in Fig. 1 in an S-N diagram
stresses are in equilibrium with themselves Loading mean stresses are not influenced by (Wohler diagram) for the plain carbon steel SAE
regarding balance of forces and moments cyclic plastic deformation in stress-controlled 1020 (German grade C 20), which presents ex-
with relation to any sectional area and axis, loading, unless stress gradients become very perimental results for persistent slip band for-
respectively. The latter is also true for ther- highfor example in the root of sharp mation (ra /log Ns curve), microcrack initiation
mally induced loading stresses. notches, where mean stress redistribution (ra /log Ni curve), and failure (ra /log Nf curve)
Loading mean stresses may be changed and may occur. Residual stresses always relax, if (Ref 5).
finally disappear, if external loadings and/or the cyclic loading exceeds certain threshold Cyclic Deformation Behavior. Investiga-
values (see the article Stability of Residual tions on the cyclic deformation behavior in stage
Stresses in this Handbook). A using computerized servohydraulic test sys-
tems enable the determination of the complete
The interaction of the residual stress state with stress/total-strain response at distinct numbers
the cyclic loading stresses determines the fatigue
Stress amplitude (a), MPa

of cycles that yields to hysteresis loops. In a


Nf behavior of components. The following sections stress-controlled test, the plastic strain amplitude
Ni of this article discuss the consequences of this ea,p plotted versus log N results in a cyclic de-
interaction on cyclic deformation behavior, on formation curve that corresponds to the micro-
Ns crack initiation and crack propagation, as well as structural changes in the material during cyclic
on the fatigue life and on the endurance limit. loading. Decreasing (increasing) ea,p values with
Throughout this article, no special account is an increasing number of cycles N are typical for
given for multiaxial loading or residual stress a cyclic work-hardening (work-softening) be-
states and equivalent stresses are not used. In- havior.
stead, this article considers simple loading states Characteristic cyclic deformation curves of
and those components of the residual stress the low-alloy steel AISI 4140 (German grade 42
Number of cycles, N
statewhich is of course always multiaxial CrMo 4) in a normalized state are shown in Fig.
Wohler curves for slip band formation (Ns), mi-
that are relevant for the interaction of the cyclic 2. During stress-controlled push-pull tests with
Fig. 1 loading and the residual stress state under con- R rmin/rmax 1 and ra  RP0.2, the plastic
crocrack initiation (Ni), and failure (Nf). Source:
Ref 5 sideration. With regard to notched specimens strain amplitudes systematically change with the
in which a multiaxial stress state always exists stress amplitudes applied. In the initial stage of
upon mechanical loadingthe treatment is cyclic loading, a quasi-linear elastic behavior is
based almost entirely on nominal stresses and found. This stage is followed by a period of in-
Plastic strain amplitude (a,p), %

0.4
stress amplitudes. Hence, the fatigue strengths creasing ea,p values due to cyclically induced
and the cyclic loadings corresponding to finite work-softening effects that are connected with
0.3 fatigue lives given in this article are nominal inhomogeneous deformation processes caused
stress amplitudes. First of all, the section Some by fatigue Luders bands. This can be proved with
0.2 Aspects of Fatigue of Steels introduces some photoelastic investigations. A typical result ob-
basic definitions and relations of fatigue. This is tained for this steel is presented in Fig. 3. Plas-
0.1 done here rather briefly. The phenomenon and tically deformed areas appertaining to the indi-
the processes of the fatigue of steels are pre- vidual stages of inhomogeneous cyclic
sented in general and in detail in Volume 19 of deformation were registered photographically
0
the ASM Handbook (Ref 4). and marked as hatched areas on the specimens
1 10 102 103 104 105 106 that are allied to the cyclic deformation curve.
Number of cycles, N Similar macroscopic softening effects occur in
Some Aspects of Fatigue of Steels all normalized states of plain carbon steels and
Fig. 2 Plastic strain amplitude versus number of cycles low-alloy steels with carbon contents below 0.6
of normalized AISI 4140 steel. Source: Ref 68 wt% (Ref 68).
The fatigue behavior of metallic materials is
The quasi-linear elastic period shortens and
characterized by different processes. Generally,
the extent of cyclic work softening grows si-
it contains four successive stages A, B, C, and
multaneously with increasing stress amplitude
Plastic strain amplitude (a,p), %

D:
0.15 Homogeneous ra. After reaching a maximum of softening, the
Inhomogeneous
hardening Stage A: Cyclic deformation with work-hard- materials behavior is mainly characterized by
softening ening or/and work-softening effects in the to- work hardening, and the ea,p values decrease
0.10 tal volume (homogeneous loading such as with increasing number of cycles. At the end of
push-pull of smooth specimens) or in the the specimens life, the plastic strain amplitude
highest loaded regions (inhomogeneous load- increases again, however fictitiously as a con-
0.05 ing such as bending, torsion, or any loading sequence of changes in the compliance of the
of notched components) and development of specimens due to crack propagation. The plot of
persistent slip bands at the surface ra versus ea,t at N Nf /2, the so-called cyclic
Stage B: Microcrack initiation and propaga- stress-strain curve, of this steel is presented in
01 10 102 tion. Normal evolution of a macrocrack in the Fig. 4 together with the monotonic-stress/total-
Number of cycles, N
surface region strain curve at nearly the same strain rate e . In the
Stage C: Stable macrocrack propagation con- elastic-plastic range, the cyclic curve is lower
Fig. 3 Cyclic deformation curve and development of
inhomogeneous strain distribution during cyclic nected with changes in the state of the mate- than the monotonic curve due to the work soft-
softening of normalized AISI 4140 steel. Source: Ref 68 rial in the crack tip plastic zone ening. The cyclic stress-strain curves of plain
Residual Stresses and Fatigue Behavior / 51

or quenched-and-tempered (at a low tempera- plitudes. However, the micro residual stress state (see Fig. 62a and c, 63, and 64c) and the im-
ture) steel, upper band in the figure) are con- may be changed significantly by the processes provement of the fatigue strength even by high
cerned that generate the macro residual stress. Hence, compressive residual stresses is limited (see Fig.
The depth distribution of the macro residual in a medium-strength steel the local resistance to 45a, 46, and 56). With increasing stress ampli-
stress rrs, characterized by its sign, magni- residual stress relaxation may be quite different tude and hence loading stress gradient, the crack
tude, and gradient grs even if the same initial material state is con- initiation site is shifted to the surface. Therefore,
The depth distribution of the micro residual cerned, and unique borderlines for the onset of the finite fatigue life is much more improved
stress rrs
micro residual stress relaxation do not exist (see Fig. than the fatigue strength (see Fig. 62a and c, 63,
The notch factor kt (smooth and mildly 60). A significant benefit from compressive re- and 45d). In notched specimens, the loading
notched specimens are regarded in the figure) sidual stresses is only obtained if their penetra- stress gradient is larger than in smooth ones. If
The gradient of the loading stress gls tion depth is sufficiently high and/or their gra- high compressive residual stresses with a suffi-
The surface topography, characterized by the dient is sufficiently low as compared to the cient depth and/or a small gradient exist, crack
roughness height Rh gradient of the loading stresses. Therefore, com- initiation will occur in the notch root resulting
pressive residual stresses produced by grinding in a strong improvement of the fatigue strength
do hardly influence the fatigue strength of (see Fig. 64a and b, 45b, 46, and 56). However,
Low-Strength Steel smooth and notched specimens (see Fig. 36). if the penetration depth of the compressive re-
Contrarily, compressive residual stresses gener- sidual stresses is low and therefore the gradient
In a low-strength steel, there will be no or very
ated by deep rolling or shot peening increase the of the local fatigue strength highfor example,
little influence of the macro residual stress, be-
notch fatigue strength (see Fig. 38, 39, and 41). after grindingthe improvement of the notch
cause it is relaxed more or less completely if the
The fatigue strength of smooth specimens is less fatigue strength will be small (see Fig. 44 and
cyclic loading approaches the fatigue strength
improved, if subsurface crack initiation occurs. 59). Frequently, after shot peening or deep roll-
(see Fig. 32 and 48). A change of the micro re-
Then, the fatigue strength of notched specimens ing, maximum compressive residual stresses oc-
sidual stress state by work hardening may sig-
may even be higher than corresponding values cur below the notch root. Then, the fatigue
nificantly increase Rf since the resistance against
of smooth specimens (see Fig. 41). This means strength may not be determined by the maximum
cyclic plastic deformation and hence, crack ini-
that the fatigue notch factor kf becomes less than cyclic loading, which does not result in crack
tiation increases (see Fig. 31). Then, also the re-
unity. At vanishing residual stresses, kf comes initiation, but by the maximum cyclic loading at
sistance against macro residual stress relaxation
rather close to kt if the loading stress gradient is which crack arrest below the surface is possible
is raised resulting in a certain sensitivity of the
small, but is significantly less at higher gls val- (see Fig. 67, 68, and 73). As a consequence of
work-hardened zone to macro residual stress.
ues. With increasing tensile residual stresses, the all these relationships, the fatigue notch factor
This may be detrimental or beneficial (see Fig.
fatigue notch factor is reduced again, as expected may vary strongly in the presence of compres-
33b) for Rf depending on the sign and the mag-
on basis of the Goodman relationship for smooth sive residual stresses regarding one-notch ge-
nitude of the macro residual stress. The influence
and notched specimens (see Fig. 58 and 60). ometry and may take values ranging from less
of the surface topography is rather small in a
There is a considerable influence of the surface than unity to the notch factor kt, as sketched in
low-strength steel (see Fig. 47). The fatigue
roughness on the fatigue strength, as shown in Fig. 75. In fact, regarding smooth and notched
notch factor kf is significantly smaller than the
Fig. 47. specimens, Fig. 44 and 54 prove that kf of ground
notch factor kt because cyclic plastic deforma-
specimens with compressive residual stresses
tion and stress redistribution occur in the notch
High-Strength Steel produced by grinding approaches kt. From Fig.
root (see Fig. 32). With increasing loading stress
56 it can be deduced that kf of shot peened spec-
gradient at a given kt, the fatigue strength in-
In high-strength steels, stress relaxation dur- imens approaches unity. At vanishing residual
creases due to the decrease of the highly stressed
ing cyclic loading in the range of the fatigue stresses, kf comes close to kt as expected in a
volume of the component or specimen (compare
strength only occurs in notched specimens bear- high-strength steel. With increasing tensile re-
kt 2.5, g 2 with kt 2.5, g 5 in Fig.
ing very high compressive residual stresses. sidual stress, the fatigue notch factor is reduced
32).
Then, the resulting fatigue strength is also high, and finally approaches unity (see Fig. 44 and 54).
and during corresponding cyclic loading very Again, this finding is in correspondence with the
Medium-Strength Steel high magnitudes of the minimum stress occur, Goodman relationship for smooth and notched
which leads to some residual stress relaxation specimens (see Fig. 58 and 59). The surface
In a medium-strength steel, there is a signifi- (see Fig. 57b). Contrarily, in the range of high roughness has principally a large influence on
cant influence of the macro residual stress on Rf tensile residual stresses and cyclic loadings that the fatigue strength of high-strength steel, as
since only a small part of rrs relaxes during cy- lead to infinite life or to technically relevant life- shown in Fig. 47. On the other hand, in practice
clic loading in the range of the fatigue limit (see times the occurring maximum stresses are much the roughness height of hard steel is rather low
Fig. 36, 38, 51a, and 52). However, in the low- lower and no residual stress relaxation is ob- even after mechanical surface treatments such as
cycle fatigue range, relaxation becomes more served even in the range of low-cycle fatigue. shot peening.
complete with increasing amplitude, and the in- Consequently, the residual stress sensitivity and
fluence of the macro residual stress vanishes (see the mean stress sensitivity of Rf are identical (see Recommendations
Fig. 34, 37, 39, 51b). Tensile residual stresses Fig. 59), and the fatigue strength is strongly re-
are always detrimental to Rf. Therefore, in the duced with increasing tensile residual stress. From all of these relationships, some recom-
presence of large tensile residual stresses a me- This is also true for the finite fatigue life (see mendations may be deduced. In medium- and
dium-strength steel may have equal or even Fig. 42). In the range of compressive residual high-strength steels tensile macro residual
lower fatigue strength than a low-strength steel stresses, complex relationships exist. A strong stresses must strictly be avoided since they al-
(compare Fig. 32 with 36). If relaxed tensile re- effect of rrs will only occur if cracks are initiated ways promote crack initiation and crack propa-
sidual stresses are concerned, the residual stress at the surface. However, in thick smooth speci- gation and are detrimental to the fatigue strength
sensitivity m of the fatigue strength approaches mens or components cyclically loaded in the andat least in higher strength steelto finite
the mean stress sensitivity M (see Fig. 60). In range of the fatigue strength, the loading stress fatigue life. In a low-strength steel, the influence
material states of similar hardness loaded in the gradient is usually lower than the gradient of the of tensile macro residual stresses is usually small
range of the fatigue limit, compressive residual local fatigue strength, which depends on the or negligible, and the change of the micro resid-
stresses relax stronger than tensile ones simply depth distribution of the residual stresses. Con- ual stress state by work hardening is much more
because of the different corresponding stress am- sequently, cracks are initiated below the surface important. In most cases, work hardening will be
30 / Effect of Materials and Processing

rate in this region is below 108 mm/cycle. The Quantitatively, the following relationships are
upper parts of the curves represent the crack valid:
Crack growth rate, log da/dN

growth behavior as the Kmax values approach the


critical stress intensity factor KIc, which marks ea,p ef N a
f (Eq 5)
the onset of unstable crack propagation. Then the
crack growth rate increases boundlessly. and
R1 The Paris law does not represent the complete
crack growth curves shown in Fig. 9. The upper ea,e rf /E N b
f (Eq 6)
R2 part of the curves depends on KIc, and the lower
part on the threshold value DKth. Therefore, sev- with ef the fatigue ductility coefficient, rf the
R3
eral complementations were suggested. An equa- fatigue strength coefficient, the fatigue ductil-
tion that incorporates all of the features of crack ity exponent, b the fatigue strength exponent,
Stress intensity factor range, log K growth curves has been used by NASA (Ref 13). and E the Youngs modulus. Rearrangement of
It has the form: both equations yields to:
Fig. 9 Crack propagation rate versus stress intensity fac-
tor range for stress ratio values R1  R2  R3
ea,p N fa constant (Eq 7)
da C (1 R)m DKn (DK DKth)p

dN [(1 R) KIc DK]q and
(Eq 2)
log a ea,e E N b
f ra N b
f constant (Eq 8)
This equation is able to account for the threshold
effect, the R ratio effect, and the final instability which are the well-known relationships from
log f effect. By suitable choice of the exponents, this Coffin-Manson (Ref 17, 18) (Eq 7) and Basquin
log Rm equation can accept a number of different vari- (Eq 8).
ants of crack propagation laws. For m p q Tensile mean stresses promote crack initiation

0, for example, the Paris law is obtained. and accelerate crack propagation. Therefore, the
The concepts of linear elastic fracture me- fatigue strength Rf decreases with increasing
chanics and the use of the stress intensity factor mean stresses. This correlation is normally rep-
log Rf resented in a Smith (Ref 19) or in a Haigh (Ref
are limited to cases in which small-scale yielding
develops at the crack tip. Moreover, the appli- 20) diagram. As an example of such fatigue
cability of Eq 1 and 2 is restricted to sufficiently strength diagrams, Fig. 12 shows a Haigh dia-
log Ns log Ne log Nf long macroscopic cracks. Short cracks can grow gram. The fatigue strength Rf is plotted versus
Fig. 10 Stress Wohler curve (S-N curve) at smaller ranges of the stress intensity factor as the mean stress rm. The Goodman approxima-
DKth. At comparable DK values, they can exhibit tion (Ref 21):
clearly higher crack growth rates than long
cracks (Ref 14). Rf R0f (1 rm /Rm) (Eq 9)
Lifetime Behavior. The lifetime behavior of
steels is commonly described in an S-N diagram which represents a straight line between the fa-
(Wohler diagram) (Ref 15), where the loading tigue strength R0f at rm 0 and the tensile
Strain amplitude (a)

f amplitude is plotted versus the number of cycles strength Rm, is frequently used to estimate fa-
to failure in double logarithmic scaling. For tigue strength. For practical use, the allowable
stress-controlled fatigue tests, the S-N curve is stresses are restricted by the demand that mac-
f/ E frequently described by three straight lines (see roscopically no plastic deformation should oc-
Fig. 10). In the range of the quasi-static strength cur. This restriction leads to a straight line con-
E N Rm (Nf  Ns) and the fatigue strength Rf(N  necting the yield strength Re at both axes.
N N0f ), the endurable stress amplitude is constant Through this, the allowable stress amplitudes lie
and given by the ultimate strength Rm and the in the hatched area in Fig. 12. The slope of the
Number of cycles to failure, Nf fatigue strength Rf, respectively. In the range of Goodman straight line is often also described as
finite life (Ns  Nf  Ne), the Basquin relation mean stress sensitivity M R0f rm /Rm. The
Fig. 11 Total strain Wohler curve
(Ref 16): Goodman approximation normally leads to con-
servative estimations. Often, the Gerber parabola
ra rf N b (Eq 3) (Ref 22):
f

Rf R0f [1 (rm /Rm)2] (Eq 10)


often describes the lifetime behavior satisfacto-
rily. rf is the fatigue strength coefficient and b is used as a nonconservative estimation of the
Re
Fatigue strength, Rf

Goodman the fatigue strength exponent. fatigue strength. Mostly, experimental results
0 In the case of total-strain-controlled fatigue range between the Goodman and the Gerber ap-
Rf Gerber
tests the ea,t values in the total strain Wohler proximations.
curve (Fig. 11) can be separated into the elastic In the range of infinite lifetime, the influence
and plastic parts: of mean stresses can be described by damage
parameters. Two frequently used damage param-
ea,t ea,e ea,p (Eq 4) eters are the one proposed by Smith, Watson,
and Topper (Ref 23):
Re Rm From experience it is known that both parts ex-
tend exponentially from the number of cycles to PSWT (ra rm) ea,t E (Eq 11)
Mean stress m
failure Nf, which means that the plots log ea,e
Fig. 12 Endurance limit versus mean stress versus log Nf and log ea,p versus log Nf are linear. and the one proposed by Ostergren (Ref 24):
Residual Stresses and Fatigue Behavior / 31

POST (ra rm) ea,p E (Eq 12) by shot peening or deep rolling. Characteristic 400 MPa show cyclic softening from the first
cyclic deformation curves for stress-controlled cycle and higher plastic strain amplitudes during
ra and rm are the imposed stress amplitude and push-pull loading of different heat treated the first cycles for stress amplitudes ra between
mean stress, respectively, ea,t and ea,p are the re- smooth specimens of the steel AISI 4140 (Ger- 250 and 350 MPa. After a certain number of cy-
sulting total strain amplitude and plastic strain man grade 42 CrMo 4) are compared in Fig. 13 cles, the opposite tendency can be detected and
amplitude, respectively, and E is the Youngs in unpeened and in shot peened conditions with the plastic strain amplitudes of the shot peened
modulus. Normally, for ea,t and ea,p the values at compressive residual stresses at the surface (Ref conditions are smaller than those of the unpeened
half of the lifetime are used. 25, 26). In the normalized state (Fig. 13a), the material states. However, for the same ra values
onset of cyclic deformation is different in both the plastic strain amplitudes of both conditions
conditions, since the shot peened specimens with approach another at relatively high numbers of
Influence of Residual Stresses on surface compressive residual stresses rrs cycles. Corresponding results for a quenched-
the Cyclic Deformation Behavior and-tempered (730 C/2 h) AISI 4140 steel are
presented in Fig. 13(b). In the unpeened condi-
Characteristic Examples. The influence of 0.20 tion, the characteristic cyclic deformation behav-

Plastic strain amplitude (a,p), %


macro and micro residual stresses on the cyclic 600 a in MPa ior of quenched-and-tempered steels occurs with
deformation behavior can be studied very well Unrolled a quasi-elastic incubation period, which is fol-
0.15
after mechanical surface treatments, for example 550 Deep rolled lowed by cyclic softening until crack initiation.
After shot peening that generates surface com-
0.10 510 pressive residual stresses rrs 410 MPa, the
onset of cyclic softening is shifted to smaller
0.4 600 550
Plastic strain amplitude (a,p), %

a in MPa 0.05 500 numbers of cycles. Furthermore, it is interesting


Unpeened 450 to note that for identical stress amplitudes and
0.3 comparable numbers of cycles, the higher plastic
Shot peened 380 0
strain amplitudes are always measured for the
0.2 350 1 10 102 103 104 105 106 shot peened specimens. Figure 13(c) shows a
325
Number of cycles, N compilation of cyclic deformation curves for an-
(a)
0.1 300 other quenched-and-tempered (570 C/2 h) AISI
250 0.04 4140 steel with a higher strength compared with
Plastic strain amplitude (a,p), %

a in MPa the steel condition in Fig. 13(b). In this case, the


Unrolled 600
0 shot peened condition that has surface compres-
0.03 Deep rolled 650
1 10 102 103 104 105 106 sive residual stresses rrs 530 MPa is char-
Number of cycles, N 550 acterized for all investigated ra values by small
0.02 600 measurable plastic strain amplitudes during the
(a)
500 first cycle that diminish or disappear first of all
Plastic strain amplitude (a,p), %

0.4 0.01 with an increasing number of cycles. After a sub-


550
a in MPa 450 sequent regime of quasi-elastic behavior, cyclic
Unpeened 500
0.3 softening, which yields to lower plastic strain
Shot peened 0
amplitudes and larger numbers of cycles to fail-
1 10 102 103 104 105 106 ure in comparison with the unpeened conditions,
0.2
Number of cycles, N is dominant.
400 (b)
450 The effect of deep rolling on the behavior of
375
0.1 the cyclic deformation curves is presented in Fig.
350 Fig. 14 Cyclic deformation curves for stress-controlled
400 350 push-pull loading of (a) normalized and (b) ni- 14(a) for the normalized plain carbon steel SAE
trocarburized specimens of the SAE 1045 steel in unpeened 1045 (German grade Ck 45) (Ref 27). At com-
0 275 and deep-rolled conditions. Source: Ref 27
parable stress amplitudes essentially smaller
1 10 102 103 104 105 106
plastic strain amplitudes are measured in the
Number of cycles, N
(b) deep-rolled condition. In this materials state, the
0.10 cyclic softening effects are extremely restricted,
Plastic strain amplitude (a,p), %

but continue during the whole lifetime until


Plastic strain amplitude (a,p), %

1.0
Untreated
900 a in MPa 0.08 crack initiation. As a consequence of the smaller
0.8 Unpeened Shot peened
800
Shot peened
plastic strain amplitudes at identical stress am-
0.06 plitudes, higher numbers of cycles to failure are
0.6
observed for the deep-rolled condition compared
750 0.04 Deep rolled with the normalized ones. Figure 14(b) shows
0.4
700 the influence of deep rolling in the case of a ni-
0.2
0.02 trocarburized SAE 1045 steel. Due to the differ-
500 400
550
ent microstructures in the near-surface areas of
600 0
0 the specimens, the same stress amplitudes yield
1 10 100 103 104 105
plastic strain amplitudes that are five times
1 10 102 103 104 105 106 Number of cycles, N
smaller than those of the normalized state. The
Number of cycles, N deep-rolling process modifies the cyclic defor-
(c) Cyclic deformation curves of different mechan-
Fig. 15 mation curves of the nitrocarburized specimens
ically surface treated materials for stress-con-
Fig. 13 Cyclic deformation curves for stress controlled trolled push-pull loading in the low-cycle fatigue (LCF) re- in a manner similar to the normalized specimens.
push-pull loading of (a) normalized, (b) gime of the plain carbon SAE 1045 steel. Stress amplitude However, the reduction of the plastic strain am-
quenched-and-tempered (730 C/2 h), and (c) quenched- ra 350 MPa, Almen intensity I 0.120 mm A, rolling
and-tempered (570 C/2 h) smooth specimens of the steel pressure p 150 bar; surface residual stresses: shot peen-
plitudes is much less pronounced.
AISI 4140 in unpeened and shot peened conditions. ing rrs  500 MPa, deep rolling, approximately 600/ Figure 15 shows plastic strain amplitudes as a
Source: Ref 25, 26 350 MPa. Source: Ref 28, 29 function of the number of cycles plotted for
32 / Effect of Materials and Processing

push-pull loading with the stress amplitudes in- damage is clearly correlated with plastic strain specimens and in Fig. 18 for quenched-and-tem-
dicated for a normalized SAE 1045 steel (Ref amplitude, the benefit of mechanical surface pered specimens of SAE 1045 with a notch fac-
28, 29). Results for annealed as well as for shot treatments becomes obvious. The influence of tor Kt 3.0 that were investigated in push-pull
peened or deep-rolled conditions are shown. One mechanical surface treatments on cyclic plastic- tests (Ref 30, 31). Figure 17(a) compares the cy-
can clearly see that both mechanical surface ity can be summarized in cyclic stress-strain clic deformation curves of material states with
treatments considerably diminish the plastic curves, which correlate stress amplitudes and very low residual stresses (produced by milling
strain amplitudes. Due to the thicker affected plastic strain amplitudes for, for example, half and subsequent annealing) with rather high re-
surface layer in the case mentioned first, the ef- the number of cycles to failure (see the section sidual stresses (produced by shot peening). In
fect for deep-rolled states is more pronounced Cyclic Deformation Behavior). Figure 16, as both conditions plastic deformation occurs in the
than for shot peened states. Because fatigue an example, shows data for an austenitic AISI first cycle at nominal stress amplitudes rn,a
304 steel in shot peened as well as deep-rolled above 150 MPa. The annealed state yields con-
conditions compared with untreated specimens tinuous cyclic softening for all investigated rn,a
(Ref 28, 29). Effects of mechanical surface treat- values. Contrarily, the initial plastic strain am-
700
ments are more distinct with a higher ratio be- plitudes of the shot peened condition that are
Stress amplitude (a), MPa

600 tween the area of affected surface layer and the much less dependent on the stress amplitude
500
cross section of the specimen. In the case pre- compared to the milled and annealed state are
Deep rolled (1 mm wall thickness)
Deep rolled (0.5 mm wall thickness) sented here, hollow thin-walled specimens were reduced during the first cycles. This is a result
400 Deep rolled (0.3 mm wall thickness) prepared from compact specimens for analysis. of residual stress relaxation that is the more pro-
300 As shown in Fig. 16, cyclic yield strength con- nounced the higher the nominal stress amplitude
siderably increases compared with compact is (see the following section Evaluation of Ex-
200 Untreated specimens, if only the fatigue behavior of the perimental Results). The consequences are
Deep rolled
100 Shot peened surface layers of mechanically surface treated smaller effective stress amplitudes and smaller
0
components is investigated. ea,p values in the following cycles. Hence, the
0 0.1 0.2 0.3 0.4 0.5 If residual stresses are present in the root of observed cyclic work hardening is fictitious.
Plastic strain amplitude (a,p), % notched specimen, marked changes always occur However, at higher numbers of cycles, cyclic
in the initial parts of the notch root cyclic defor- softening occurs with lower ea,p values than in
Fig. 16 Cyclic stress-strain curves of deep-rolled AISI mation curves (ea,p /log N curves) and in the the unpeened condition.
304 (rolling pressure 150 bar, surface residual notch root cyclic mean strain curves (em /log N On the other hand, as can be seen from Fig.
stresses r  350 MPa) using compact as well as hollow
rs
l curves) that depend on the sign and the magni- 17(b), the shot peening notch root residual
specimens with different percentages of strain-hardened
layers compared with an untreated state and a shot peened tude of the residual stresses. Characteristic ex- stresses have an influence on the initial parts of
condition (rrs  450 MPa). Source: Ref 28, 29 amples are presented in Fig. 17 for normalized the mean strain curves that were determined for
different nominal stress amplitudes (Ref 30, 31).
As a consequence of the cyclic deformation in-
duced relaxation of the compressive residual
True plastic strain amplitude (a,p), %

0.06
True plastic strain amplitude (a,p), %

stresses in the notch, the initial parts of the em /


n,a = 200 MPa
0.4 log N curves show negative plastic mean strains,
n,a in MPa where the magnitudes of em increase with in-
0.3 rs = 15 MPa 0.04 creasing nominal stress amplitude. The minima
rs = 490 MPa
of em are shifted to lower numbers of cycles with
220 rs = 710 MPa
0.2
increasing rn,a values. However, from the min-
240 460 ima of em at approximately 5% of the number of
220 0.02
200
320
cycles to failure, the em /log N curves show in-
0.1 175
150 creasing mean strains for all nominal stress am-
170 150
190 0 plitudes. Simultaneously, a decrease of the ten-
1 10 102 103 104 105 106 sile compliance in the hysteresis loops that
0 intensifies with an increasing number of cycles
Number of cycles, N
1 10 102 103 104 105 106 (a) is observed. By that, the compressive compli-
Number of cycles, N
0.06 ance is approximately constant. This finding
(a)
proves that this increase of the mean strains with
Plastic mean strain (m), %

220 n,a = 200 MPa the number of cycles is caused by an increase of


0.30 190 170 0.04 the number of microcracks and/or the growth of
n,a in MPa
Plastic mean strain (m), %

150
microcracks. These sections are indicated by
0.02
dashed lines in the em /log N curves of Fig. 17(b).
0.15 The true plastic strain amplitudes in the cyclic
460
320 deformation curves of the quenched-and-tem-
0 pered (400 C/2 h) specimens in Fig. 18(a) are
0.00 rs = 710 MPa influenced by the magnitude of the notch root
0.2 residual stresses, but not by their sign (Ref 30,
1 10 102 103 104 105 106 31). All investigated conditions in Fig. 18(a)
0.15 Number of cycles, N
1 10 102 103 104 105 106
show for the initial cycles fictitious cyclic-work-
(b)
Number of cycles, N
hardening effects due to the relaxation of resid-
(b) True plastic strain amplitude (a) and plastic ual stresses rrs of the surface layers (rrs 460
Fig. 18 MPa and 320 MPa produced by upcut milling
mean strain versus number of cycles (b) for
Fig. 17 True plastic strain amplitude (a) and plastic stress-controlled push-pull loading of quenched-and-tem- as well as 710 MPa by shot peening). Up to
mean strain versus number of cycles (b) for pered (400 C/2 h) notched specimens (notch factor Kt
stress-controlled push-pull loading of normalized notched 3.0) of the SAE 1045 steel in cut-milled and shot peened
N 20 cycles, increasing magnitudes of the
specimens (notch factor Kt 3.0) of the SAE 1045 steel in conditions with different surface residual stresses. Source: compressive residual stresses cause higher neg-
unpeened and shot peened conditions. Source: Ref 30, 31 Ref 30, 31 ative mean strains. However, positive milling re-
Residual Stresses and Fatigue Behavior / 33

sidual stresses yield to positive mean strains. The stricted mean free path of the mobile dislocations plastic strain of notched specimens (see Fig.
increases in the plastic mean strains that appear in the work-hardened surface layers. 18b).
for the numbers of cycles above the minimum of In hardened as well as in quenched-and-tem- A study of the residual stress effects on the
em are caused also by microcracks as described pered at low temperature conditions, the changes cyclic deformation behavior seems to be possi-
previously. in the cyclic deformation behavior result not ble by their simulation by applied mean stresses,
Evaluation of Experimental Results. In rela- only from surface hardening or softening effects, which are homogeneously distributed over the
tively soft material states, as for example in nor- but also from the more stable residual stresses as cross section of the specimen. An example is
malized as well as in quenched-and-tempered at can be seen in Fig. 19(b) for notch root residual presented in Fig. 20 for smooth, normalized SAE
high-temperature conditions, the consequences stresses of the quenched and tempered (400 C/ 1045 steel specimens that were stress-controlled
of mechanical surface treatments by shot peen- 2 h) SAE 1045 steel under the loading condi- loaded in push-pull tests with a constant mean
ing or deep rolling on the cyclic deformation be- tions of Fig. 18(a) (Ref 30, 31). The conse- stress of rm 300 MPa (Ref 32). The cyclic
havior are mainly caused by near-surface micro quences are a considerable influence of the deformation curves at stress amplitudes ra
residual stresses, that is, work hardening of the residual stresses on the cyclic behavior, espe- 125 and 150 MPa in Fig. 20(a) show first of all
surface layers, because the macro residual cially with regard to the development of mean a quasi-elastic incubation interval, followed by
stresses are relaxed very soon by cyclic plastic considerable cyclic softening within some cy-
deformation (see the article Stability of Resid- cles. Subsequently, within some further cycles,
ual Stresses in this Handbook). This relaxation cyclic work hardening occurs to such an extent
of macro residual stresses is shown in Fig. 19(a)
Plastic strain amplitude (a,p), % 0.20 that extremely low plastic strain amplitudes re-
250
for shot peening residual stresses in the notch a in MPa sult. For ra 200 MPa, the compressive stress
root of a normalized SAE 1045 steel (Ref 30, 0.16 peak | rm ra |, which is higher than the yield
31) under the loading conditions of Fig. 17(a). 125 strength, induces plastic deformation in the first
The dislocation structures in the ferrite after the 0.12 cycle. This procedure is followed by a rapid cy-
mechanical surface treatment are not stable and 150 clic work hardening in such a manner that at cy-
change during cyclic loading in energetically 0.08 cles above 100 extremely low ea,p values are ob-
more favorable arrangements. The formation of 350 served. As a consequence of the constant
typical fatigue-induced dislocation structures is 0.04 compressive mean stress, the specimens shorten
200 100
combined with cyclic softening effects as pre- for all investigated stress amplitudes, as proved
0
sented in Fig. 13(a), 14(a), and 17(a). The small 1 10 102 103 104 105 106
by Fig. 20(b). For ra 200 MPa, the mean
plastic strain amplitudes of the shot peened or Number of cycles, N
strains increase considerably and continuously.
deep-rolled conditions and the resulting in- (a) However, for ra  200 MPa the ea,p values sat-
crease in fatigue life are caused by the re- urate above 20 cycles.
3
100 150 125 200
Plastic mean strain (m), %

0
3
Influence of Residual Stresses on
250 the Crack Initiation
Residual stress (l ), MPa

0 9
n,a = 220 MPa
Characteristic Examples. Crack initiation
350
occurs as a consequence of microstructural
rs

200 15 changes in metallic materials during cyclic load-


a in MPa
400 ing. Different mechanisms are responsible for
n,a = 150 MPa 450 their formation (e.g., Ref 33). If it is accepted
400 21
1 10 102 103 104 105 106 that for given materials states at comparable load
Number of cycles, N amplitudes increasing amounts of plastic strain
(b) amplitudes lead to decreasing numbers of cycles
600
to crack initiation Ni, it follows that residual
0 1 10 102 103 104 105 106 Fig. 20 Plastic strain amplitude (a) and plastic mean
stresses may extend, shorten, or leave unchanged
Number of cycles, N strain versus number of cycles (b) for axial
stress-controlled cyclic loading with mean stress rm 300 the number of cycles to crack initiation. How-
(a) MPa for specimens of the normalized SAE 1045 steel. ever, experimental investigations concerning the
Source: Ref 32 influence of macro and micro residual stresses
n,a = 200 MPa on crack initiation are scarce. This is due to the
Residual stress (l ), MPa

300
difficulties connected with the observation of the
formation and the propagation of small cracks.
rs

100
0 A recently published investigation gives a re-
A Untreated
port on the influence of mechanical surface treat-
Ratio of cycles N/Nf, %

300 80 B Deep rolled


ments on crack initiation and crack propagation
C Shot peened
in push-pull loading of steels (Ref 28). In un-
600 60 Macrocracks
(>750 m) treated materials, crack initiation normally takes
Short cracks place at positions of high localized slip, for ex-
900 40 (200-750 m) ample, at extrusions and intrusions that are con-
0 1 10 102 103 104 105 106 Microcracks nected with persistent slip bands. However, as
Number of cycles, N 20 (>200 m)
(b) shown in Fig. 21 in shot peened and deep-rolled
No damage
A B C marks
conditions of the austenitic steel AISI 304, crack
Fig. 19 Residual stress relaxation during stress-con- 0 formation occurs later than in the untreated state
trolled push-pull loading of notched specimens due to the consequence of numerous obstacles
(notch factor Kt 3.0) for (a) normalized SAE 1045 in shot Fig. 21 Influence of mechanical surface treatments on
peened conditions and (b) quenched-and-tempered (400 the damage evolution of the push-pull loaded
for slip (dislocations, grain and twin boundaries)
C/2 h) SAE 1045 in cut-milled and shot peened conditions. austenitic steel AISI 304. ra 320 MPa; R 1; A, Nf in the work-hardened surface layer that impede
Source: Ref 30, 31 3859; B, Nf 4445; C, Nf 20,265. Source: Ref 28 localized slip. In these surface-work-hardened
34 / Effect of Materials and Processing

conditions, no persistent slip band is observed at compressive residual stresses near the surface is literature. If the corresponding crack length
all. Furthermore, crack propagation is slower high enough. Characteristic results are presented amounts to only some grain diameters, then Ni
than in the untreated state due to the effect of in Fig. 23 for shot peened bending specimens in is determined by shear-stress-controlled pro-
microstructure and compressive residual a hardened state of the steel SAE 1045 with the cesses in stage I of crack propagation. However,
stresses. Similar results are found in Ref 28 for bending fatigue strength Rf 960 MPa after if crack lengths of several hundred microns ap-
the normalized steel SAE 1045, in Ref 34 for the shot peening with the shot size d 0.3 mm and pear (stage II of crack propagation), Ni will be
plain carbon steel SAE 1080, and in Ref 35 and Rf 1050 MPa with d 0.6 mm, respectively determined by normal-stress-controlled pro-
36 for a high-strength spring steel AISI 6150 (Ref 38, 39). For stress amplitudes ra  Rf, crack cesses. In the last case, stable macro residual
(German grade 50 CrV 4). initiation was observed directly at the surface. stresses may have a considerable influence on
However, as reported in Ref 1, the crack ini- However for ra  Rf, subsurface cracks occurred the Ni values.
tiation time of shot peened specimens is some- as shown in Fig. 24. If the centers of the rosettes The experimental results achieved until now
times shorter than that of unpeened specimens on the scanning electron micrographs (SEMs) for crack formation directly at the surface with
despite increased lifetimes. One example is are considered to be the crack initiation points, an appropriate consideration of the Ni definition
given in Fig. 22 for the quenched-and-tempered it is obvious that the crack initiation depth in- show increasing numbers of cycles to crack ini-
steel SAE 1045 in the case of bending fatigue creases with decreasing stress amplitude (Ref tiation, if the surface area:
tests in seawater (Ref 37). With the exception of 39). At stress amplitudes that result in a fatigue
Becomes work hardened by mechanical sur-
high stress amplitudes, the cracks are formed life Nf above 107 cycles, crack initiation is lo-
face treatment
earlier in shot peened specimens than in ground cated approximately 0.3 mm below the surface.
Bears additional compressive macro residual
ones. This finding is attributed to an enhanced These findings stimulated the development of a
stresses
crack initiation at micronotches resulting from concept of the local fatigue strength (see the par-
Is smoothed by polishing or deep rolling
shot peening, which is obviously supported by agraphs on experimental results in the section
corrosion pittings in the case of seawater envi- Influence of Residual Stresses on S-N curves However, decreasing Ni values are observed,
ronment. However, the number of cycles to fail- in this article). if in the surface area:
ure of the shot peened conditions are higher than Evaluation of Experimental Results. The def-
The roughness is increased, for example, by
that of ground conditions. inition of the number of cycles to crack initiation
Residual stresses may have a remarkable in- Ni and particularly the establishment of the af- shot peening
Tool marks or material overlaps from shot
fluence on the location of crack initiation. In the filiated crack length is of decisive meaning for
case of bending fatigue, for example, cracks may the evaluation of the influence of different pa- peening or deep rolling exist
Tensile macro residual stresses occur that fa-
start below the surface, if the magnitude of the rameters on Ni from experimental results from
vor interface cracks between matrix and hard
second phases.

800 In this context, an investigation of the fatigue


(a) crack initiation in double-edge notch specimens
Stress amplitude (a), MPa

Fracture
of SAE 1080 steel with a fine-grained, spheroid-
Crack initiation
600 200 m ized microstructure shows considerable effects
of the macro residual stresses (Ref 34). In a
(b) press-fitted condition, large compressive resid-
400
ual macro stresses occur that remained relatively
200 m stable throughout the fatigue life and thus greatly
200 increased the numbers of cycles to crack initia-
Shot peened
Ground (c) tion Ni. With reference to the Basquin-relation-
0 ship, Ni is estimated from the following relation-
104 105 106 107 200 m ship:
Number of cycles, N
1/b;

r (r r )
(d) 2ra
Stress amplitude versus number of cycles to Ni (Eq 13)
Fig. 22 i a
rs
crack initiation and to failure of a quenched-
and-tempered SAE 1045 steel under bending fatigue load- 200 m
ing in seawater. Source: Ref 37 where Ni is defined as the number of cycles at a
(e) crack length of a 0.1 mm, rrs is the macro
residual stress, ra is the applied stress amplitude,
0.4 200 m and ri 2315 MPa and bi 0.197 are ma-
terial constants.
Distance from surface, mm

(f)
0.3
Influence of Residual Stresses on
200 m
0.2 the Crack Propagation
Site of
macrocrack
(g)
0.1 initiation Characteristic Examples. During cyclic
Shot size 0.3 mm 200 m
loading, the lifetime of components is deter-
Shot size 0.6 mm mined significantly by the stage of crack prop-
0
agation, especially in the case of existing macro
800 1000 1200 1400
Fig. 24 Scanning electron micrographs of fractured residual stresses. Crack propagation as a conse-
Stress amplitude (a), MPa specimens of hardened and shot peened SAE quence of fatigue loading without mean stresses
1045 at different stress amplitudes. (a) ra 1300 MPa. (b)
Crack initiation sites versus stress amplitude of ra 1250 MPa. (c) ra 1200 MPa. (d) ra 1100 MPa.
can be described, with the exception of the crack
Fig. 23 initiation and the near-fracture stage, by the Paris
a hardened SAE 1045 steel in different shot (e) ra 1000 MPa. (f) ra 950 MPa. (g) ra 900 MPa.
peened conditions. Source: Ref 38, 39 Source: Ref 39 equation (see Eq 1):
Residual Stresses and Fatigue Behavior / 35

da these descriptions imply that a propagating crack sidual stress component rrs y , which acts perpen-
C(DK)m (Eq 14)
dN itself is surrounded by a typical residual stress dicular to the crack flanks, shows maximum
field as shown in Fig. 25(a) for a high-strength compressive residual stresses of approximately
where DK is the stress intensity range and C and structural steel of the European grade S690QL1 350 MPa at the crack tip. The alteration of this
m are constants. Mean stresses are taken into ac- (Ref 40, 41). A crack was produced by cyclic distribution after application of 20 overload cy-
count by the Foreman equation: loading up to a stress intensity range of DKI cles with an overload ratio k 2 (DKI 94.8
47.4 MPa m. The distribution of the macro re- MPa m) is given in Fig. 25(b). In front of the
da C(DK)m crack tip, a larger maximum value and a larger
(Eq 15) area with compressive residual stresses com-
dN (1 R)Kc DK
pared with Fig. 25(a) are developed. The influ-

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


103
ence of different overload cycles on the crack
where R is the stress ratio and Kc is the stress propagation rate is shown in Fig. 26 for k 2
intensity factor for plane stress. Both relation- and 3. For both overload ratios, a delayed retar-
ships are valid for crack propagation in materials dation of crack propagation occurs that is more
states without macro residual stresses. However, 104 As received pronounced for k 3 than for k 2 due to the
effect of overload-induced compressive residual
stresses. Thus, by sufficiently high overloads,
200 105 Autofrettaged crack arrest can be produced (Ref 42).
If a crack propagates into a macro residual
Residual stress (y ), MPa

stress field, the crack propagation behavior can


0
106 be considerably influenced by magnitude and
rs

0 0.2 0.4 0.6 0.8 1.0 distribution of the residual stresses. This is
Ratio a/W shown in Fig. 27, in which crack propagation
200 rates within an autofrettaged tube of SAE 4337
Fig. 27 Crack propagation rate da/dN versus ratio a/
W in an autofrettaged and an untreated tube of steel are compared with those in the untreated
the SAE 4337 steel. Source: Ref 43 material (Ref 43). As a consequence of the au-
400
tofrettaging process, a triaxial residual stress
state is created with very high compressive re-
12 8 4 0 4 8 12
sidual stresses near the inner surface of the tube.
Distance from crack tip, mm 10
(a)
150 bar In agreement with Eq 14, loading with a constant
stress intensity range results in a constant crack
200 8
Crack propagation rate

Untreated 75 bar propagation rate in residual-stress-free tubes.


(da/dN), mm/cycle
Residual stress (y ), MPa

However, if the crack propagates through the


6 225 bar
0
compressive residual stress field at the inner sur-
rs

face of the autofrettaged tube, considerably


4
smaller crack propagation rates occur.
200 Crack velocity is influenced by work harden-
2 ing as well as by residual stresses in mechani-
300 bar
cally treated surface layers. This is demonstrated
0
400 0 1 2 3 in Fig. 28 for differently surface-rolled AISI 304
Crack length (a), mm stainless steel. The crack propagation rate da/
12 8 4 0 4 8 12 dN, which was determined by analyzing stria-
Distance from crack tip, mm Fig. 28 Fatigue crack propagation rates da/dN in an- tions on cracked surfaces, increases with increas-
(b) nealed and with different rolling pressures ing rolling pressure and is considerably
deep-rolled steel AISI 304. Stress amplitude ra 320 MPa,
Residual stress component rrsy versus distance surface residual stresses rrs  200 MPa (75 bar), 350 diminished compared with untreated specimens
Fig. 25 (Ref 29).
from crack tip of the steel S690QL1 after a MPa (150 bar), 400 MPa (225 bar), and 300 MPa (300
mode I base load of DK 47.4 MPa m (a) and an over- bar). Source: Ref 29 Finally, the influence of welding residual
load of DK 94.8 MPa m (b). Source: Ref 40, 41
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
Crack propagation rate (da/dN), mm/cycle

versus DK diagram (Ref 44). The crack propa-


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

103
102 gation rate in the heat-affected zone is consid-
erably lower in the as-welded state than in the
=2 stress-relieved state due to the distribution of
103 104 Stress relieved welding compressive residual stresses.
Evaluation of Experimental Results. For
104 =3 practical purposes, it is very important to know
105 As welded that crack propagation through residual stress
fields can be modeled quantitatively by intro-
105
ducing an effective stress intensity range DKeff.
106 This can be seen in Fig. 30, which shows crack
106 5 10 20 30 40 60 80 propagation rates da/dN in welded and un-
8 4 0 4 8 12 Stress intensity range (K), MPa m welded specimens of SAE 1019 steel (Ref 45).
Crack length (aol), mm
Exactly at the welding seam, tensile residual
Fig. 29 Crack propagation rate da/dN versus stress in-
Crack propagation rate da/dN versus crack tensity range DK in the heat-affected zone of
stresses of about 340 MPa exist that disappear
Fig. 26 in a distance of 10 mm. Figure 30 presents crack
length a0l for overloads with k 2 and 3. EN S355 steel in an as-welded condition and after a stress-
Source: Ref 40, 41 relief heat treatment. Source: Ref 44 propagation rates in the unwelded base material
36 / Effect of Materials and Processing

as a function of DK for different stress ratios R. with high tensile residual stresses, individual
Crack propagation rate (da/dN), mm/cycle

103
As expected, with increasing R values, higher curves for each R value used are observed. For
crack propagation rates are observed. It is im- R (pulsating compression), first increas-
portant to note that a mode I crack can only grow ing and then decreasing crack propagation rates
0.1 0 0.5 1.0 during that portion of loading cycle where the are observed indicating the influence of the ten-
104 0.2
crack is open. This portion is influenced by the sile residual stress field. However, also in this
R = 0.4
2.0 R value itself and the near-crack-tip residual case, the crack propagation behavior can be de-
stress distribution. The effective stress intensity scribed by a single, relatively narrow scatter
105 range can be determined quantitatively as the band if the influence of the welding residual
difference between the maximum stress intensity stresses on the DK values appearing at the crack
and the stress intensity where the crack opens: tip is taken into account (see Fig. 30d). In this
106 way, the crack propagation behavior in macro
6 10 20 30 40 50 60 DKeff Kmax Kop (Eq 16) residual stress fields can be quantitatively de-
Stress intensity range (K), MPa m scribed if crack propagation data of macro resid-
(a) If the influence of the residual stress contribution ual-stress-free materials are available and the re-
on Kop is known, DKeff can be estimated and used sidual stress distribution is known.
Crack propagation rate (da/dN), mm/cycle

to predict the fatigue crack growth. Thus, the


103 double logarithmical plot of the crack propaga-
tion rate da/dN versus the effective stress inten- Influence of Residual Stresses on
R=0 sity range DKeff in Fig. 30(c) shows that all mea- S-N Curves
0.5
104 1.0 sured data points form a narrow scatter band. In
welded specimens, the (da/dN)/DK relations are
completely changed (e.g., Fig. 30a and b). Ob-
viously, for small DK values the crack propa- Characteristic Examples
105 2.0
gation behavior is entirely controlled by the
welding residual stress state, which leads to Low-Strength Steel. Figure 31 shows S-N
identical crack propagation rates irrespective of curves for alternating bending of normalized
106 the R ratio. However, outside the volume areas SAE 1045 steel (German grade Ck 45) (Ref 47
6 10 20 30 40 50 60
49). The notched specimens had a stress-con-
Stress intensity range (K), MPa m centration factor kt 2.5. The stress gradient at
(b) the notch root dr/dz related to the maximum
Nominal stress amplitude (n,a), MPa

300 stress r (normalized stress gradient g):


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

103 Upcut milling


z0
1 dr
rs = 234 MPa g (Eq 17)
200 r dz
Annealed
rs = 0 MPa
104 was 5 mm1. All data are nominal stress ampli-
Downcut milling tudes and are valid for a failure probability of
100
rs = 242 MPa 50%. The bending fatigue strength was evalu-
105
kt = 2.5 ated at an ultimate number of cycles Nu 107.
=5
By downcut milling and upcut milling, surface
0 residual stresses of 242 and 234 MPa, respec-
104 105 106 107
tively, were generated. The corresponding S-N
106 Number of cycles to failure, Nf
6 10 20 30 40 50 60 curves are almost identical. A third batch of
Effective stress intensity range (Keff), MPa m Alternating bending S-N curves of notched specimens was annealed 2 h at 700 C after
Fig. 31 downcut milling. The annealing results in a re-
specimens of normalized plain carbon SAE
(c) 1045 steel after annealing, downcut milling and upcut mill- duction of bending fatigue life and bending fa-
ing. Source: Ref 4749 tigue strength (Ref 4749).
Crack propagation rate (da/dN), mm/cycle

103
The alternating bending fatigue strengths of
milled smooth and notched specimens with dif-
Bending fatigue strength (Rf), MPa

300 ferent geometries are plotted in Fig. 32 as a func-


=1 = 0.4 tion of the surface residual stresses (Ref 4652).
104 = 0.1 Again, all data are given for a failure probability
kt = 1.7, = 2
200
of 50%, and the bending fatigue strengths are
nominal stress amplitudes at Nu 107. With
105 kt = 2.5, = 5 increasing stress-concentration factor and de-
creasing stress gradient, the bending fatigue
100 kt = 4.4, = 15
kt = 2.5, = 2 strength decreases. The influence of the stress-
106 Smooth concentration factor is clearly visible from a
6 10 20 30 40 50 60 Notched comparison of the specimens with the same
Effective stress intensity range (Keff), MPa m 0 value g 2 mm1, but different values kt
750 500 250 0 250 500 750
(d) rs
1.7 and 2.5, respectively. On the other hand, the
Surface residual stress (l ), MPa increase of g from 2 to 5 mm1 at specimens
with kt 2.5 results in a significant increase of
Fig. 30 Crack propagation rate da/dN versus stress in- Fig. 32 Alternating bending fatigue strength of milled
tensity range DK in the base material (a) and (c) smooth and notched specimens of normalized
bending fatigue strength. It is also interesting to
and versus effective stress intensity range DKeff (b) and (d) plain carbon SAE 1045 steel versus surface residual stress. note that specimens with kt 4.4, g 15 have
in the welded zone of the SAE 1019 steel. Source: Ref 45 Source: Ref 4752 a somewhat higher strength than specimens with
Stability of Residual Stresses / 65

conventionally peened samples relax for ra,s (70 or 555 F), the stability of the residual MPa (50 ksi). After N 1000, the |rrs| value is
700 MPa (100 ksi). With increasing stress am- stresses was investigated in detail (Ref 47, 48). 150 MPa (20 ksi). The depth where the residual
plitude, an increasing relaxation is observed. The depth distribution of the residual stresses stresses change their sign is not significantly af-
The reduction of the half-widths and, hence, of and the half-widths in the region close to the fected. As shown in Fig. 24(a), a shot peening
the micro residual stresses is much more pro- surface was determined at different numbers of treatment at 290 C (555 F) considerably delays
nounced for the samples peened at room tem- cycles at a surface stress amplitude 1000 MPa the relaxation of residual stresses. During the
perature compared to the other peening variant. (145 ksi). The curves for the conventionally first cycle, the initial absolute value of the resid-
For the warm-peened states, the measured val- peened samples in Fig. 24(a) show that the com- ual stress at the surface, |ra,s| 660 MPa (95
ues after cycling are often a bit higher than the pressive residual stresses in the region close to ksi), is reduced only by about 140 MPa (20 ksi).
initial values. Only for ra,s 900 MPa (130 the surface are strongly reduced with increasing After N 1000, |ra,s| 380 MPa (55 ksi) is
ksi) and N 104 can a significant decrease be number of cycles. During the first cycle, the ab- measured. The affiliated depth distributions of
seen. solute value of the residual stress of the surface the half-widths of the conventionally peened
For the same samples peened at 20 or 290 C decreases from about 600 MPa (90 ksi) to 350 condition in Fig. 25(a) verify that with increas-
ing number of cycles, reductions of the half-
widths and consequently of the micro residual
stresses occur due to larger rearrangements of
mobile dislocations in the surface layers. In op-
position to this finding, under the same loading
1.0 conditions the warm-peened condition shows
practically no relaxation of the micro residual
stresses. Obviously, the dislocation arrange-
ments in the warm-peened condition under cy-
a,s, MPa clic loading are considerably more stable than in
0.9
300
the conventionally peened condition. This is due
to dynamic and static strain aging effects that
400 occur during and after warm peening (Ref 47,
HWs/HWs,0

48). Dynamic strain aging results in a more dif-


0.8 500
fuse and stable dislocation structure caused by
600 pinning of dislocations by solute carbon atoms
and formation of extremely fine carbides.
700
The effect of stabilized residual stresses on fa-
0.7 800 tigue behavior is illustrated in Fig. 26, with S-N
curves at alternating bending of the convention-
900
ally peened and two warm-peened conditions in
1000 comparison with the ground condition. The S-N
curves for a failure probability of P 50% were
0.6 determined in each case from approximately 30
1 10 102 103 104 105 106 107
tests after the arcsin P procedure (Ref 49). Shot
Number of cycles, N
(a) peening at room temperature increases the fa-
tigue 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, peen-
1.0 ing 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
400 Resistance to Residual Stress Relaxation
500 The cyclic relaxation of macro residual
stresses can be divided into four phases (Ref 16,
HWs/HWs,0

600
17, 50):
0.8 700
Quasi-static relaxation during the first half of
800 the first cycle, which is caused by quasi-static
deformation processes
900 Quasi-static relaxation during the second half
0.7 of the first cycle, which is influenced by the
1000
load reversal
Cyclic relaxation at 1 N Ni (Ni num-
ber of cycles to crack initiation) due to cyclic
0.6 deformation processes, which usually leads to
1 10 102 103 104 105 106 107 a linear reduction of the residual stresses as a
Number of cycles, N function of the logarithm of the number of
(b) cycles according to Eq 21 due to cyclic creep
Cyclic relaxation at Ni N Nf (Nf num-
Fig. 23 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 shot- ber of cycles to failure) for sufficiently high
peened condition (Tpeen 20 C, or 70 F). (b) Warm-peened condition (Tpeen 290 C, or 555 F) stress amplitudes at the surface, which is
38 / Effect of Materials and Processing

fluence on finite fatigue life. Figure 37(b) com- grinding. Again, in the range of finite fatigue life, MPa) at the surface than the latter ones (221
pares S-N curves of notched specimens which the influence of the different manufacturing pro- MPa).
were milled, ground, and shot peened after cesses is almost negligible. The relative increase Figure 38 shows plots of the alternating bend-
of the bending fatigue strength by shot peening ing fatigue strengths evaluated from Fig. 37 as a
is more pronounced compared with smooth function of the surface residual stresses. The ar-
specimens. It is interesting to note that milled rows mark the shift in bending fatigue strengths
Bending fatigue strength (Rf), MPa

400
specimens have a higher bending fatigue and surface residual stresses produced by shot
strength than ground ones, even though they peening. Additionally, data points of ground
Smooth
= 0.4 have lower compressive residual stresses (159 specimens with negligible or tensile residual
300 stresses shown in Fig. 36 are included. In the
case of notched specimens, all data points lie on
200 a common line with the slope 0.154 except for
800

Bending fatigue strength (Rf), MPa


ground specimens with compressive residual
Notched
100
kt = 1.7 stresses at the surface. Regarding smooth spec-
=2 600 imens, the influence of tensile residual stresses
0 on the bending fatigue strength is much more
750 500 250 0 250 500 750 pronounced than the influence of compressive
rs 400 residual stresses.
Maximum residual stress (l ), MPa Notched Smooth
Figure 39 shows S-N curves for alternating
Fig. 38 Alternating bending fatigue strength of 200 bending of smooth specimens (g 1 mm1) of
quenched-and-tempered (600 C/2 h) plain
carbon SAE 1045 steel versus surface residual stress eval-
blank-hardened AISI 5115 steel (German grade
uated from Fig. 36 and 37 0
16 MnCr 5) determined in the unpeened and
0 1000 2000 3000 4000 5000 various shot peened states including one with
Deep-rolling force, N electrolytically removed surface layer (Ref 54
56). The corresponding depth distributions of re-
Fig. 41 Bending fatigue strength smooth and notched sidual stresses are given in Fig. 40. From the
specimens made from quenched-and-tem- comparison of both figures, it becomes clear that
pered SAE 5135 steel versus deep-rolling force. After Ref
57 the surface residual stress is not a suitable pa-
rameter for the assessment of the influence of the
various treatments on the fatigue behavior.
The influence of the deep-rolling force on the
Nominal stress amplitude (n,a), MPa

600 rotating bending fatigue strength of smooth and


notched specimens made from quenched-and-
500 rs = 400 MPa tempered SAE 5135 steel (German grade 37 CrS
4) is shown in Fig. 41 (Ref 57). In both cases, a
400
9 maximum of the fatigue strength occurs at a cer-
300 tain force. However, the increase of the fatigue
strength of notched specimens (kt 2) by deep
200 rolling is much more pronounced than that of
Fig. 39 Alternating bending S-N curves of smooth 923
specimens made from blank-hardened AISI smooth specimens. In the end, the optimal bend-
5115 steel in the as-blank-hardened and with different con- 100 kt = 1.7
ing fatigue strength of notched specimens
ditions shot peened states including one with a subse- =2
0 which is a nominal stress amplitudeis higher
quently electropolished surface. 1, as-blank-hardened; 2, 104 105 106 107
shot velocity v 23 m/s, coverage c 100%, mean di- than the bending fatigue strength of smooth
Number of cycles to failure, Nf specimens.
ameter 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, High-Strength Steels. S-N curves for alter-
v 81 m/s, c 600%, d 0.6 mm; 6, v 53 m/s, c Fig. 42 Alternating bending S-N curves of notched
specimens made from quenched plain carbon nating bending of notched specimens of
100%, d 0.6 mm, 100 lm surface layer electrolyti-
cally removed. After Ref 5456 SAE 1045 steel after different grinding processes. Source: quenched SAE 1045 steel in differently ground
Ref 4749 conditions are compared in Fig. 42 (Ref 4749).
The grinding parameters (final feed, cutting
speed) and the resulting depth distributions of
400
residual stresses are given in Fig. 43. Similar to
1000
1 ground quenched-and-tempered specimens, the
Residual stress (l ), MPa

200
Residual stress (l ), MPa

2 15 m, 30 m/s fatigue behavior of specimens with very small or


0 compressive residual stresses (which have a very
rs

500 small penetration depth) hardly differ. Tensile


rs

200 3
2 9 m, 15 m/s
4 residual stresses, however, not only cause a
400 5 strong reduction of bending fatigue strength, but
600 0 also of finite fatigue life.
800 In Fig. 44, the bending fatigue strength that
1000 2 3 m, 15 m/s was evaluated from Fig. 42 and corresponding
0 0.1 0.2 0.3 0.4 0.5 data from tests on smooth specimens are plotted
500
Distance from surface, mm 0 0.1 0.2 as a function of the surface residual stresses. The
Distance from surface, mm negative influence of tensile residual stresses on
Fig. 40 Depth distribution of the residual stress in spec- the bending fatigue strength of smooth speci-
imens made from blank-hardened AISI 5115 Fig. 43 Depth distribution of the residual stresses in
steel in the as-blank-hardened (1) and in different condi- notched specimens made from quenched plain
mens is even more pronounced compared to
tions of the shot peened state (3), (4), and (5) corresponding carbon SAE 1045 steel and ground with the two steps of notched specimens. The influence of compres-
to Fig. 39. After Ref 5456 final feed and cutting speed indicated. Source: Ref 4749 sive residual stresses generated by grinding is
Residual Stresses and Fatigue Behavior / 39

much smaller than the influence of tensile fatigue life, which comes up to one and a half crease of finite fatigue life. Additionally, the S-
stresses. orders of magnitude at high stress amplitudes. N curve of milled specimen is included. Finite
In Fig. 45(a), the S-N curves of smooth spec- The S-N curves of notched specimens of the fatigue life and bending fatigue strength of these
imens in the ground state and after additional same steel state after grinding and after addi- specimens are lower compared to shot-peened
shot peening are compared (Ref 4649). Similar tional shot peening with shot of different hard- ones, even though they contain very high surface
to quenched-and-tempered specimens (see Fig. ness are shown in Fig. 45(b). Compared to compressive residual stresses of 1200 MPa.
37), shot peening produces a significant increase smooth specimens, shot peening produces a In Fig. 46, the bending fatigue strength data
of the bending fatigue strength. Contrary to the much stronger increase of the bending fatigue already plotted in Fig. 44 are complemented by
results of the medium-strength steel, however, strength. Again, there is also a remarkable in- data evaluated from Fig. 45. The arrows mark
there is also a very pronounced increase of finite the shift in bending fatigue strengths and surface
residual stresses produced by shot peening.
Similar to the discussion of Fig. 39 and 40, it

Bending fatigue strength (Rf), MPa


800
Smooth
becomes obvious that the magnitude of surface
Bending fatigue strength (Rf), MPa

800 residual stress is not a suitable parameter for the


= 0.4
Smooth 600 assessment of the influence of shot-peening-in-
600 = 0.4 duced residual stresses on the fatigue strength.
400
400 Evaluation of Experimental Results
Notched Ground Notched
200
kt = 1.7 Shot peened kt = 1.7
200 Low-Strength Steels. Regarding Fig. 31 to
=2 Milled =2
0 33, different machining processes or process pa-
0
1000 500 0 500 1000 rameters were applied to the specimens investi-
rs
1000 500 0 500 1000 1500 Surface residual stress (l ), MPa gated producing different surface topographies.
rs
Surface residual stress (l ), MPa
In Fig. 47, the ratio DRf /DRh, which gives the
Fig. 46 Alternating bending fatigue strength of smooth change of the alternating bending fatigue
and notched specimens made from quenched strength of differently treated SAE 1045 and
Fig. 44 Alternating bending fatigue strength of ground plain carbon SAE 1045 steel with different surface condi-
smooth and notched specimens of quenched tions versus surface residual stress AISI 5115 steels per lm increase of roughness
plain carbon SAE 1045 steel versus surface residual stress. height is plotted as a function of the hardness at
Source: Ref 4749
the surface (Ref 58, 59). As is well known from
many other investigations, the susceptibility of
100
the bending fatigue strength to roughness in-
creases with increasing hardness. For the steel
Ratio Rf/Rh, MPa/m

SAE 1045
Nominal stress amplitude (n,a), MPa

1000 40 AISI 5115 states investigated, a value of 2 MPa/lm can


900 Shot peened serve as a guide. Since the roughness height of
10 the specimens did not vary by more than 2 lm,
800 the results plotted in Fig. 31 and 32 are hardly
influenced by surface topography. It should be
700 4
kept in mind, however, that a large increase of
Ground
600 surface roughness will significantly decrease the
1 fatigue strength even in low-strength material
500 200 400 600 800 states.
Surface hardness, HV 5 In low-strength steels, cyclic plasticity ap-
400
104 105 106 107 pears if the cyclic loading approaches the fatigue
Fig. 47 Changes of alternating bending fatigue strength strength. Extensive cyclic plasticity occurs in the
Number of cycles to failure, Nf DRf relative to changes of surface roughness
height DR h versus surface hardness HV 5 for different heat
range of finite fatigue life, being combined with
(a)
treated SAE 1045 and AISI 5115 steels. After Ref 58, 59 cyclic softening and/or hardening processes (see
the section Some Aspects of Fatigue of Steels
Nominal stress amplitude (n,a), MPa

1000 in this article). Cyclic plasticity results in macro


900 residual stress relaxation, which is more rapid
0
5458 HRC the higher the cyclic loading is (see the article
Surface residual stress (l ), MPa

800 n,a = 360 MPa Stability of Residual Stresses in this Hand-


200 book). Figure 48 (Ref 49) shows the relaxation
rs

700 4650 HRC Shot peened 325 of the surface residual stresses during alternating
600 bending fatigue in notched upcut milled speci-
Milled 400 280 men made from normalized SAE 1045 steel
500
Ground
200 starting at an initial value of 590 MPa. The
400
bending fatigue strength amounts to 190 MPa
600 (see Fig. 32). During loading with an amplitude
104 105 106 107 185
Number of cycles to failure, Nf of 185 MPa just below the fatigue limit, the
(b) 0 1 10 102 103 104 105 106 107 change of the macro residual stresses is almost
Number of cycles, N negligible. During all other fatigue loadings, ex-
Fig. 45 Alternating bending S-N curves of specimens tensive residual stress relaxation occurs. There-
made from quenched plain carbon SAE 1045 Fig. 48 Surface residual stress in notched (kt 1.7) fore, almost no influence of macro residual
steel. (a) Smooth specimens after grinding and after addi- upcut milled specimens made from normal-
tional shot peening. (b) Notched specimens after grinding, ized plain carbon SAE 1045 steel during alternating bend-
stresses is found in Fig. 31 and 32 (Ref 4752).
milling, and grinding with additional shot peening with ing fatigue at different nominal loading amplitudes versus The amount of cyclic plasticity, which occurs
shot of the indicated hardness. Source: Ref 4649 number of cycles. Source: Ref 49 during cyclic loading resulting in a given finite
40 / Effect of Materials and Processing

fatigue life or which corresponds to the fatigue cycles to crack initiation and decreasing the imen roughness, corrosion, oxidation) on the
limit, decreases with increasing strength of the propagation rate of short cracks (Ref 60, 61). crack initiation. Secondly, if crack initiation
steel. This results in decreasing residual stress Therefore, as is evident from Fig. 31, a residual occurs at the surface, the influence of the work-
relaxation rates and finally in stable macro resid- stress relief heat treatment that eliminates the hardened zone may also be rather small, be-
ual stresses. On the other hand, in a low-strength machining-induced work hardening reduces the causeat a given loadinga stress redistribu-
steel, local work hardening may occur during finite fatigue life and the fatigue limit of nor- tion may occur from the softer core region of the
manufacturing processes such as milling, turn- malized SAE 1045 steel. At rather high loading specimen to the work-hardened zone. This effect
ing, or grinding. In other words, the micro resid- amplitudes, this influence vanishes because then is combined with cyclic plasticity in the core re-
ual stress state is changed, mainly by the increase the cyclic deformation behavior is almost en- gion and, hence, with rather high loading ampli-
of the dislocation density. As discussed in the tirely governed by cyclic softening and harden- tudes. The first effect predominates in the range
article Stability of Residual Stresses in this ing processes during the fatigue loading (see the of the fatigue limit, where cyclic plasticity is
Handbook, the resistance of the micro residual section Cyclic Deformation Behavior in this small, and the second one in the range of finite
stress state against cyclic relaxation is much article) irrespective of initial variations of the lo- fatigue life. On the other hand, during rotating
higher compared to that of the macro residual cal work-hardening state. bending (Fig. 33b) with a distinct stress gradient,
stress state. So one may assume that macro re- A smaller influence of the micro residual crack initiation always occurs at the surface. If
sidual stresses that exist in a part of a component stress state appears in Fig. 33(a) and a somewhat the work-hardened zone is thick enough, there is
that has undergone work hardening by manufac- stronger one in Fig. 33(b). Here, it becomes clear no stress redistribution from the core region
turing processes are more stable than expected that even in a rather soft material state such as which is then purely elastically loadedto the
from the basic strength of the material. Looking normalized SAE 1015 steel, an improvement of work-hardened zone. Again, at very high loading
at Fig. 32, this relationship may be responsible the fatigue behavior can be obtained from local amplitudes, the cyclic softening/hardening pro-
for the small reduction of the bending fatigue work hardening. In this case, the depth of the cesses produced by the cyclic loading itself push
strength by tensile residual stresses that occurs surface layer of the smooth specimens influenced the influence of the initial work-hardening state
in some of the specimen series investigated. by deep rolling is relatively large compared to into background.
A much more important consequence of the the milled specimens, which are the basis of Fig. In Fig. 49, the ratio DRf /DHV, which gives
change of the micro residual stress state by local 31 and 32. Nevertheless, push-pull loading (Fig. the increase of alternating bending fatigue
work hardening is the direct alteration of the lo- 33a) without stress gradient reveals a small in- strength per unit hardness, is plotted as a func-
cal fatigue behavior. Increasing strength reduces fluence of deep rolling, because of two major tion of the hardness itself (Ref 58, 59). The di-
the local cyclic plastic deformation at a given effects. First, the crack initiation site may be agram shows that the sensitivity of the fatigue
external loading, thus increasing the number of shifted below the work-hardened zone. Then, the strength to the hardness decreases with increas-
small increase of the fatigue strength mirrors the ing hardness. Taking into account that the resis-
benefit of omitting surface effects (such as spec- tance of the macro residual stress state against
3 cyclic relaxation is low in low-strength steels, it
becomes clear from Fig. 49 thatbesides the
600 change of the surface topographythe altera-
Surface residual stress (l ), MPa
Ratio Rf/HV 5

Smooth tion of the micro residual stress state is the gov-


2 = 0.4
erning parameter for the assessment of the influ-
rs

n,a = 200 MPa


ence of machining on the fatigue behavior of
Rh = 3 m
400 components made from low-strength steel.
1 Notched Medium-Strength Steel. A detrimental effect
kt = 1.7, = 2 of tensile grinding residual stresses appears in
Rh = 15 m
200 n,a = 150 MPa Fig. 34 and 36 on alternating bending fatigue
0 strength, being more pronounced for smooth
0 200 400 600 800 1000 kt n,a = 255 MPa than for notched specimens. Therefore, the fa-
0
Surface hardness, HV 5 tigue notch factor decreases with increasing re-
0 1 10 102 103 104 105 106 107 sidual stress, as shown in Fig. 50. There is re-
Fig. 49 Changes of alternating bending fatigue strength Number of cycles, N
DRf relative to changes of surface hardness laxation of surface residual stresses during
DHV 5 versus surface hardness HV 5 for different surface (a) bending fatigue loadings of smooth and notched
roughnesses heights Rh. After Ref 58, 59 specimens, which results in lifetimes of approx-
Surface residual stress (l ), MPa

imately 2 million cycles, as shown by Fig. 51(a).


600 The stress relaxation in the notched specimens
Smooth
rs

2.5 = 0.4 is significantly more pronounced than in smooth


400 n,a = 400 MPa ones. An elastic estimation of the stress in the
notch root r* kt rn,a yields a much higher
Fatigue notch factor, kf

kt = 2.5 =2 value than the stress amplitude in the smooth


2.0
200 specimens. Even though r* is only an upper
Notched
kt = 1.7 n,a = 350 MPa bound for the true stress at the notch root, it be-
=5 =2 kt n,a = 595 MPa
comes clear that it is the higher local cyclic load-
1.5 0
ing in notched specimens that causes the stronger
kt = 1.7
0 1 10 102 103 104 105 106 cyclic stress relaxation and, hence, the smaller
=2 Number of cycles, N residual stress sensitivity of the bending fatigue
1.0 strength compared to smooth specimens. Re-
400 200 0 200 400 600 (b)
rs
garding the relatively high loadings indicated in
Surface residual stress (l ), MPa Surface residual stress in smooth and notched Fig. 51(b), stress relaxation in notched specimen
Fig. 51
specimens made from quenched-and-tem- is almost complete at the end of fatigue life, and
Fig. 50 Fatigue notch factor kf of specimens made from pered (600 C/2 h) plain carbon SAE 1045 steel versus
quenched-and-tempered (600 C/2 h) plain number of cycles (a) at a lifetime of approximately 2 106
therefore only a small influence of initial macro
carbon SAE 1045 steel versus surface residual stress. cycles and (b) at a lifetime of approximately 105 cycles. residual stresses on fatigue life at high loadings
Source: Ref 4749 Source: Ref 4749 is expected. This is shown in Fig. 34.
Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 71

of hydrogen in iron becomes unstable (Ref 17). ing elements. The energy of hydrogen interac- micron size zone of maximal three-axis stresses,
Thermodynamic data of H diffusion in iron at tion with different structural components varies where hydrogen dissolved in the metal lattice
high temperatures are well reproducible in vari- in a narrow interval (0.10.4 eV per atom). In diffuses and its concentration in steel at equilib-
ous works (Ref 1820) that cannot be told about particular, the one with dislocations in iron was rium conditions attains maximal values as:
the data obtained below 200 C (392 F). Dif- evaluated as about 1.6 1020 J (0.1 eV per
fusion mobility of hydrogen in iron at low tem- atom) (Ref 25).
peratures varies in a wide range and depends on The Role of Stresses. Hydrogen atoms interact C rmax C0 exp (1.9 103 ry) (Eq 10)
imperfection of the crystal lattice and the content with any anisotropic strain field, similar to its
of impurities. For example, Ref 2124 there interaction with dislocations. If concentration of
Therefore, in steel having the yield strength ry
were obtained values of the diffusion coefficient hydrogen atoms is insignificant, then distribution
of approximately 1725 MPa (250 ksi) the equi-
at 25 C (77 F) varying from 109 up to of the dissolved ones in a strain field of a crystal
librium concentration of hydrogen in the zone of
105cm2 /s. When studying diffusion in pure lattice can be found using:
maximal tensile stresses exceeds its content in
iron by intensity of hydrogen emission out of
volume by 26 times (Ref 28) (Fig. 3).
massive samples, it was found that its diffusion C C0 exp (U/kT) (Eq 7) At absence of a crack or any other concentra-
mobility is sharply depressed at lower than 200
tor, the stresses are localized in front of a dis-
C (392 F) (Fig. 2). Above 200 C (392 F) the where C is a concentration of hydrogen in a cer- location pile-up, where considerable enrichment
diffusion coefficient is the following: tain point of the field, C0 is an average concen- with hydrogen also can be observed. Both fac-
tration, k is the Boltzmann constant, T is the kel- tors, concentration of tensile stresses and dis-
D 1 4 103 exp (3200/ kT) cm2 /s (Eq 5) vin temperature, and U is the interaction energy solved hydrogen, initiate nucleation of cracks
of hydrogen atoms with the strain field in the and promote its propagation.
Below 200 C (392 F) that is: place of a defined concentration, C. Hydrogen Effect on Structure and Phase
Accounting for only hydrostatic components Transformations in Steel. The basic conse-
D 0.12 exp (7820/ kT) cm2 /s (Eq 6) of the field, the interaction energy, U, makes: quence of hydrogen presence in iron and steels
at solidification is the development of porosity,
Just as impurities forming interstitial solid so- U p dV (Eq 8) due to a gas-eutectic transformation (Ref 29)
lutions, hydrogen interacts with various crystal
similar to the well-investigated eutectic one. To-
defects, creating nonuniform solid solutions. where p is an average hydrostatic pressure, that tal amount of pores is proportional to those of
Presence of microdeformation near the crystal is, p 13(rxx ryy rzz), and dV is the lattice dissolved hydrogen and is inversely proportional
imperfections increases hydrogen solubility volume alteration under the influence of hydro- to the pressure of crystallization. Hydrogen in
there. Therefore, the solubility of hydrogen in gen. steel stabilizes austenite and decreases the criti-
steel rises with an increase in density of defects. From these equations it follows that at a con- cal point A3, the temperature of martensite trans-
When diffusing inside the crystal, hydrogen at- stant temperature the following condition is formation, and the critical rate of cooling for
oms are retained by the lattice imperfections, and valid:
its effective diffusion coefficient essentially de- quenching (Ref 29). This is the cause of marten-
creases. This results in sharply diminishing dif- site formation even in a carbon-free iron at stan-
fusion mobility of hydrogen in iron at low tem- ln (C/Co) A 13(rxx ryy rzz) (Eq 9) dard rates of cooling. Unlike a carbonic marten-
peratures. Defects retaining hydrogen are known site, at low temperatures of tempering there
as traps. Its concentration controls an excessive where A kT/dV. arises a hydrogen-free martensite and a gaseous
hydrogen solubility, compared with the lattice According to (Eq 9), hydrogen in the crystal hydrogen escaping from a sample or accumulat-
one (i.e., in an ideal lattice). Moreover, the num- lattice is redistributed under stress influence. ing in micropores, creating high pressure and mi-
ber of places capturing hydrogen is inversely These deductions were confirmed with experi- crocracks like floccules (Ref 29).
proportional to the value of the effective diffu- ments (Ref 26, 27). It is settled that the effective States of Hydrogen in Steel. Hydrogen ab-
sion coefficient. diffusion coefficient of hydrogen essentially de- sorbed by metal can be present in various states:
The role of traps can be played by edge dis- creases, and its solubility increases under load-
ing the metal in a macro-elastic interval (below An interstitial solid solution
locations, boundaries of mosaic blocks or grains,
single pores uniformly distributed in crystals, in- yield strength). Thus, stresses promote nuclea- A segregating one on imperfections of the
clusions of a second phase, and atoms of alloy- tion of new reversible hydrogen traps, disap- crystalline structure
pearing after elimination of an external stress. Adsorbing one on the surface of micropores
Similar sites arise near the crack tip, in a several and particles of second phases
T, C
800 400 200 100 25
107

1 1
107 2
D, cm2/s

107 H
H
H H
HH H
H2 HH H2 H
H H H
107 H2 HH H2 H H
H
H2 HH H2 H
H H
H
107 H H
0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 H
H
1000/T, K1

Fig. 2 Diffusions of hydrogen in pure annealing iron.


Source: Ref 1825 Fig. 3 Influence of hydrogen on crack propagation. Source Ref 28
42 / Effect of Materials and Processing

almost compensated (state 4) or overbalanced crack propagation, resulting in crack arrest at or stop the propagation of microcracks existing
(state 5) by a slower rate of crack propagation stress amplitudes increasing with the deep-roll- after crack initiation.
due to the larger penetration depth andin the ing force. At relatively high forces, however, the The extrapolated bending fatigue strength ver-
case of state 5the higher amount of the com- bending fatigue strength decreases, because then sus surface residual stresses curves in Fig. 44
pressive residual stresses. Since the surface the impairment of the surface zone by the strong cross in the range of tensile residual stresses at
roughness, the surface work-hardening state and increase of the surface roughness and micro- 1125 MPa. This means that the fatigue notch
the surface residual stress state (see Fig. 40) of cracking as well as an unfavorable residual stress factor becomes unity. On the other hand, at neg-
the material states 2 to 5 do not differ much, they state at the surface enhances the formation of ligible or compressive residual stresses, the fa-
have similar resistance to crack initiation at the rather long initial cracks that only arrest at re- tigue notch factor approaches the stress-concen-
surface. Hence, the different fatigue strength ap- duced stress amplitudes. The bending fatigue tration factor, as shown in Fig. 54. Qualitatively
pearing in Fig. 39 is determined by the apper- strength of smooth specimens increases only similar to the quenched-and-tempered state, the
taining resistance to microcrack propagation, weakly with increasing rolling force. Regarding influence of tensile residual stresses is stronger
which, in turn, depends on the depth distribution the maximum value, an increase of only 15% is than that of compressive residual stresses and the
of the compressive residual stresses. This is dis- achieved, compared to 120% in the case of residual stress sensitivity of notched specimens
cussed in more detail in the section Propagating notched specimens. Since the stress gradient is is smaller than that of smooth ones. However,
and Nonpropagating Cracks in this article. rather small in smooth specimens, there is some this cannot be explained by different residual
A very interesting example of the interaction speculation (Ref 56, 57) that the crack initiation stress relaxation so far. Careful x-ray analyses
of changing surface roughness as well as micro site is shifted below the surface by deep rolling. prove that in notched specimens loaded with a
and macro residual stress state in a medium- Then, the small increase of the fatigue strength stress amplitude of 450 MPa and in smooth spec-
strength steel are given in Fig. 41. The strong mirrors the benefit of omitting surface effects imens loaded with 750 MPa (corresponding to
increase of the rotating bending fatigue strength (such as specimen roughness, corrosion, oxida- the fatigue limit of each) the relaxation of com-
of notched specimens by deep rolling is clearly tion) and the effect of a decreasing local stress pressive residual stress is small and of equal
attributed to the generation of micro- and macro- amplitude on the crack initiation. This mecha- amount. This is not astonishing, since an elastic
stresses with a rather large penetration depth. It nism is discussed in detail in the section Haigh estimation of the stress amplitude in the notch
is interesting to note that the resistance to micro- Diagram in this article. At high rolling forces, root (r* kt r 765 MPa) yields almost the
crack initiation is hardly influenced by the deep- however, the crack initiation occurs at the sur- same value as the stress amplitude of the smooth
rolling treatment (Ref 57), and the increasing face again, where it is enhanced by a high rough- specimen. No corresponding measurements for
bending fatigue strength is almost entirely com- ness, an increased roughness sensitivity due to smooth and notched specimens with tensile re-
bined with an increasing resistance to micro- work hardening, and an unfavorable residual sidual stresses were performed. However, since
stress state. Therefore, the bending fatigue the fatigue limit is reduced by tensile residual
strength decreases, and there is no improvement stresses, it is clear that their relaxation is negli-
2.5 by deep rolling at all. In the end, the maximum gible.
bending fatigue strength of the notched speci- A comparison of Fig. 45(a) with Fig. 37(a)
mens (remember, evaluated as nominal stress shows that the shot peening treatments per-
Fatigue notch factor, kf

amplitude) is higher than the maximum bending formed increases the bending fatigue strength of
2.0 fatigue strength of the smooth specimens. These smooth specimens in the quenched state less
kt = 1.7 results show that the gradients of the micro and than in the quenched-and-tempered state. In
=2 the macro residual stresses in comparison to the view of the different residual stress states and
1.5 gradient of the loading stresses are of great im- their different stabilities in both heat treating
portance, and this will become even more evi- states, this finding is rather astonishing. In Fig.
dent in the next section. 55, the depth distribution of residual stresses in
1.0 High-Strength Steels. The S-N curves of shot peened quenched specimens before fatigue
500 0 500 1000 quenched SAE 1045 steel in Fig. 42 prove that loading and after 107 cycles at the fatigue limit
rs there is a detrimental effect of tensile grinding is shown. In the range of specimen scatter, the
Surface residual stress (l ), MPa
residual stresses on bending fatigue strength and, results are identical. Contrary to quenched-and-
Fig. 54 Fatigue notch factor kf of specimens made from contrary to the quenched-and-tempered state (see tempered specimen (see Fig. 52), the shot peen-
quenched plain carbon SAE 1045 steel versus Fig. 34) also on finite fatigue life. This is because ing residual stresses are much higher and much
surface residual stress. Source: Ref 4749
there is only a weak relaxation of the initial re- more stable in quenched specimens. However,
sidual stresses even at rather high stress ampli- in the former, crack initiation occurs at the sur-
tudes. Similar to quenched-and-tempered speci- face. Hence, the shot-peening-induced changes
0 mens, compressive grinding residual stresses of the micro and the macro residual stress state
Before loading with very small penetration depth (
20 lm, see have a strong influence on the measured bending
Residual stress (l ), MPa

Fig. 43) have almost no effect on bending fatigue fatigue strength. In smooth quenched specimens,
500 n,a = 750 MPa strength and finite fatigue life. Therefore and cyclic loading at the fatigue limit initiates cracks
rs

N 107 since in the quenched state the compressive re- below the surface. Therefore, the influence of the
sidual stresses are retained to the largest part dur- shot-peening-induced changes of the residual
1000 ing fatigue loading in the range of the fatigue stress state have a rather small influence on the
limit, one may expect that crack initiation occurs measured bending fatigue strength. This is
below the surface. Analogous to the quenched- treated more detailed in the section Modeling
1500 and-tempered state, this could not be proved def- the Influence of Residual Stresses on Fatigue Be-
0 0.1 0.2 0.3 0.4 0.5 initely (Ref 49). Since in the quenched state a havior. Contrarily, in the range of finite fatigue
Distance from surface, mm grinding-induced work softening occurs close to life, crack initiation in both states occurs at the
the surface and an influence of the surface rough- surface, and there is a strong influence of shot
Fig. 55 Residual stress of smooth ground and addi- ness always exists, it is very probable that crack peening regarding quenched specimens and a
tional shot peened specimens made from
quenched plain carbon SAE 1045 steel before loading and
initiation occurs at the surface. Then, one must very small one regarding quenched-and-tem-
after 107 cycles versus distance from surface. Source: Ref conclude that compressive residual stresses with pered specimens. This finding is directly corre-
4649 a penetration depth of only 20 lm do not retard lated with very different residual stress relaxa-
Residual Stresses and Fatigue Behavior / 43

tion. While surface residual stress relaxation is duces a strong increase of both fatigue limit and after
107 cycles. In this case, a significant re-
definitely complete in quenched-and-tempered finite fatigue life. Milling, which may generate laxation and redistribution of the residual
specimens after 50,000 cycles at a stress ampli- rather high compressive residual stresses, is less stresses occur.
tude of 570 MPa, there is hardly any change of effective. In all cases, crack initiation was ob-
the surface residual stresses in the quenched served at the surface. However, as evident from
specimens after the same number of cycles even Fig. 46, the surface residual stress state is not an Modeling the Influence of Residual
at 1200 MPa (Ref 49), although there may be appropriate parameter to assess bending fatigue Stresses on Fatigue Behavior
some residual stress relaxation below the sur- life in the presence of compressive residual
face. stresses. In Fig. 56, the bending fatigue strength
Regarding notched specimens, a different pic- is plotted as a function of the maximum residual Haigh Diagram
ture appears (compare Fig. 45b with Fig. 37b). stresses. Again, the arrows mark the shift of the
Now, shot peening of quenched specimens pro- residual stresses and bending fatigue strength Basic Relationships. One obvious way to ac-
obtained from shot peening. Now, the data of count for the influence of (macro) residual
ground notched specimens with tensile or neg- stresses on the fatigue behavior is to treat them
ligible residual stresses and of notched speci- as local mean stresses. In doing so, one has to
Bending fatigue strength (Rf), MPa

800
mens peened with shot of different hardness are realize that there are several important differ-
Smooth described satisfactorily by one common line ences between (loading) mean stresses and re-
600 = 0.4 covering a residual stress range of 2350 MPa. sidual stresses, as discussed in the section Com-
The data points of ground and those of milled parison between Loading Stresses and Residual
400
specimens bearing compressive residual stresses Stresses in this article. Figure 58 shows a Haigh
fall below this line, which has the slope 0.214. diagram for smooth and notched specimens
Ground In Fig. 56, it becomes even more clear than in made from a medium-strength steel (Ref 1, 62).
200 Notched
Shot peened kt = 1.7 Fig. 46 that the improvement of the bending fa- The Goodman approximation (see Eq 9) is used
Milled =2 tigue strength of smooth specimens by shot to account for the influence of residual stresses
0 peening is very limited. As a consequence, the on the fatigue strength. If the amount of the min-
1500 1000 500 0 500 1000 notch effect on bending fatigue strength is com- imum stress or the maximum stress in smooth
rs
Maximum residual stress (max ), MPa pletely eliminated at high compressive residual specimens does not exceed the critical stress am-
stresses. plitude ra,crit, which is a function of the cyclic
Fig. 56 Alternating bending fatigue strength of smooth
and notched specimens made from quenched It is interesting to look at the stability of very yield strength (see the article Stability of Re-
plain carbon SAE 1045 steel with different surface condi- high compressive residual stresses. As already sidual Stresses in this Handbook), the residual
tions versus surface residual stress shown in Fig. 55, fatigue loading of smooth stresses do not relax, and the line AB gives the
specimens in the range of the fatigue limit does influence of the residual stress on the fatigue
not cause any significant change of the residual strength. Then, all combinations of residual
0 stress state. As shown in Fig. 57(a), the same stress and stress amplitude inside the shaded area
holds for notched specimens peened with shot of result in neither residual stress relaxation nor fa-
Residual stress (l ), MPa

Before loading relatively low hardness, which have a very simi- tigue failure. However, if the amount of the min-
lar depth distribution of the residual stresses. imum stress or the maximum stress exceeds the
500 n,a = 640 MPa
rs

Elastic calculation of the minimum stress 70 lm critical stress amplitude, it is assumed that the
N 107 below the notch root (at the maximum of the residual stresses relax to the value given by the
compressive stresses) yields the value 2035 points A and B, respectively, and the fatigue
1000 MPa. This stress value may occur more than 107 strength remains constant at the value given by
times without effecting any change of the resid- these points. In the case of notched specimens,
ual stress state. As shown in Fig. 57(b), notched the cyclic yield strength and the notch fatigue
1500 specimens peened with shot of higher hardness strength (both in terms of nominal stress ampli-
0 0.1 0.2 0.3 0.4 0.5 bear even higher compressive residual stresses. tudes) are less than the respective values of
Distance from surface, mm In this case, elastic calculation of the minimum smooth specimens. However, the ultimate tensile
(a)
stress in a depth of 100 lm yields the initial strength of notched specimens is larger than that
0 value 2475 MPa and the value 2130 MPa of smooth ones in such a material state because
Residual stress (l ), MPa

Before loading

500
rs

n,a = 760 MPa


1000
N 107

1500
0 0.1 0.2 0.3 0.4 0.5
Distance from surface, mm
(b)

Fig. 57 Residual stress of notched ground and addi-


tional 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: Fig. 58 Haigh diagram. Bending fatigue strength Rf of smooth and notched specimens made from a medium-strength
Ref 4649 steel versus residual stress. After Ref 1, 62
Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 77

high-strength steel Fe-0.3C-2Ni-Cr-Mn-Si (Ref cess of initiation and growth of a crack when tion, the microstress relaxation process com-
79). testing on a delayed fracture in a freshly hard- pletes in a time of no more than 10 to 20 h since
Some Specific Cases of Delayed Fracture. ened state is quite the same as that in the grain quenching. In this state a material still possesses
In the scientific literature, including the educa- boundaries. The typical initiating crack found in the sensitivity to delayed fracture. In the process
tional one, it was a widespread opinion that de- the stress concentrator top is shown in Fig. 9. of holding samples at room temperature, the evo-
layed fracture of steels can develop involving no The work has been methodically performed in lution (desorption) of hydrogen from them into
hydrogenthat is, under the effect of only struc- the following way. First, regularities of the ini- the atmosphere has been found to occur. The
ture factors (Ref 28, 54, 80). In particular, the tiation and crack growth have been studied thor- sensitivity to delayed fracture is just concur-
question about the nature of the enhanced sen- oughly by methods of mechanical tests in con- rently with termination of hydrogen desorption
sitivity to delayed fracture of steels in the so- junction with the acoustic emission. In (i.e., about 70 h after quenching has been
called newly quenched state is long discussed. particular, the effect of the deformation-rate in- elapsed). According to the data of the laser spec-
Cases of quenched products cracking, which are hibition of the crack growth has been estab- troscopy, the amount of desorbent hydrogen over
met in practice of thermal processing (Ref 52, lished, durations of the incubation period, si, un- the period of holding comes to approximately 2
54), have been associated with the development der different conditions of loading have been 106 at.% (Ref 84).
of delayed fracture of untempered martensite. determined and so on. Second, the simulation of The source of diffusive-mobile hydrogen in
Initiation and growth of crack provoking delayed the newly quenched state has been performed. the structure of the quenched steel is inner hy-
fracture occurs mostly over the boundaries of For this purpose, after approximately 70 h ex- drogen, whose amount at conventional process
primary austenitic grains. The short-term tem- posure at room temperature (to eliminate the sen- of melting makes up no less than 2 to 3 104
pering or resting of quenched samples at room sitivity to delayed fracture) quenched samples at.%. The lattice solubility of hydrogen increases
have been hydrogenated by the special proce- from approximately 107 at 20 C (68 F) to 4.4
temperature enhances resistance to delayed frac-
dure. 104 at.% at 860 C, (i.e., by about 3 orders
ture; therefore, its development calls for further
Subsequent tests have shown that all the char- of magnitude), as a result of heating to the aus-
hydrogenation. Within the span of more than 50
acteristics of delayed fracture of quenched tenization temperature. A reverse decrease of
years of investigations, alternate hypotheses for steelsthe threshold stress, rth, the duration of solubility in the process of the fast quenching
explaining this phenomenon have been pro- the incubation period, si, and so on, as well as results in a part of hydrogen being kept for some
posed. The structure hypothesis based on the the mechanism of fracture, are analogous with time in the oversaturated state in the -solid so-
prevailing role of residual microstresses inherent the case of delayed fracture caused by hydrogen. lution and has the increased diffusive mobility
in the untempered martensitic structure has re- Therefore, it is beyond the reason to consider and the property of localizing in the region of
ceived the most acceptance (Ref 79, 80). The that in the case of the untempered martensite maximal tensile stresses. High interior micro-
role of internal hydrogen, which is always pres- structure a certain different type of delayed frac- stresses inherent of the quenched steel structure
ent in steel as an inevitable impurity, has not ture that is not connected with the presence of make the process of the crack initiation and the
been discussed practically in the literature. hydrogen in steel takes place. The fact that the crack growth easier but are not the factors that
The phenomenon of delayed fracture of kinetics of relaxation of interior microstresses in control kinetic parameters of delayed fracture.
quenched steels has been studied in detail in Ref the process of resting at room temperature does Another interesting case is delayed fracture of
8183. Steels Fe-0.3C-Cr-Mn-Si and Fe-0.4C- not correspond to the kinetics of the change of high-strength maraging steels, which develops
Cr-Mn quenched in oil at 860 C (1580 F) have the threshold stress, rth, bears witness to this over the certain temperature interval of aging
been chosen as objects of investigation. The pro- also. Judging by the change of the interior fric- 400460 C (Ref 85). The development of de-
layed fracture in these steels requires no special
hydrogenation. Therefore, within a long period
of time this phenomenon was also connected
80
only with the peculiarities of the structure state
forming at indicated aging temperatures. Fur-
3 thermore, this problem is urgent because aging
60
in the indicated interval provides for obtaining
the maximum elastic limit of steel. However,
, %

40 owing to the low resistance to delayed fracture,


1 this mode of aging usually is not practiced. An
20 analysis of data in hand has shown that the key
question in explaining the phenomenon of de-
2 layed fracture of maraging steels is a question on
0 the role of the testing environment (air, vacuum)
2500 as well as the role of titanium (Ref 85).
3 The investigations have been carried out with
2000 steels Fe-16Ni-10W-Mo-Ti and Fe-18Ni-9Co-
5Mo-Ti (Ref 86, 87). The tests have been per-
B, MPa

1500
formed under static loading at the reduced rate
2
of e 2.8 105 s1. Dependencies of the me-
1 chanical properties of steel Fe-18Ni-9Co-5Mo-
1000 Ti on the aging temperature when testing in the
air and in vacuum are presented in Fig. 10. A
500 crevasse in curves of the change of mechanical
340 380 420 460 500 540 properties on tests in the air (curves 1,1) cor-
T, C responds to the aging temperature of 420 to 460
C. Plastic properties of steels in this range come
Fig. 9 Nucleating grain boundary crack revealed at the Fig. 10 Influence of testing media on delayed failure of
top of the stress concentrator on testings of steel Fe-18Ni-9Co-5Mo-Ti. 1,1 testings in the
practically to the zero level. An effect of delayed
quenched steel Fe-0.3C-Cr-Mn-Si for delayed failure. air, u 4045%; 2.2, the same, u 6570%; 3,3, fracture for steel Fe-16Ni-Mo-10W-Ti is ob-
1100. Source: Ref 81 vacuum testings, P 1.3 Pa. Source: Ref 86 served within a more narrow temperature inter-
Residual Stresses and Fatigue Behavior / 45

strength steel in the section Evaluation of Ex- pressive grinding residual stresses. The same in shot peened as well as deep-rolled conditions
perimental Results, there is a detrimental effect holds for notched specimens with kt 2.5, g in comparison with the untreated states. For each
of a very high surface roughness. The second 5. The data points of shot peened specimens fall steel, a unique curve is obtained for the three
arrow illustrates the correction, which is based above the expected relationship as a conse- material conditions. Hence, concerning these
on Fig. 47, for this effect. Compressive residual quence of pronounced work hardening, which steels the quantities that are included in the SWT
stresses produced by grinding show some relax- increases both fatigue strength and resistance of parameterthe maximum stress and the total
ation, although their initial value is inside of the the compressive residual stresses against relax- strain amplitude (see Eq 11)form a reliable
border lines. This finding is consistent with the ation. basis for the assessment of the influence of re-
consideration given so far, since work hardening From the relationships shown in Fig. 60, the sidual stresses on the fatigue life.
is almost negligible in this specimen. Moreover, following conclusions can be drawn for a me-
the data point corresponding to the relaxed com- dium-strength steel:
pressive grinding residual stresses still falls be- Concept of Local Fatigue Strength
Residual stress relaxation plays a significant
low the Goodman line. This finding was already
role for the influence of residual stresses on Basic Relationships. As shown in the previ-
attributed to the low penetration depth of the
fatigue behavior. ous section, in the paragraphs on applying the
grinding residual stresses.
Residual stress relaxation depends on the Haigh diagram to high-strength and medium-
The residual stress sensitivity of notched spec-
work hardening produced by the various strength steels, the use of the Haigh diagram for
imens with kt 1.7 is significantly smaller than
manufacturing processes. Not only the the assessment of the effect of residual stresses
that of smooth specimens, even though the con-
amount, but also the penetration depth of on the fatigue behavior has serious limitations.
sideration is based on the relaxed residual
work hardening is important. As a conse- Regarding, for example, smooth specimens
stresses. Therefore, the different residual stress
quence, even for one initial material state made from a high-strength steel, shot peening
sensitivity of the fatigue strength of smooth and
there are no unique borderlines for the onset produces much less improvement in fatigue
notched specimens, respectively, is not solely
of residual stress relaxation. strength than expected from the residual stress
based on different residual stress relaxation, and
If comparable work-hardening states (or mi- state. This finding is combined with subsurface
the third conclusion given previously for a high-
cro residual stress states) are concerned, the crack initiation. Regarding medium-strength
strength steel is also valid for a medium-strength
influence of relaxed residual stresses on the steel, difficulties arise because not only the
steel to some extent. From an extrapolation of
bending fatigue strength corresponds satisfac- (macro) residual stress state is changed by manu-
the Goodman line plotted, an ultimate tensile
torily to the Goodman approximation. This facturing processes, but also the micro residual
strength of 1133 MPa is obtained, which is sig-
means that the relaxed residual stress sensi- stress state. The latter influences the fatigue
nificantly higher than the ultimate tensile
tivity and the mean stress sensitivity of the strength at zero mean stress and the resistance of
strength of smooth specimens. Again, this find-
bending fatigue strength is almost the same. the residual stress state against relaxation. There-
ing results from the triaxial stress state in the
In this formalism, the different initial residual fore, it is appropriate to look for the local fatigue
interior of the notched specimens and is consis-
stress sensitivity of smooth and notched spec- strength, which depends on the local micro re-
tent with data measured on other medium-
imens, which causes a reduction of the fatigue sidual stress state, the local macro residual stress
strength steels (Ref 63, 64). Regarding material
notch factor with increasing residual stress, is state, and (as far as crack initiation at the surface
states with compressive residual stresses, the
related to both the different residual stress re- is concerned) on the topography of the surface
data point of milled specimens comes rather
laxation and the different stress states in of the component. Strictly speaking, the local re-
close to the Goodman line. No measurements
smooth and notched specimens via the dif- sistance against fatigue crack initiation is con-
were done concerning the stability of this resid-
ferent ultimate tensile strengths. cerned, as shown later.
ual stress state. Since the data point falls inside
of the border lines, the residual stress state The concept of the locally effective fatigue
should be stable. Contrarily, there is some relax- Damage Parameters strength, which has its origin in Ref 38 and 39,
ation of the compressive residual stresses in enables quantitative predictions of the effect of
ground specimens even though their initial value As discussed in the section Lifetime Behav- the depth distributions of residual stresses on the
falls inside of the border lines, too. Again, this ior, there are some so-called damage parameters locus of crack initiation as well as on the fatigue
is a consequence of different micro residual that correlate fatigue life with loading parame- strength. The basic assumption of the concept is
stress states. Similar to smooth specimens, the ters. The well-known Manson-Coffin relation- that a crack can only be initiated at or below the
data points fall below the Goodman line as a ship (Eq 5) is not applicable to material states surface if the local loading stress exceeds the lo-
result of the small penetration depth of the com- with residual stresses, because it does not ac- cal fatigue strength. Especially in the case of
count for mean stresses. On the other hand, the relatively hard materials, for example, hardened
Ostergren parameter (Eq 12) and the Smith-Wat- steels, this concept yields to a good estimation
son-Topper parameter (Eq 11) include the mean of the corresponding properties. For that pur-
1000
stress via the maximum stress. With the limita- pose, it is necessary to have a good knowledge
800 SAE 1045
of the depth distributions of the fatigue strength
SWT parameter, MPa

tions already outlined in the previous section,


600
one could try to use these parameters to assess in the residual-stress-free condition R0f (z) as well
the effect of residual stresses on fatigue life. as of the macro residual stress rrs(z) and the re-
400
AISI 304 sidual stress sensitivity m(z). The locally effec-
300 However, the application of the Ostergren pa-
rameter requires the knowledge of the local tive fatigue strength Rf(z) as a function of the
plastic strain amplitude, which in turn needs so- distance x from the surface is calculated by the
200
Untreated
phisticated modeling of the local cyclic elastic- relationship:
Shot peened
Deep rolled plastic deformation behavior. This seems not to
100 be a useful approach. However, the Smith-Wat- Rf(z) R0f (z) m(z) rrs(z) (Eq 18)
103 104 105 106 107 son-Topper parameter (SWT parameter) is easier
Number of cycles to failure, Nf to evaluate and was used in Ref 28 to account where the residual stress sensitivity m(z) ap-
for surface treated steels (normalized SAE 1045 proaches the mean stress sensitivity M of the
Fig. 61 Smith-Watson-Topper parameter versus num-
ber of cycles to failure for specimens made
and AISI 304). Figure 61 shows characteristic Goodman relationship (e.g., Eq 9) if the residual
from normalized SAE 1045 and AISI 304 steel in different plots of the SWT parameter versus the number stresses are stable. Then, the residual stress sen-
surface conditions. After Ref 28 of cycles of stress-controlled tests on both steels sitivity is determined approximately by m(z)
46 / Effect of Materials and Processing

R0f (z)/Rm(z) (Ref 39). However, if residual stress sitivity m again approaches the mean stress sen- measured depth distributions of the hardness
relaxation occurs, the residual stress sensitivity sitivity M as shown in the paragraphs applying (see e.g., Ref 65).
m is smaller than the mean stress sensitivity M, the Haigh diagram to medium-strength steels in Application to High-Strength Steels. In a
if the initial residual stress distribution is used the previous section. The depth distributions of high-strength steel, manufacturing processes
(Ref 39, 62). However, if the relaxed residual the tensile strength Rm(z) and of R0f (z) can be may generate very large (macro) residual
stress distribution is used, the residual stress sen- estimated from appropriate correlations with 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
Smooth Notched rather small. Therefore, the local fatigue strength
0 0 is almost entirely determined by the initial ma-
Residual stress, rs

Residual stress, rs
terial state and the depth of the (macro) residual
stresses. Figure 62(a) and (b) show depth distri-
butions of compressive residual stresses in
smooth and notched bending specimens pro-
duced 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
(a) Distance from surface, mm (b) Distance from surface, mm corresponding depth distributions of the loading
N f = 104 stress. It becomes clear from this figure that the
improvement of the fatigue strength for infinite
N f = 104 life of smooth specimens (low stress gradient)
a and Rf,loc

a and Rf,loc

by shot peening is small because crack initiation


occurs below the surface in a depth with vanish-
Nf =
ing residual stresses. Contrarily, with increasing
stress amplitude, the crack initiation site is
shifted toward the surface, and the shot-peening-
Local fatigue strength induced stresses influence crack initiation and
Nf =
Loading stress
early fatigue crack growth. Hence, there is a
0 Distance from surface, mm 0 Distance from surface, mm large influence of shot peening on the low-cycle
(c) (d) fatigue strength and a small influence on the
high-cycle fatigue strength. This is shown in Fig.
Fig. 62 Distribution of residual stress (a) and (b) and distribution of loading stress and local fatigue strength for loadings 45(a). Contrarily, crack initiation in notched
in the range of finite and infinite lifetime (c) and (d) for smooth (a) and (c) and notched (b) and (d) specimens specimens (high stress gradient) occurs always
made from a high-strength steel
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
Propagating crack
1200 Not propagating 1200 that fatigue life determining cracks initiate only
crack in places where the loading stress exceeds the
d = 0.3 mm local fatigue strength is verified by several ex-
a and Rf,loc, MPa

a and Rf,loc, MPa

d = 0.6 mm
1000 1000 perimental results (e.g., Ref 38, 39, 62). Figure
63 presents for two series of specimens of the
Rf Rf aforementioned investigations from Ref 38 and
800 Loading 800 39 on hardened and shot peened SAE 1045 steel
stresses graphs of the depth distribution of the local fa-
tigue strength (see Fig. 63a and b). Rf,loc can be
600 600 calculated from the compressive residual stress
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 distributions (Fig. 63c and d) under the assump-
Distance from surface, mm Distance from surface, mm tion of uniform hardness. The closed circles
(a) (b)
mark the initiation sites of damage-relevant
400 400 cracks for each load stress distribution, which
0 0 could be taken from SEM of the fracture surface
(see Fig. 24). As can be seen, the actual crack
l, MPa

l, MPa

400 400 initiation positions agree with the expectation af-


rs

rs

800 800 ter the concept of the local fatigue strength. For
high stress amplitudes (1150 MPa), the cracks
1200 1200 that determine fatigue life can initiate at the sur-
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 face. For lower stress amplitudes (1150 MPa),
Distance from surface, mm Distance from surface, mm the damage behavior is changed. Of course,
(c) (d) stress amplitudes between 1000 and 1100 MPa
are able to initiate cracks at the surface (open
Fig. 63 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
circles in Fig. 58a and b). However, these stress
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 amplitudes are too small to propagate the cracks
(d), respectively. After Ref 38, 39 into greater depth because of the high compres-
Effect of Residual Stress on Hydrogen Embrittlement and Stress Corrosion Cracking / 83

Those additions in steel have dissimilar influ- streaks. Sulfide streaks have S:Fe:Mn ratio of heat-hardenable steel (0.4% C, 1.0% Cr, 0.37%
ence on the resistance to different kinds of frac- 2:1:1, which comply with the formula (FeMn)S; Mo, Fe-bal) to hydrogen-sulfide cracking, and it
ture. If its content does not exceed an optimal their thickness is 10 to 15 nm, and they are has been shown that increasing its content from
value (Table 1), resistance to all kinds of brittle- mainly located at the boundaries of original aus- 0.009 to 0.06% lowers the threshold stress from
ness increases (Ref 143). At higher contents, the tenite grains (Ref 149). The sulfide-matrix inter- 518 MPa (75 ksi) to 140 MPa (20 ksi). Such an
resistance of steel to hydrogen is increased, but faces are sites for the accumulation of diffusion- effect of phosphorus is attributed, first, to the fact
it is reduced to other kinds of brittle fracture. mobile hydrogen atoms, which at subsequent that it activates the hydrogen absorption process
Carbide-non-forming elements affect steel re- stages cause hydrogen blistering in low-alloy and, second, that it changes the shape and dis-
sistance to a ductile, brittle, and intergranular ferritic-pearlitic steels due to hydrogen mole- tribution of carbides. Phosphorus or manganese
fracture. Addition of 0.1 % Si, Mn, Co or Al in cules formation from its ions and greater (than segregation can be formed at a pearlite band in
a solid solution of an improved engineering steel the yield strength) planar pressure. Under load- low-alloy steels (Ref 154). Such formation
reduces its fracture resistance; the transition tem- ing conditions the hydrogen also will accumulate causes structural changes, an increase of steel
perature, T50 rising by 5; and the impact strength at the sulfide-matrix boundaries and other sites occlusive capacity by 2 times, and a decrease of
at decreasing by 0.07MJ/m2. Nickel is an excep- where there is a high stress level and will cause the time for failure under load in a hydrogenating
tion; 0.1% of it makes T50 lower by 4 to 10. sulfide stress cracking, which is characteristic medium by 2 times.
Similarly, these elements influence hydrogen re- both to mild and alloy high-strength steels (Ref The adverse effect of phosphorus on steel re-
sistance of steel. Thus, manganese, aluminum, 150, 151). Shape modification of sulfide inclu- sistance to brittle fracture is comparable with the
silicon, and cobalt decrease the time for failure, sions from plate-like to globular lowers the sen- favorable effect of nickel and molybdenum:
st, and increase the loss of plasticity, Fw, under sitivity of steels to hydrogen cracking, since the Each 0.01% lowering of phosphorus content in
hydrogenation. Nickel increases steel resistance globules are less favorable to hydrogen accu- steel is equivalent, in its influence on the T50
to hydrogen. mulation than plates are. Calcium and rare-earth level, to steel alloying with 1% nickel or 0.1%
Thus, when these elements are added in a solid metals are suitable for shape modification of in- of molybdenum.
solution, the direct dependence between the hy- clusions (Ref 152). Antimony and Tin. The resistance of engi-
drogen and brittleness resistance characteristics, Thus, the nonmetallic inclusions, especially neering steels to brittle fracture depends, to a
T50, ac, is observed. sulfides, enhance accumulation of hydrogen in large extent, on its cleanliness with regard to an-
Impurities such as sulfur, phosphorus, anti- steels and increase the susceptibility of steels to timony and tin impurities. These impurities are
mony, tin, and other elements have great influ- hydrogen embrittlement. Therefore, in order to brought into steel in considerable quantities
ence on the resistance of engineering steels to lower the sensitivity of steels to hydrogen em- (0.010.03%) during melting when using scrap.
hydrogen embrittlement. Their action is mani- brittlement, it is necessary to decrease the sulfur Both elements increase the cold-shortness
fested via weakening of grain-boundary binding content in them and to modify the shape and type threshold and lower the impact strength level
(temper-embrittlement processes) and via non- of sulfide inclusions. (Ref 155). The degree of its influence may vary
metallic inclusions formation. According to ex- Sulfur, therefore, is a harmful impurity in with the steel alloying system to a great extent.
isting opinions (Ref 144), loaded grain junctions steels, lowering their resistance to fracture and In low-alloy steels with ferritic-pearlitic struc-
can be zones for the origination of three-axial hydrogen embrittlement. In mild steels with fer- ture, the addition of antimony, which is a more
stress state. These grain junctions are energeti- ritic-pearlitic structure, the sulfur, forming man- electro-positive noble element than iron, im-
cally preferable for the accumulation of hydro- ganese sulfides of plate-like shape (stringers), proves the corrosion resistance in hydrochloric
gen atoms having a low chemical potential enhances hydrogen blistering, which is mani- and sulfuric acids, assumed to be due to a film
value. It is assumed that hydrogen decreases the fested even at 0.001% S. Sulfide (hydrogen) containing antimony forming on the steel sur-
cohesive forces between iron atoms at grain stress cracking of such steels is sharply reduced face. The surface layer inhibits penetration of hy-
boundaries (Ref 145). Segregation of impurities with the decrease of sulfur content below drogen into steel (Ref 156).
(phosphorus, antimony, tin, etc.) at boundaries 0.007%. An effective way for prevention of blis- To a greater extent, antimony and tin lower
of original austenite grains results in recurring tering and hydrogen stress cracking is modifi- the steel resistance to brittle fracture: Each ad-
temper embrittlement, which further weakens in- cation of engineering steels with rare-earth ele- ditional 0.01% Sb or Sn raises the T50 by an av-
tergranular cohesive forces. These considera- ments (0.10.3%), especially with cerium erage of 18 C (32 F). Their adverse effect in
tions have been proved by direct experiments spheroidizing sulfide inclusions. this respect is comparable with the favorable in-
(Ref 146). The results indicate that hydrogen Phosphorus, Antimony, and Tin. The adverse fluence of nickel and molybdenum.
does not interact with impurities or segregations effect of these elements on steel properties is It is possible to assume that at an antimony
at grain boundaries (i.e., the grain-boundary em- well studied. Impurities enriching grain bound- content above 0.3%, the effect of lowering steel
brittlement actions of impurities and of hydrogen aries during tempering cause reversible temper resistance to brittle and ductile fracture will pre-
mutually supplement each other). brittleness. This phenomenon has a considerable vail over the effect of lowering the quantity of
Published data indicate that steel improving effect on the level of steel resistance to fracture. absorbed hydrogen. Hence, additional alloying
with impurities and gases cleanliness enables it Phosphorus in engineering steels greatly low- with approximately 0.20 to 0.30% antimony is
to boost its fracture resistance (Ref 147). The use ers brittle and ductile fracture. Phosphorus fa- advantageous for low-alloy steels to be used in
of metallurgical refining and remelting processes cilitates steel hydrogen embrittlement by en- hydrogen-sulfide media. The associated reduc-
greatly increases hydrogen-embrittlement resis- hancing temper brittleness development and tion in the level of toughness properties is insig-
tance due to the removal of nonmetallic inclu- establishing chemical and structural heteroge- nificant as compared with the improvement in
sions from the steel, especially sulfides and neity. the resistance to hydrogen embrittlement by 2
stringer oxide inclusions (Ref 148). Quantitative relationships have been found for times.
Improving the quality of heat-hardenable en- the embrittling effect of phosphorus, tin, and an- With respect to these considerations the fol-
gineering steels (unalloyed with molybdenum timony on steel: Each 0.01% of the impurities lowing conclusions can be drawn:
and tungsten) enables them to boost by 2 to 3 increases brittle fracture temperature, T50, by 20 Antimony and tin impurities lower the resis-
times their resistance to brittle and ductile frac- to 23 C; (3641 F), decreases crack-develop- tance of structural steels to brittle and ductile
ture. ment energy, Ap, by 10 J/cm2; and increases fracture: Each 0.01% increase in antimony or
Sulfur is present in steel mainly in the form of plasticity loss value during hydrogenation by tin raises T50 of heat-hardenable steels by 18
manganese sulfides. Nonmetallic inclusions of 8%. C (32 F), and each 0.01% increase in anti-
this type are very ductile and, hence, during roll- Reference 153 describes a study on the influ- mony raises T50 of low-alloy steels by 1 to 5
ing are rolled out in the form of elongated ence of phosphorus content on the resistance of C (1.89 F).
86 / Effect of Materials and Processing

146. M. Vornet, Mem. Sci. Rev. Met., Vol 74 150. J. Hirth, Corrosion, Vol 32 (No. 1), 1976, 154. H. Gondo and M. Iino, Nippon Steel Cor-
(No. 5), 1977, p 307316 p 3437 poration Studies on Materials for Steel
147. A.P. Guliaev, Clean Steel, Metallurgia, 151. Z. Szklarska-Smilowaska, Werkst. Kor- Pipes for Service in Wet Sour Gas Me-
Moscow, 1975, p 184 (in Russian) ros., Vol 32 (No. 11), 1982, p 478485 dium, Technical Report B2 (No. 11), Nip-
148. V.G. Starchak, A.B. Kuslitsky, U.A. 152. G.M. Itskovich, Deoxidation of Steel and pon Steel Corporation, 1979, p 71
Kliachko, and I.A. Tamarina, Metal. Ter- Modification of Nonmetallic Inclusions, 155. E. Stephenson, Metall. Trans. A, Vol 11
moobrabotka, No. 4, 1977, p 1113 (in Metallurgia, Moscow, 1981, p 296 (in (No. 3), 1980, p 517524
Russian) Russian) 156. S. Shichi, M. Takeshi, and D. Utaka, Jap-
149. A. Joshi, Corrosion, Vol 34 (No. 2), 1978, 153. E. Snape, Corrosion, Vol 24 (No. 9), 1968, anese Application kl 12A41 (C23Fg/02),
p 4752 p 261282 No. 54, 1979
Residual Stresses and Fatigue Behavior / 49

69. It becomes obvious that a crack, once initi- between 300 and 200 lm below the surface and tiate a crack. Again, similar to the quenched state
ated, can propagate. Regarding crack propaga- finally crack arrest. However, in this stage of (see Fig. 66), DKeff has a local maximum in a
tion toward the surface, the simple model pre- crack propagation the crack extension will be depth of 70 lm, and up to a depth of 150 lm the
dicts very slow crack growth in the range large compared to the depth of the compressive residual stress field increases significantly DKeff
residual stresses. Therefore, loading stress redis- as compared with the loading of a specimen
tribution will occur, as will relaxation of the without residual stresses (dotted line).
8 compressive residual stress field in the process The corresponding relationships for shot
With residual zone ahead of the crack front. The crack will peened notched specimens with the residual
stresses continue to propagate toward the surface causing stress distributions shown in Fig. 53 are given in
6
final failure. Then, the fatigue limit is correlated Fig. 71. The loading stress amplitude ra 310
Keff, MPa m

with the maximum stress amplitude that does not MPa which corresponds to the bending fatigue
4 Without residual stresses
cause crack initiation below the surface. There- strength, is considered. From Fig. 60 it becomes
fore, the improvement of the fatigue strength of clear that crack initiation at the notch root occurs
smooth specimens by shot peening is limited, as under such loading conditions because the cor-
2
shown in Fig. 45(a). However, Fig. 69 illustrates responding data point (open square) falls signifi-
Range of Kth,eff that the overall value of DKeff and, hence, the cantly above the Goodman relationship. On the
0 driving force for crack propagation is rather basis of the initial residual stress distribution,
0 0.05 0.1 0.15 0.2 0.25 0.3 small (compare, for example, Fig. 69 with Fig. there is no driving force for the propagation of
Distance from surface, mm 66 in view of the very different loading stress such a crack up to a distance of 170 lm from
amplitudes). Therefore, already in the transition the notch root. However, if the distribution of
Fig. 70 Range of the effective stress intensity factor
DKeff with and without consideration of the re- from infinite to finite fatigue lifewhere crack the relaxed residual stresses is considered, DKeff
sidual stresses versus distance from surface for ground initiation still occurs below the surface (see Fig. comes close to the threshold in depths ranging
notched specimens made from a quenched-and-tempered 63)there is a strong influence of shot peening from 30 to 200 lm. Therefore, the fatigue limit
(600 C/2 h) SAE 1045 steel at cyclic bending loading (kt on fatigue life, as shown also in Fig. 45(a). The corresponds to a loading condition where cracks
1.7, g 2)
influence of shot peening on finite fatigue life that initiate at the notch root may propagate or
increases with increasing stress amplitude, be- not.
cause the crack initiation site is shifted toward In the case of smooth shot peened specimens
20 the surface, as discussed in the paragraphs on with the residual stress distributions shown in
Without residual
stresses high-strength steel in the section Concept of Fig. 52, it is difficult to determine on the basis
15 Initial residual
Local Fatigue Strength. of the concept of the local fatigue strength
Keff, MPa m

stresses Application to a Medium-Strength Steel. In whether crack initiation occurs below or at the
Relaxed a medium-strength SAE 1045 steel, the threshold surface, if the loading approaches the fatigue
10
residual value Kth,eff ranges from 2.3 to 3.3 MPa m (Ref limit (see paragraph on medium-strength steel in
stresses 67). In the following, ground notched specimens the section Concept of Local Fatigue Strength
5 Range of Kth,eff (kt 1.7, g 2) with tensile residual stresses and Fig. 65b). Actually, crack initiation at the
(see Fig. 35, curve with open circles), are treated. surface is observed that is favored by the rather
0
The interaction of this initial residual stress state high roughness. Figure 72 gives the depth dis-
0 0.1 0.2 0.3 0.4 with the loading stress amplitude ra 143 tribution of DKeff for the loading stress amplitude
Distance from surface, mm MPawhich corresponds to the fatigue ra 414 MPa, which corresponds to the bend-
strengthresults in the depth distribution of ing fatigue limit. Regarding the initial residual
Fig. 71 Range of the effective stress intensity factor DKeff shown in Fig. 70. At a depth of 15 lm, stress state, the longest nonpropagating crack
DKeff without and with consideration of the ini- DKeff exceeds the range of threshold values. would have a length of 250 lm. Due to stress
tial and the relaxed, respectively, residual stresses versus
distance from surface for ground and additional shot
Therefore, a crack that initiates at the notch root relaxation, this value is reduced by 50%. Still,
peened notched specimens made from a quenched-and- will always propagate into the interior of the rather long cracks have to be formed during the
tempered (600 C/2 h) SAE 1045 steel at cyclic bending specimen, and the fatigue limit corresponds to initiation stage to be able to propagate into the
loading (kt 1.7, g 2) the maximum stress amplitude that does not ini- 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
Relaxed residual
stress amplitude at which crack arrest is ob-
Nominal stress amplitude (n,a), MPa

16 1200
stresses R m = 1650 MPa
Without residual
Final failure served may be very large in deep-rolled speci-
mens, which have rather small or even tensile
Keff, MPa m

12 stresses Transition range


1000 1400 MPa residual stresses at the notch root, but large com-
No final
8 failure
pressive residual stresses down to a relatively
Initial residual 800 1150 MPa great depth below the surface. Then, the length
stresses of an arrested crack will strongly depend not
4
only on the depth distribution of the residual
Range of Kth,eff 600 stresses, but also on the loading stress amplitude.
0 This is proved by Fig. 73 concerning notched
0 0.1 0.2 0.3 0.4 0.5
400 specimens (kt 2) made from differently heat
Distance from surface, mm
0 0.2 0.4 0.6 0.8 1.0 1.2 treated SAE 5135 steel with the ultimate tensile
Length of arrested crack, mm strength indicated (Ref 66). The diagram gives
Fig. 72 Rangewithout
of the effective stress intensity factor
and with consideration of the ini- the length of arrested cracks as a function of the
DKeff
tial and the relaxed, respectively, residual stresses versus Fig. 73 Influence of the nominal bending stress ampli- nominal stress amplitude. The higher the
distance from surface for ground and additional shot tude on the length of arrested cracks in deep-
peened smooth specimens made from a quenched-and- rolled notched specimens (kt 2) made from SAE 5135
strength of the material state, the less the crack
tempered (600 C/2 h) SAE 1045 steel at cyclic bending steel with the ultimate tensile strengths indicated. After Ref length increases with increasing stress amplitude
loading (g 0.4) 66 because the amount and the stability of the com-
50 / Effect of Materials and Processing

pressive residual stresses produced by deep roll- removed electrolytically, resulting in a shift of fatigue limit, but higher finite fatigue lives than
ing increase with increasing hardness. From this the maximum compressive residual stresses into the specimens corresponding to curve 6. Apply-
figure it becomes clear that the large increase of the surface. Taking the slope of the straight line ing the same procedure with the aid of Fig. 36
the notch fatigue strength visible in Fig. 41 is for smooth specimens, g 0.4, in the range of as above to the surface residual stress state, an
almost entirely based on an increase of the re- (initial) tensile residual stresses in Fig. 36 as a increase of the bending fatigue limit of 220 MPa
sistance against crack propagation produced by rough approximation, one would expect a dif- compared to ground specimens is expected. Ac-
deep rolling. ference of both specimen series of 316 MPa tually, a difference of 279 MPa is determined.
In view of the relationships presented so far, caused by the different residual stresses. On the This finding means that the bending fatigue limit
it is interesting to look again at Fig. 39 and 40. other hand, on the basis of the Goodman line in is not determined by crack initiation at the sur-
The lowest bending fatigue limit was determined Fig. 60, one would expect a much higher differ- face, but by the largest amplitude at which crack
with ground specimens having low tensile resid- ence of 429 MPa. This shows that stress relax- arrest below the surface occurs. These relation-
ual stresses at the surface (curve 1 in both fig- ation occurs in this medium-strength steel state, ships become more clear in Fig. 74. There, DKeff
ures). An increase of 324 MPa (curve 6 in Fig. as expected. Anyway, the bending fatigue limit for surface cracks developing in conditions 5 and
39) was achieved with specimens that were shot of both specimen series is determined by crack 6 at nominal stress amplitudes corresponding to
peened according to curve 4 in Fig. 40 and from initiation at the surface. The specimens corre- the bending fatigue strength are plotted as a
which a surface layer of 100 lm thickness was sponding to curve 5 in both figures have a lower function of the distance from the surface. Re-
garding the initial residual stress distribution,
cracks smaller than 320 lm (condition 5) or 100
lm (condition 6) would not be able to propagate.
30
Without residual stresses Actually, stress relaxation occurs, the extent of
which, however, is not given in Ref 54 to 56. If
n,a = 640 MPa
n,a = 600 MPa it is assumed that the residual stresses relax by
Initial residual
stresses 50%, one gets the curves for the relaxed residual
stress state in the figure. Now, one can see that
20
Keff, MPa m

a crack once initiated in condition 6 will con-


Condition 6 tinue propagation. Hence, the bending fatigue
n,a = 640 MPa strength of this condition is determined by the
Condition 5 resistance against crack initiation, as stated
Assumed n,a = 600 MPa
relaxed above. Contrarily, in condition 5 the bending fa-
10 residual tigue strength corresponds to the maximum
stresses nominal stress amplitude at which crack arrest
Range of Kth,eff below the surface occurs.
From these considerations, it becomes clear
that the depth distribution of relaxed residual
0
0 0.1 0.2 0.3 0.4 0.5 0.6 stresses must be taken into account to gain a re-
Distance from surface, mm
liable assessment of the influence of residual
stresses on the fatigue behavior of medium-
Fig. 74 Range of the effective stress intensity factor DKeff without and with consideration of the initial and the relaxed, strength steel. Frequently, these data are not
respectively, residual stresses versus distance from surface for (corresponding to Fig. 39) shot peened smooth known. Then, the use of initial tensile residual
specimens made from a blank-hardened AISI 5115 steel at cyclic bending loading
stresses for the dimensioning of components is
a conservative procedure, because their detri-
mental effect on the fatigue behavior of a com-
ponent will be reduced by residual stress relax-
ation. However, the beneficial effect of
compressive residual stresses will be overesti-
mated by using the initial values, and this results
in a nonconservative dimensioning. Therefore, it
rs ls is important to estimate the remaining residual
kt Smooth stresses conservatively, for example with the ar-
Fatigue strength, Rf

Rh ticle Stability of Residual Stresses in this


Handbook.
rs
micro
rs
ls
Summary and Recommendations
kt
Rh
Notched Notched Some important conclusions that can be
drawn from this article are illustrated by Fig. 75,
Smooth
Smooth
in which the fatigue strength Rf is plotted as a
rs
micro ls kt Rh function of the macro residual stress. The fol-
lowing parameters are regarded:
Notched
The material strength; a low-strength steel
(for example, a normalized steel, lower band
Residual stress ,rs in the figure), a medium-strength steel (for
Influence of the macro residual stress and some other parameters on the fatigue strength of smooth and notched
example, a quenched-and-tempered (at a me-
Fig. 75 dium temperature) steel, middle band in the
specimens made from a low-strength steel (lower band), a medium-strength steel (middle band), and a high-
strength steel (upper band) figure), and a high-strength steel (a quenched
Residual Stresses and Fatigue Behavior / 51

or quenched-and-tempered (at a low tempera- plitudes. However, the micro residual stress state (see Fig. 62a and c, 63, and 64c) and the im-
ture) steel, upper band in the figure) are con- may be changed significantly by the processes provement of the fatigue strength even by high
cerned that generate the macro residual stress. Hence, compressive residual stresses is limited (see Fig.
The depth distribution of the macro residual in a medium-strength steel the local resistance to 45a, 46, and 56). With increasing stress ampli-
stress rrs, characterized by its sign, magni- residual stress relaxation may be quite different tude and hence loading stress gradient, the crack
tude, and gradient grs even if the same initial material state is con- initiation site is shifted to the surface. Therefore,
The depth distribution of the micro residual cerned, and unique borderlines for the onset of the finite fatigue life is much more improved
stress rrs
micro residual stress relaxation do not exist (see Fig. than the fatigue strength (see Fig. 62a and c, 63,
The notch factor kt (smooth and mildly 60). A significant benefit from compressive re- and 45d). In notched specimens, the loading
notched specimens are regarded in the figure) sidual stresses is only obtained if their penetra- stress gradient is larger than in smooth ones. If
The gradient of the loading stress gls tion depth is sufficiently high and/or their gra- high compressive residual stresses with a suffi-
The surface topography, characterized by the dient is sufficiently low as compared to the cient depth and/or a small gradient exist, crack
roughness height Rh gradient of the loading stresses. Therefore, com- initiation will occur in the notch root resulting
pressive residual stresses produced by grinding in a strong improvement of the fatigue strength
do hardly influence the fatigue strength of (see Fig. 64a and b, 45b, 46, and 56). However,
Low-Strength Steel smooth and notched specimens (see Fig. 36). if the penetration depth of the compressive re-
Contrarily, compressive residual stresses gener- sidual stresses is low and therefore the gradient
In a low-strength steel, there will be no or very
ated by deep rolling or shot peening increase the of the local fatigue strength highfor example,
little influence of the macro residual stress, be-
notch fatigue strength (see Fig. 38, 39, and 41). after grindingthe improvement of the notch
cause it is relaxed more or less completely if the
The fatigue strength of smooth specimens is less fatigue strength will be small (see Fig. 44 and
cyclic loading approaches the fatigue strength
improved, if subsurface crack initiation occurs. 59). Frequently, after shot peening or deep roll-
(see Fig. 32 and 48). A change of the micro re-
Then, the fatigue strength of notched specimens ing, maximum compressive residual stresses oc-
sidual stress state by work hardening may sig-
may even be higher than corresponding values cur below the notch root. Then, the fatigue
nificantly increase Rf since the resistance against
of smooth specimens (see Fig. 41). This means strength may not be determined by the maximum
cyclic plastic deformation and hence, crack ini-
that the fatigue notch factor kf becomes less than cyclic loading, which does not result in crack
tiation increases (see Fig. 31). Then, also the re-
unity. At vanishing residual stresses, kf comes initiation, but by the maximum cyclic loading at
sistance against macro residual stress relaxation
rather close to kt if the loading stress gradient is which crack arrest below the surface is possible
is raised resulting in a certain sensitivity of the
small, but is significantly less at higher gls val- (see Fig. 67, 68, and 73). As a consequence of
work-hardened zone to macro residual stress.
ues. With increasing tensile residual stresses, the all these relationships, the fatigue notch factor
This may be detrimental or beneficial (see Fig.
fatigue notch factor is reduced again, as expected may vary strongly in the presence of compres-
33b) for Rf depending on the sign and the mag-
on basis of the Goodman relationship for smooth sive residual stresses regarding one-notch ge-
nitude of the macro residual stress. The influence
and notched specimens (see Fig. 58 and 60). ometry and may take values ranging from less
of the surface topography is rather small in a
There is a considerable influence of the surface than unity to the notch factor kt, as sketched in
low-strength steel (see Fig. 47). The fatigue
roughness on the fatigue strength, as shown in Fig. 75. In fact, regarding smooth and notched
notch factor kf is significantly smaller than the
Fig. 47. specimens, Fig. 44 and 54 prove that kf of ground
notch factor kt because cyclic plastic deforma-
specimens with compressive residual stresses
tion and stress redistribution occur in the notch
High-Strength Steel produced by grinding approaches kt. From Fig.
root (see Fig. 32). With increasing loading stress
56 it can be deduced that kf of shot peened spec-
gradient at a given kt, the fatigue strength in-
In high-strength steels, stress relaxation dur- imens approaches unity. At vanishing residual
creases due to the decrease of the highly stressed
ing cyclic loading in the range of the fatigue stresses, kf comes close to kt as expected in a
volume of the component or specimen (compare
strength only occurs in notched specimens bear- high-strength steel. With increasing tensile re-
kt 2.5, g 2 with kt 2.5, g 5 in Fig.
ing very high compressive residual stresses. sidual stress, the fatigue notch factor is reduced
32).
Then, the resulting fatigue strength is also high, and finally approaches unity (see Fig. 44 and 54).
and during corresponding cyclic loading very Again, this finding is in correspondence with the
Medium-Strength Steel high magnitudes of the minimum stress occur, Goodman relationship for smooth and notched
which leads to some residual stress relaxation specimens (see Fig. 58 and 59). The surface
In a medium-strength steel, there is a signifi- (see Fig. 57b). Contrarily, in the range of high roughness has principally a large influence on
cant influence of the macro residual stress on Rf tensile residual stresses and cyclic loadings that the fatigue strength of high-strength steel, as
since only a small part of rrs relaxes during cy- lead to infinite life or to technically relevant life- shown in Fig. 47. On the other hand, in practice
clic loading in the range of the fatigue limit (see times the occurring maximum stresses are much the roughness height of hard steel is rather low
Fig. 36, 38, 51a, and 52). However, in the low- lower and no residual stress relaxation is ob- even after mechanical surface treatments such as
cycle fatigue range, relaxation becomes more served even in the range of low-cycle fatigue. shot peening.
complete with increasing amplitude, and the in- Consequently, the residual stress sensitivity and
fluence of the macro residual stress vanishes (see the mean stress sensitivity of Rf are identical (see Recommendations
Fig. 34, 37, 39, 51b). Tensile residual stresses Fig. 59), and the fatigue strength is strongly re-
are always detrimental to Rf. Therefore, in the duced with increasing tensile residual stress. From all of these relationships, some recom-
presence of large tensile residual stresses a me- This is also true for the finite fatigue life (see mendations may be deduced. In medium- and
dium-strength steel may have equal or even Fig. 42). In the range of compressive residual high-strength steels tensile macro residual
lower fatigue strength than a low-strength steel stresses, complex relationships exist. A strong stresses must strictly be avoided since they al-
(compare Fig. 32 with 36). If relaxed tensile re- effect of rrs will only occur if cracks are initiated ways promote crack initiation and crack propa-
sidual stresses are concerned, the residual stress at the surface. However, in thick smooth speci- gation and are detrimental to the fatigue strength
sensitivity m of the fatigue strength approaches mens or components cyclically loaded in the andat least in higher strength steelto finite
the mean stress sensitivity M (see Fig. 60). In range of the fatigue strength, the loading stress fatigue life. In a low-strength steel, the influence
material states of similar hardness loaded in the gradient is usually lower than the gradient of the of tensile macro residual stresses is usually small
range of the fatigue limit, compressive residual local fatigue strength, which depends on the or negligible, and the change of the micro resid-
stresses relax stronger than tensile ones simply depth distribution of the residual stresses. Con- ual stress state by work hardening is much more
because of the different corresponding stress am- sequently, cracks are initiated below the surface important. In most cases, work hardening will be
Deflection Methods to Estimate Residual Stress / 91

carbon steel (e.g., electric arc discharge) may re- face during peening. Peening is continued until quench distortion is the Navy C-ring test (Ref 5
sult in a level of triaxial residual stress close to the material has reached saturation; that is, when 7). The amount of distortion (deflection) of a
the cohesive strength of the material, possibly the lift height increases by no significant amount quenched test piece is measured by the change
leading to immediate or delayed cracking. and the compressive stress in the surface layers in the gap width (Fig. 9).
Although the stress may be multidirectional, corresponds closely to the elastic limit of the
hoop stress that arises at several stages of man- steel strip (Fig. 6).
ufacture is the primary cause of many of the Consider the residual stress distribution Methods for Measuring Residual
problems. When a length of tube is parted off through the Almen strip. Deflection of the strip Stresses from Deflection Data
and slit in a longitudinal direction, any hoop occurs in order to relieve some of the high sur-
stress present tends to open the slit (Fig. 5). face compressive stress. The peened surface area Sectioning to allow relaxation of residual
Opening of the slit indicates compressive stress endeavors to expand by bowing in order to nul- stress in actual components where stresses are
in the inside diameter (ID) of the tube and tensile lify the compressive stress. The magnitude of thought to be present can be performed by sev-
stress on the outside diameter (OD). On rare oc- bowing is limited by the restraint of the remain- eral methods, from simple saw cutting to the
casions, the slit may close, indicating the reverse der of the material, and equilibrium is reached more sophisticated compliance method (Ref 8
condition. The stress is not always uniform along when the remaining compressive stress is in bal- 10). In the latter method, the residual stress pro-
the length of the tube. Stress measurements have, ance with the elastic compressive stress devel- file is calculated from the strains caused by in-
on occasions, indicated a cyclic variation corre- oped in the lower surface. The depth and inten- troducing a cut of progressively increasing depth
sponding to differences in cooling of areas in sity of the compressive layer is proportional to into the component. The strains are measured by
contact with the cooling bed cross ties. the lift height at saturation. The Almen strips are using suitably positioned strain gages cemented
Shot peening intensity is monitored by using primarily used to ensure that the process is in to the surface adjacent to the cut. The cutting is
a method developed by J.O. Almen of General control (shot condition, uniformity in applica- performed by using various techniques, how-
Motors Company. In this method, 75 mm long tion, impeller operation, etc.). A photograph of ever, electrical discharge machining is the pre-
by 18.75 mm wide (3.00 by 0.75 in.) strips of three typical strips is shown in Fig. 7. ferred method. Usually, two computer-based ap-
1070 spring steel are quenched and tempered to Metal Improvement Company, Inc., in col- proaches are used to analyze the data. These are
a deep-blue oxide finish (blue tempered) for a laboration with ENSAM, a French advanced en- the forward and inverse solution. The forward
hardness of 45 to 50 Rc and exposed on one side gineering school, has developed a software pro- solution derives the measurable strains (compli-
to the same shot intensity as the component un- gram called PeenStress (Metal Improvement ance functions) that develop from introducing a
dergoing treatment (Ref 3). Three different Co., Inc., Paramus, NJ) that is used to assist in successively deeper slot into a part containing an
thicknesses are used: N, 0.79 mm (0.031 in.); shot peening callouts (Ref 4). The user selects a arbitrary known stress distribution. The inverse
A, 1.30 mm (0.051in.); and C, 2.38 mm material from a library of about 80 materials, solution develops the original residual stress dis-
(0.0938 in.) to allow for differences in degree then selects a shot size and shot intensity and tribution that best matches the actually measured
and shot peening required and materials being inputs some basic geometry considerations. Fig- strains.
treated. For example, peen forming of an alu- ure 8 is a curve generated on a chromium-silicon Similar methods based on drilling small holes
minum alloy requires considerably less peen in- spring wire shot peened with a hardened, 0.023 in the stressed material have been around for
tensity than a carburized gear tooth. During treat- in. shot to a 10 A intensity. A 10 A intensity some years. Strain relaxation is measured using
ment, the Almen strip assumes a concave shape. equates to a 0.25 mm (0.010 in.) arc height on strain gages or photoelastic coatings. Strain
The lift height is proportional to the level of the A-strip. gages have been used to measure strain relaxa-
compressive stress developed in the upper sur- One of the older standard tests for evaluating tion in bevel gears following successive removal
of layers by electrochemical machining (Ref 11).
Saw cutting or slitting is an easy quality con-
trol test that gives a global overview of the state
of residual bulk hoop stress in rings. Interest-
Ar3 ingly, such a technique was used recently to val-
Ar1 idate the measurements of residual stress levels
in railway wheels using electromagnetic-
Austenite acoustic transducers (Ref 12).
Ferrite

Pearlite Mathematical Derivations for


Core Simple Cases Based on the Saw-Cut
Temperature

Methods
Bainite

Surface Definitions of the symbols used in the follow-


ing derivations are given in Table 1.
To develop the mathematical algorithms for
Martensite interpreting deflection measurements after slit-
ting 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
Time, T
(Eq 1)
I c R

Fig. 3 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 where M is the bending moment to which the
and Ar1, upper and lower transformation temperatures, respectively, on heating (refroidissant) a hypoeutectoid steel beam is subjected at x. Bending moment M is
Residual Stresses and Fatigue Behavior / 53

23. K.N. Smith, P. Watson, and T.H. Topper, J. Ed., DGM-Informationsgesellschaft Verlag, 52. B. Syren, Dr.-Ing. Thesis, University Karls-
Mater., Vol 5, 1970, p 767778 Oberursel, 1985, p 237283, in German ruhe (TH), 1975
24. W.J. Ostergren, J. Test. Eval., Vol 4 (No. 39. P. Starker, E. Macherauch, and H. Wohl- 53. H. Traiser and K.-H. Kloos, Z. Werkstoff-
5), 1976, p 327339 fahrt, Fatigue Eng. Mater. Struct., Vol 1, techn., Vol 16, 1985, p 135143
25. A. Ebenau, Dr.-Ing. Thesis, University 1987, p 319327 54. R. Schreiber, H. Wohlfahrt, and E. Mach-
Karlsruhe (TH), 1989 40. S. Jaegg, Dr.-Ing. Thesis, University GH erauch, Arch. Eisenhuttenwes., Vol 49,
26. A. Ebenau, D. Eifler, O. Voehringer, and E. Kassel, 1999 1978, p 207210
Macherauch, Proc. ICSP 4, K. Iida, Ed., The 41. S. Jaegg and B. Scholtes, ICRS 5, Linkop- 55. R. Schreiber, H. Wohlfahrt, and E. Mach-
Japan Society of Precision Engineering, To- ing, Sweden, 1997, p 10781083 erauch, Arch. Eisenhuttenwes., Vol 48,
kyo, 1990, p 327336 42. E. Welsch, D. Eifler, B. Scholtes, and E. 1977, p 653657
27. F. Hoffmann, Dr.-Ing. Thesis, University Bre- Macherauch, Proc. Sixth European Conf. on 56. H. Wohlfahrt, Mechanical Surface Treat-
men, 1987; Fortschritts-Berichte VDI Reihe 5 Fracture (ECF 6), V. Elst and A. Bakker, ments, H. Wohlfahrt and P. Krull, Ed., Wi-
Nr. 123, VDI-Verlag, Dusseldorf, 1987 Ed., Eng. Mater. Adv. Services Ltd ley-VCH, Weinheim, Germany, 2000, p
28. I. Altenberger, Dr.-Ing. Thesis, University (EMAS), Amsterdam 1986, p 13031320 5584, in German
GH Kassel, 2000 43. A. Stacey and G.A. Webster, Materials Re- 57. K.-H. Kloos, B. Fuchsbauer, and J. Adel-
29. I. Altenberger, B. Scholtes, U. Martin, and search Society Proc., Vol 22, 1984, p 215 mann, J. Fatigue, Vol 9, 1987, p 3542
H. Oettel, Mater. Sci. Eng. A, Vol 264, 44. H.P. Lieurade, Advances in Surface Treat- 58. P. Starker and E. Macherauch, Z. Werkstoff-
1999, p 116 ments, A. Niku-Lari, Ed., Pergamon Press, techn., Vol 14, 1983, p 109115
30. G. Kuhn, Dr.-Ing. Thesis, University Karls- Oxford, 1987, p 455482 59. P. Starker, Dr.-Ing. Thesis, University
ruhe (TH), 1991 45. K.J. Kang, J.H. Song, and Y.Y. Earmme, Karlsruhe (TH), 1981
31. G. Kuhn, J.E. Hoffmann, D. Eifler, B. Fatigue Fract. Eng. Mater. Struct., Vol 12, 60. A. Navarro and E.R. de los Rios, Philos.
Scholtes, and E. Macherauch, ICRS 3, H. 1989, p 363376 Mag. A, Vol 57 (No. 1), 1988, p 1536
Fujiwara, T. Abe, and K. Tanaka, Ed., El- 46. J.E. Hoffmann, D. Loehe, and E. Macher- 61. K.J. Miller and E.R. de los Rios, The Be-
sevier Applied Science, London, 1991, p auch, Proc. ICSP3, H. Wohlfahrt, R. Kopp, havior of Short Fatigue Cracks, Mechanical
12941301 and O. Voehringer, Ed., Garmisch-Parten- Engineering Publications Limited, London,
32. A. Glaser, D. Eifler, and E. Macherauch, kirchen, DGM Verlag, Oberursel, Germany, UK, 1986
Mater.wiss. Werkst.tech., Vol 22, 1991, p 1987, p 631638 62. B. Scholtes, Structural and Residual Stress
266274 47. J.E. Hoffmann, D. Loehe, and E. Macher- Analysis by Nondestructive Methods: Eval-
33. H. Mughrabi, Fatigue Behavior of Metallic auch, Residual Stresses in Science and uation, Application, Assessment, V. Hauk,
Materials, D. Munz, Ed., DGM-Informa- Technology, Proc. ICRS 1, E. Macherauch Ed., Elsevier Science B.V., Amsterdam,
tionsgesellschaft Verlag, Oberursel, 1985, p and V. Hauck, Ed., Vol 1, DGM Verlag, Ob- 1997
738, in German erursel, Germany, 1987, p 801808 63. W. Backfisch and E. Macherauch, Arch. Ei-
34. J.D. Almer, J.B. Cohen, and B. Moran, Ma- 48. J.E. Hoffmann, D. Eifler, and E. Macher- senhuttenwes., Vol 50, 1979, p 167171
ter. Sci. Eng. A, Vol 284, 2000, p 268279 auch, Residual Stresses: Origination, Mea- 64. K. Wellinger and D. Dietmann, Calculation
35. H. Berns and L. Weber, Residual Stresses surement, Assessment, E. Macherauch and of Strength, A. Kroener Verlag, Stuttgart,
in Science and Technology, E. Macherauch V. Hauck, Ed., Vol 2, DGM Verlag, Ober- 1969, in German
and V. Hauk, Ed., DGM-Informationsge- ursel, Germany, 1983, p 287300, in Ger- 65. B. Winderlich, Mater.wiss. Werkst.tech.,
sellschaft Verlag, Oberursel, 1987, p 751 man Vol 21, 1990, p 378389
758 49. J.E. Hoffmann, Dr.-Ing. Thesis, University 66. K.H. Kloos, Notches and Structural Dura-
36. H. Berns and L. Weber, Shot Peening, ICSP Karlsruhe (TH), 1984 bility, H. Nowack, Ed., Deutscher Verband
3, H. Wohlfahrt, R. Kopp, and O. Voehrin- 50. B. Syren, H. Wohlfahrt, and E. Macherauch, fur Materialforschung und-pruefung e.V.,
ger, Ed., DGM-Informationsgesellschaft Arch. Eisenhuttenwes., Vol 46, 1975, p 1989, p 740, in German
Verlag, Oberursel, 1987, p 647654 735739 67. J. E. Hoffmann and D. Loehe: to be pub-
37. R. Herzog, Dr.-Ing. Thesis, University 51. B. Syren, H. Wohlfahrt, and E. Macherauch, lished
Braunschweig, 1997 Proc. Second Int. Conf. Mech. Behavior of 68. T. Fett and D. Munz, Stress Intensity Fac-
38. E. Macherauch and H. Wohlfahrt, Fatigue Materials (ICM2) (Boston), 1976, Ameri- tors and Weight Functions, Computational
Behavior of Metallic Materials, D. Munz, can Society for Metals, p 807811 Mechanics Publications, 1997
Handbook of Residual Stress and Deformation of Steel Copyright 2002 ASM International
G. Totten, M. Howes, T. Inoue, editors, p54-69 All rights reserved.
DOI: 10.1361/hrsd2002p054 www.asminternational.org

Stability of Residual Stresses


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

RESIDUAL STRESSES are generated in creep (Ref 1617). If these processes occur to Relaxation of Residual Stresses by
structural components during manufacturing only a limited extent or not at all, relaxation of Annealing
processes such as forging, machining, heat treat- residual stresses is also conceivable by crack for-
ing, shot peening, and many others. These mation and propagation. Relaxation of residual
If a pure metal is annealed for several hours
stresses are always a consequence of inhomo- stresses in the real case occurs by complex in-
at a temperature of
geneously distributed dimensional changes due teraction of a large number of factors. It depends
to inhomogeneous plastic deformations, ther- not only on the residual stress state itself but also Ta  0.5 Tm [K] (Eq 1)
mochemical treatments, and/or phase transfor- on the material state, loading condition, geom-
mations. They can be either beneficial or detri- etry, and environment of the component under where Tm is the melting or solidus temperature,
mental to component behavior in service, consideration. and then cooled slowly to room temperature, al-
depending on the materials state as well as the The best known and most important tech- most total relaxation of the residual stresses aris-
sign, magnitude, and stability of the residual niques for inducing residual stress relaxation are ing from forming, machining, heat treatment, or
stresses, mechanical loading, and environmental annealing (tempering), uniaxial deformation joining operations can be achieved. The neces-
conditions (Ref 115). Compressive macro re- (drawing, stretching), and cyclic deformation sary annealing time depends essentially on the
sidual stresses in the surface region of materials (Ref 1618). Relaxation can also be caused by workpiece dimensions and the material state.
with medium and high hardnesses increase the thermal cycling, quenching, neutron bombard- Since an annealing temperature of 0.5 Tm cor-
fatigue life and the fatigue limit at cyclic loading ment, the effect of alternating magnetic fields (in responds to the recrystallization temperature of
compared to materials states that are free of re- the use of ferromagnetics), vibration, and partial ferritic steels, complete relaxation of macro re-
sidual stresses. This improvement is caused by damage. As discussed in other articles in this sidual stresses can be expected. In the case of
an increased resistance against crack initiation Handbook, residual stresses can be differentiated pearlitic steels, macro residual stresses are com-
and, to a certain extent, against crack propaga- by their technological origins (i.e., whether they pletely eliminated during the pearlite-austenite
tion if the residual stresses are sufficiently stable transformation above the A1 temperature. Micro
were produced during forming, machining, heat
in the areas of highest loading (in most cases, the residual stresses are considerably reduced but
treatment, joining, coating, or casting).
component surface areas). Moreover, compres- not entirely removed, since lattice defects (par-
Before the effect of applied thermal or me-
sive macro residual stresses can increase the re- ticularly dislocations) and, in the case of hetero-
chanical energy on the residual stress state can
sistance of certain materials against corrosion fa- geneous materials, the different expansion co-
be assessed, the latter must be determined quan-
tigue and stress corrosion cracking. They can efficients of the various phases are always
titatively. In the case of mechanical methods, the
also improve the wear resistance. Therefore, in responsible for some micro residual stresses. Re-
many cases compressive macro residual stresses necessarily large number of measurements in-
volve considerable expenditure of both time and sidual stress relaxation by annealing is brought
are intentionally generated in near-surface re- about by so-called thermally activated processes
gions by controlled heat treatments or by post- money. Most investigations have thus employed
x-ray methodsfor example, the sin2w method for which the annealing temperature and the an-
treatments such as shot peening or deep rolling. nealing time are interchangeable within certain
The stability or relaxation behavior of these (Ref 19) with CrK radiation on the {211}-in-
limits. In order to achieve comparable residual
residual stresses at purely thermal or mechanical terference plane of the ferrite, as well as
stress relaxation at a lower annealing tempera-
loadings as well as superimposed thermal and interference-profile analyses (e.g., Ref 20, 21).
ture, the annealing time must be increased cor-
mechanical loadings is thus of decisive impor- In this article, current knowledge about resid- respondingly (Ref 16, 17).
tance for the service behavior of components and ual stress relaxation of steels is presented in a Thermal residual stress relaxation is funda-
hence of great interest from a scientific as well condensed and, as far as possible, systematic mentally affected by the residual stress state it-
as a practical point of view (Ref 1618). If re- form. The abundance of available experimental self and by the material state. This is convinc-
sidual stresses are relaxed by annealing or me- data nevertheless necessitates a limited scope. ingly demonstrated by the findings presented in
chanical treatment, they naturally have little if Thus, the findings of residual stress relaxation Fig. l and 2 (Ref 16, 17), which show the effect
any influence on subsequent component failure. by annealing, tensile loading, compressive load- of 1 h anneals on surface macro residual stresses
Residual stresses can be reduced or com- ing, and cyclic loading will be taken as examples of a variety of origins in a variety of steels in-
pletely relaxed by the application of mechanical to demonstrate universally valid rules. Shot vestigated using x-ray methods. Figure 1 is a bar
and/or thermal energy. The elastic residual peening residual stresses are particularly suitable chart showing the macro residual stresses present
strains, ee, that are associated with the residual for such investigations due to their sign and their in the original state and after the normal indus-
stresses via Hookes law can be converted into relatively large absolute value. Possible formulas trial stress relieving for hardened components of
micro plastic strains, ep, by suitable deformation that can be used to quantify residual stress relax- 1 h at 200 C (390 F). The materials deformed
processes. For example, this transformation can ation are discussed, and the underlying micro- in tension have undergone negligible residual
be achieved by dislocation slip, dislocation structural as well as micromechanical mecha- stress relaxation. The greater the original com-
creep, grain-boundary sliding, and/or diffusion nisms are considered. pressive residual stresses, the greater the residual
Stability of Residual Stresses / 55

stress relaxation produced by annealing of hardening micro residual stresses are relaxed at and thus, from the hardness of the material (Ref
surface-machined specimens: 11% for a plain lower temperatures, such as those produced by 25). The macro residual stresses (left-hand side
carbon steel with 0.45 wt% Carbon, SAE 1045 machining; likewise, machining micro residual of Fig. 3) and the micro residual stresses (right-
(Fig. 1C); 17% for SAE 1045 (Fig. 1D); and stresses are relaxed at lower temperatures, such hand side; characterized by the half-width ratio
19% for the rolling bearing steel AISI 52100 as those produced by deformation in tension. In DHW/DHW0), yield with the hardness increasing
(Fig. 1E). The most pronounced relative residual comparable steel states, micro residual stresses from the normalized to the quenched and tem-
stress relaxation, however, is observed in the are relaxed only after a longer period or at a pered and to the hardened condition, greater re-
hardened materials: 25% for plain carbon steel higher temperature than macro residual stresses laxation rates. Figure 3 also illustrates the delay
with 0.22 wt% C, SAE 1023 (Fig. 1F); and 37% (Ref 16, 17). For example, in order to achieve of the relaxation of the micro residual stresses,
for SAE 1045 (Fig. 1G). 50% micro residual stress relaxation in tensile- which was mentioned above in comparison with
Figure 2 shows the relative residual stresses deformed SAE 1045 annealed for 1 h, the an- the macro residual stresses. Further experimental
rrs(Ta)/rrs(293 K) for the same steels as a func- nealing temperature must be 100 C (180 F) investigations of the macro and micro residual
tion of the homologous annealing temperature, higher than that needed to achieve the same de- stress relaxation of metallic materialstemper-
Ta /Tm. It can be seen that a 1 h anneal at 0.5 Tm gree of macro residual stress relaxation. ing steel AISI 4140 (Ref 2628), titanium alloy
(about 600 C, or 1110 F) results in complete Data pertaining to the thermal relaxation of Ti-6Al-4V (Ref 29), and several CuZn alloys and
relaxation of the macro residual stresses in every micro residual stresses for hardened steels are AlMg alloys (Ref 30, 31)lead to similar re-
case. Characteristic, material-specific stress- available by analyzing x-ray interference lines of sults.
temperature curves are obtained. Clearly, resid- different steels in (Ref 2224). The mean lattice Materials with surface residual stresses pro-
ual stresses produced by hardening are relaxed distortions e21 / 2, which are proportional to the duced by grinding, milling, deep rolling, or shot
at lower temperatures, machining residual micro residual stresses according to peening frequently show different residual stress
stresses at medium temperatures, and deforma- relaxation behavior when comparing the surface
micro E110 e 
rrs 2 1/2
tion residual stresses at higher temperatures. In (Eq 2) layer with subsurface areas. In most cases, this
order to achieve the same degree of residual behavior is connected with work-hardening gra-
stress relaxation, (e.g., 50%) in 1 h in hardened E110 is the Youngs modulus of iron in the 110 dients, where an increasing work hardening due
and in deformed SAE 1045, a temperature dif- direction, increase with growing carbon content to increasing dislocation densities causes grow-
ference of 150 C (270 F) is necessary. Fur- mainly due to the increasing dislocation density ing mean strains and greater half-width values.
thermore, according to Fig. 1, larger initial re- and, secondarily, to the increasing number of Characteristic examples of depth distributions
sidual stresses result in a shift of the rrs versus solute carbon atoms in octahedral sites of the of the macro residual stresses and the half-width
T curve to lower annealing temperatures in the tetragonal martensite lattice. Usually, it can be values in SAE 1045, which were produced after
case of hardened or machined materials. assumed that with increasing carbon content the different heat treatments and electrolytical re-
Micro residual stresses and their relaxation be- temperature for the onset of residual stress relax- moval of thin material layers, are shown Fig. 4
havior can be determined directly from the half- ation is lowered and the recovery rate increases. and 5 (Ref 25). Normalized SAE 1045 samples
width values (HW) of the x-ray interference However, it must be remembered that with the in the ground, milled, or shot-peened condition
lines, which are a measure of the microstructural annealing-induced formation of carbides on dis- result in typical depth distributions that depend
work-hardening state as well as the dislocation locations, which leads to their pinning, the re- on the heat treatment (Fig. 4). In this case, the
density. The change of the HW values are ap- laxation procedure slows. relaxation behavior of the macro residual
proximately proportional to the changes of the The relaxation behavior of residual stresses stresses (left-hand side), and of the half-widths
micro residual stresses (DHW  Drrs micro). depends for a similar surface treatment, as for (right-hand side), is more pronounced at the sur-
Again, the micro residual stresses in steels are example by shot peening, as shown in Fig. 3 for face than in deeper layers. After selected an-
more effectively relaxed at a higher temperature the plain carbon steel SAE 1045 with 0.45 wt% nealing treatments, this behavior holds to com-
and, like macro residual stresses (see Fig. 2), C, from the heat treatment before shot peening pletely 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
400 Ck45 C, or 1020 F).
Ck22 G The relaxation behavior of normalized SAE
F 1045 in the shot-peened condition is reproduced
Ck45 Ck22 Ck45 Ck45 100Cr6
A B C D E
0
Ta, C
Hardened 0 100 200 300 400 500 600
1.0
Normalized
rs, MPa

+5% deformed ta = 1 h
rs(Ta)/rs(293 K)

400 0.8
in tension C
Normalized B
0.6
+ ground A
E
0.4
800 F D
G
0.2

Untempered Hardened + 0
shot peened 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
200 C/1 h tempered
1200 Ta/Tm, K

Fig. 2 Macro residual stress ratio versus homologous


Fig. 1 Macro residual stress (rrs) of different steels before and after annealing for 1 h at 200 C (390 F). Ck 45 SAE annealing temperature of different steels (see
1045; Ck 22 SAE 1023; 100 Cr 6 AISI 52100 Fig. 1) after annealing for 1 h
Measurement of Residual Stresses / 103

was then measured at each core depth until the removing thin slices parallel to the original sur- from the surface of the plate and measuring the
core was milled completely through the plate. face and from the bisected block as shown in the strains existing at each layer. The process as-
The cored plug was then assumed to be com- upper part of Fig. 4. sumed that the stress perpendicular to the surface
pletely free of residual stress. The measured dis- The change in strain of the blocks is measured was zero. It should be noted that in order for this
tances were then used to reconstruct the original using shallow holes or dimples in the original procedure to be valid, the areas measured by
biaxial stress condition of the plug at each cored surfaces of the component. These gage points are XRD would have to be free of plastic deforma-
depth. The mechanical measuring gage used in located along the long axis of the block on the tion caused by layer removal (see Sectioning
this technique was similar to that used in Gun- original faces of the plate (see Fig. 4). The dis- and Material-Removal Methods).
nerts first technique (Ref 31). tance between these gage marks is measured be- Triaxial Conditions. In reality, in most com-
Another approach to measuring residual fore and after removal of the blocks from the ponents in which residual stresses have been in-
stresses that has a broader application with re- plate and after each sectioning of the block (Ref duced, usually due to manufacturing processes,
spect to the shape of the component and the 33). the stress field is triaxial and varies from point
stress-field distribution was proposed by Rosen- This procedure assumes a constant biaxial to point (element to element) in all three direc-
thal and Norton (Ref 33). It is applicable to stress field over the length of the blocks, which tions. Thus, a number of destructive procedures
plates and plate-shaped weldments. is not the case in welded plates in the direction and stress-field condition assumptions have been
applied in order to measure the three-dimen-
The procedure involved cutting two narrow transverse to the weld. The far side of the block
sional residual-stress-field condition existing in
blocks having the full thickness of the plate, each in Fig. 4 shows a sectioning procedure that
most components of practical engineering inter-
with its long axis parallel to one of the assumed would reveal the stresses parallel to the weld
est. Two of these are described in this section.
biaxial principal, residual-stress directions in the along a gradient transverse to the weld. The far
Chen (Ref 34) revised Rosenthal and Nortons
surface of the plate (see the near side of Fig. 4). side of the block in Fig. 4 shows a sectioning approach to deriving the triaxial residual-stress
Thus, the long axes of the blocks are perpendic- procedure that would reveal the stresses parallel condition (Ref 33) as follows. The typical
ular to each other and parallel to the face, the to the weld along a gradient transverse to the method of residual-stress measurement is by me-
largest area surface of the plate. The smallest weld. Also, electrical-resistance strain gages chanically removing part of a body and measur-
dimension of the block should be several times could be used instead of measuring the distance ing the change of stress in the rest of the body.
smaller than the plate thickness and its largest between shallow holes or dimples. The method of Rosenthal and Norton instead
dimension at least twice the thickness. The block Another approach to a constant biaxial stress deals only with a small element that has been cut
then can be further sectioned in order to deter- field in a flat plate, varying only through thick- free from a plate. The sectioning procedure con-
mine the biaxial stress variation through the ness, was described by Moore and Evans (Ref sists of removal of a narrow block from a plate
thickness of the component. This proceeds by 2). They relied on XRD for the measurement of with gages attached, followed by splitting the
first cutting the block at the location representing the strains from which the stress was calculated, block in half with gages attached on the top and
the midthickness of the component plate, then and the procedure consisted of removing layers 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 0.1 in., gages are removed and stresses are mea-
sured (at least two points) on the surface by
40 10 20 20 20 50 XRD. The basic assumptions are:
Partial stress relief occurs in the direction of
the long axis of the block and a total stress
relief occurs in the direction transverse to the
long axis.
90 The small amount of stress relaxed in the re-
Initial
sectioning
mainder of the block follows a linear law
through the thickness when a thin slice of
270 metal is removed.
Variation of transverse stress along the axis
Weld of the weld is small in the middle portion of
the plate weldment.

45
10 Saw cut
10 Fine cutter 180
(i.e., green abrasive saw 0.5 mm thick) Gage

5 Further
5
sectioning
10 Weld
Final
sectioning 10

10
10 Rosette gage

Fig. 4 Welded steel plate, the near side of which shows


Fig. 3 Residual-stress measurement of a girth-welded pipe by strain gaging and sectioning. Note that strain gages the two narrow blocks suggested in Rosenthal
shown in the final sectioning should be placed on the pipe prior to initial sectioning; and for a more complete and Nortons procedure (Ref 33). The far side shows several
analysis several of the layer sections detailed in the final sectioning step should be strain gaged and sectioned. Dimensions blocks sectioned to reveal the stresses parallel to the weld
given in mm with a gradient transverse to the weld.
Stability of Residual Stresses / 57

yields DH 3.3 eV, m 0.122, and B 1.22 rrs(Ta, ta)/rrs0 in Eq 4. The new algorithm was AISI 4140 in normalized as well as quenched
1021 min 1 for the surface values of the macro used to determine the material parameters DH and tempered states.
residual stresses. The curves in Fig. 8(a) were 2.64 eV, m 0.096, and B 5.32 1012 In order to obtain some information about the
calculated using these constants in Eq 4 and 5. min1, which allow description of the measured relaxation behavior of macro residual stresses in
They describe the time and temperature depen- values as shown in Fig. 8(c). subsurface layers, the values of the residual
dence of the relaxation process very well. Comparison of these relaxation data in Ref 26 stresses after the annealing processes of AISI
The alterations of surface half-widths by an- and 27 shows that the relaxation of macro resid- 4140 mentioned above were measured for dif-
nealing at temperatures between 250 and 450 C ual stresses at the surface of the shot-peened state ferent distances from the surface (Ref 26, 27).
(480 and 840 F) for different annealing times is a bit faster than the relaxation of half-widths The evaluated parameters of the Avrami ap-
are shown in Fig. 8(b) for the same steel. The and mean strains; this corresponds with the fact proach vary only at the surface itself from those
reductions of the HW values are similar to the that the constant B differs by nearly nine orders values measured below the surface. There, the
relaxation of the macro residual stresses pre- of magnitude (see Table 1). The reason for the evaluated data show no significant tendency and
sented in Fig. 8(a). Again, the Avrami approach differences in the relaxation rate is that for the amount to the mean values DH 2.99 eV, m
is used to describe the relaxation of half-widths, relaxation of macro residual stresses, dislocation 0.172, and B 6.09 1017 min1. As
applying the new iterative method to the differ- movement is sufficient. For a distinct relaxation shown in Fig. 9, the dependence of the macro
ences between the half-widths after annealing of micro residual stresses, however, additional residual stresses after different annealing times
and the value HW 1.65 of a normalized spec- dislocation annihilation is necessary. The expo- at 450 C (840 F) on the distance from surface
imen related to their starting values instead of nents m show no significant alterations, and the can be described quantitatively using the surface
the ratio rrs(Ta,ta)/rrs
0 in Eq 4. The curves in Fig. activation enthalpies approach the values of the material properties and the mean values for all
8(b) were calculated by means of the material activation enthalpy of self-diffusion of iron: DHS subsurface layers. The agreement between mea-
properties DH 2.48 eV, m 0.116, and B  2.8 eV. Additional data for the same steel in sured values and modeled curves is very good.
1.09 1013 min 1 and agree well with the mea- a normalized and another quenched and tem- During heating to sufficiently high annealing
sured values. The dependence of the relaxation pered state in Table 1 (with exception of the temperatures, a distinct relaxation of macro re-
of mean strains e21 / 2 and micro residual hardened state) confirm this statement (Ref 28). sidual stresses occurs. An attempt was made to
stresses rrs
micro, respectively, on the annealing Accordingly, volume diffusion-controlled dis- model this behavior by extending the Avrami ap-
time was measured in the temperature range be- location creep in the residual stress field that is proach to nonisothermal stress relaxation (Ref
tween 250 and 450 C (480 and 840 F). The dominated by climbing of edge dislocations 27). The transient relaxation of the macro resid-
data can also be modeled by the Avrami ap- should be the rate-controlling process for the re- ual stresses was calculated for specimens that
proach using e2(Ta, ta)1/2/e201/2 instead of laxation of the shot peening residual stresses of 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 relaxa-
tion of residual stresses, the real T(t) relationship
3.0 was partitioned into a staircase curve with small
400 Ground
equidistant steps and isothermal sections. Cal-
350 C/22 min culation of the relaxation was determined by a
2.5 numerical integration after the so-called stress-
200 transient method (Ref 27) under the application
of the Avrami quantities DH, m, and B (Table 1)
2.0
which are valid for the surface state. The mag-
nitudes of the macro residual stresses measured
0 after interruption of the heating and cooling
0 down to room temperature are marked with tri-
Milled angles. It becomes evident that the stress-
2.4
350 C/22 min
transient method describes the measured values
rs, MPa

HW, 2

100 in an excellent manner. Figure 10 also shows the


2.2 isothermally calculated relationship between
heating time and absolute values of residual
2.0 stresses. While there are distinct differences dur-
200
ing the first stage of heating time of approxi-
mately 20 s, the curves calculated isothermally
0 0.1 0.2 0 0.1 0.2 and by the stress-transient method closely ap-
proach at long annealing times.
0 A similar discussion of thermally activated
Shot peened
3.0
processes was carried out for other steels (Ref
400 C/30 min 16, 17, 28, 30, 31) and for AlMg alloys and
500 C/30 min CuZn alloys (Ref 30, 31) with surface residual
200
550 C/30 min 2.5
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
400 2.0 in the range of 1.1 to 2.6 eV. DH is lowest for
relaxation of residual stresses due to hardening
0 0.1 0.2 0.3 0 0.1 0.2 0.3 and highest for those due to deformation in soft
annealed states. That proves unequivocally that
Distance from surface, mm
residual stress relaxation in steels can occur by
Macro residual stress and half-width of the x-ray interference line versus distance from surface before and after several processes.
Fig. 4 Characteristic structural changes occur during
annealing of ground, milled, and shot-peened SAE 1045 samples in a normalized state
58 / Effect of Materials and Processing

the deformation, machining, or hardening of case of randomly distributed dislocations or tan- is required in a component without significant
steels. Typically, an increase in the dislocation gles of extremely high density, qt for example, change in yield strength or tensile strength, the
density and a change in the dislocation arrange- in hardened steels where qt  1012 cm2 (Ref annealing temperature and time must be chosen
ment are observed. In the case of hardening, the 23, 24)residual stress relaxation is expected to correspond with the recovery stage and not
concentration of solute interstitial atoms differs to involve dislocation-core diffusion-controlled with recrystallization.
from the equilibrium value, and this, together climb by edge dislocations. Predominantly vol-
with the presence of dislocations, has a decisive ume diffusion will determine recovery if the dis- Resistance to Residual Stress Relaxation
influence on the residual stress fields. If a single location configurations are relatively stable and
heat treatment is carried out in the temperature consist of plane arrangements, cell walls, or sub- Residual stress relaxation by heat treatment is
range corresponding to recovery (T  0.5 Tm), grain or low-angle grain boundaries. This recov- fundamentally impossible if in a predominantly
the dislocations adopt arrangements of lower en- ery process probably occurred in the investigated uniaxial residual stress state rrs is smaller than
ergy by elementary processes such as glide and shot-peened quenched and tempered steels and the creep yield strength. This resistance for the
cross slip by screw dislocations and glide and normalized steels as well as nonferrous alloys. onset of plastic creep deformation, designated in
climb by edge dislocations. In the case of hard- In view of the activation energies, the residual the following by Rce, is characterized by the
ened material, diffusion of carbon atoms depen- stress relaxation in the hardened steels can prob- creep strain limit at vanishingly small plastic de-
dent on the annealing temperature and accom- ably be classified between these two extremes formation. As shown schematically in Fig. 11,
panied by the formation of characteristic and occurred by two recovery mechanisms in Rce decreases with increasing temperature and
carbides is superimposed on these processes. competition with each other. load time. In contrast to relevant times regarding
The rate-determining process, with the exception Up to now, discussion has centered on resid- the creep condition of a high-temperature com-
of the early stages in the annealing of hardened ual stress relaxation at temperatures Ta  0.5 ponent at service, only very short times are nec-
steels (Ref 36), is clearly the thermally activated Tm that is, those temperatures brought about essary for residual stress relaxation by creep pro-
climb of edge dislocations (Ref 17). If diffusion by typical recovery processes. In this case, me- cesses (dislocation creep, grain-boundary glide,
of matrix atoms occurs along the edge disloca- chanical parameters such as hardness and yield or diffusion creep). With increasing T and/or t,
tions to the dislocation core, the activation en- strength are not significantly altered. During a Rce approaches a localized residual stress peak
thalpy should be DH  0.5 DHS, where DHS is recrystallization anneal at Ta  0.5 Tm, the dis- of magnitude rrs. For T Tti, rrs equals Rce
the activation enthalpy of self-diffusion. If vol- location density rapidly takes very small values associated with the localized onset of creep de-
ume diffusion predominates, the DH value for as a result of the growth of new grains. This leads formation. Further increases in temperature or
time result in an increasing and measurable mi-
climb is determined by DHS. In the real case, to complete removal of macro residual stresses
croplastic creep strain. As illustrated in Fig. 11,
both processes occur simultaneously but to dif- and to small micro residual stresses, but is as-
residual stress relaxation begins at higher tem-
ferent degrees. The dislocation density and ar- sociated with pronounced changes in mechanical
peratures (Ti) the smaller the load time (ti) or the
rangement are of considerable importance. In the properties. If extensive residual stress reduction residual stress (rrs) and the greater the creep re-
sistance (Rce) of the material. Changes in Rce can
be achieved by deliberate alterations in the state
0

Shot peened
400 C/30 min
400 3
500 C/30 min
550 C/30 min
800 Ta,4
Normalized 2
Normalized Ta,3
log ln (0/rs)

Ta,2
0
rs

6 Ta,1
Shot peened
rs, MPa

400 300 C/60 min


HW, 2

5 m

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


Quenched and tempered 4
800 log ta
Quenched and tempered 3

0
6

400 5
log ta

4 Ha/ln10
Shot peened
800
Hardened Hardened 300 C/60 min 3

0 0.2 0.4 0 0.2 0.4


Distance from surface, mm 1/kTa

Fig. 5 Macro residual stress and half-width versus distance from surface before and after annealing of shot-peened Fig. 6 Schematic of conventional determination of
SAE 1045 samples in normalized, quenched and tempered, and hardened states Avrami approach parameters
Measurement of Residual Stresses / 107

N
condition can be filled and the measurement can 2x vx
be fit using an approximate method, such as the rxy
x2

n1
exy
y
(Eq 52)
least-squares method. If this can be performed
Cnx[(nb sinh ny ny cosh ny) one obtains:
sufficiently accurately, these trial solutions are
close to the exact solution, since it has been (1 nb coth nb)ny sinh ny]cos nx (Eq 48)
v(1 v)

y2
shown that the solution is uniquely determined vx(x, y) Cx C0xy
by the stated boundary conditions. 2x
N E 2
sxxy Cnx 1v
N


2
For the loading case Tx(x): xy n1

E
Cnx nby sinh ny 2b
[2nby sinh ny (1 nb coth nb) n1
N
y3 y2 Cnx v
x(x, y) Cx
6
C0x
2

n1 n
2
(sinh ny ny cosh ny)]sinh nx (Eq 49) cosh ny
1v
nby sinh ny

Since plane strain has been assumed xz can be


[nb(sinh ny ny cosh ny)
(1 nb coth nb)ny sinh ny]cos n (Eq 46)
calculated from: (1 nb coth n y cosh ny
1 v
xz m(rx ry) (Eq 50) sinh ny
n 1v
is valid where n np/a and 2a and 2b are the
cos x
width and thickness of the plate after removing where m is the Poissons ratio. Further, the strain 1
exy can be obtained from: sinh ny y cosh ny n (Eq 53)
material. n

Equation 39 gives:
1 m2

m A similar treatment is used for the loading case
exy ry rx (Eq 51) r
T (x). In that case, the solution is given by:
E 1 m
N
2x
rxx
y2
Cxy C0x Cnx[(nb(coth
n1 where E is the elastic modulus. y(x, y) Cy
y3
C0y
y2

N
Cny
Since the condition for compatibility is satis- 6 2 n1 n2
bb 1)(2 cosh ny ny sinh ny) nb
fied, the displacement in the y-direction is [(1 nb coth nb)(sinh ny ny
(sin hny ny cosh ny)]coshnx uniquely determined, excepting rigid body dis-
(Eq 47) placements, from the relationship: cosh ny) 2n by sinh ny]cos nx (Eq 54)

from which are obtained:


N
2y
y yx
y2
Cy C0y
n1

Cny[2nb cosh ny 2n by sinh ny


(1 nb coth nb) (sinh ny
Removed layer
ny cosh ny)]cos nx (Eq 55)
N
2y
x
yy
x2

n1
Cny

x(x)
Ty (x) T (x ) [(1 nb coth nb)(sinh ny ny
y cosh ny) 2nby sinh ny]cos nx (Eq 56)
Internal forces Tx(x) 
2 y N
syxy
xy

n1

Cny[nb sinh ny n2by cosh ny)


(1 nb coth nb) ny
x
x(x) sinh ny]sin nx (Eq 57)

Finally, the displacement is given by:


x(x)corr
v(1 v)

y2
y vy(x, y) Ey C0yy
E 2
N
1 v

2

E n1
Cny b sinh ny

v
nby cosh ny
1 v
x
(x) (b sinh ny nby cosh ny)
x(x)

x(x)
(1 coth nb) y sinh ny
2 v
cosh ny
Fig. 6 Residual-stress distributions, forces, and distortion of a plate before and after layer removal. Top: residual-stress n 1 v
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) y sinh ny cos nx (Eq 58)
60 / Effect of Materials and Processing

cannot, however, be compared directly to a creep nation of the two expressions that increasing re- Residual Stress Relaxation by
test. It is much more like a stress relaxation ex- sidual stress values lead to a more effective re- Uniaxial Deformation
periment. In the latter case, the total strain re- sidual stress relaxation. As a result of the greater
mains constant while elastic strain is converted driving force, shorter times and/or lower tem-
into plastic strain. With a homogeneous stress peratures are necessary. This is in agreement In certain cases in practice, uniaxial defor-
distribution over the cross section of a specimen, with the experimental results presented. mation is often employed in addition to stress-
the following expression would be valid: The influence of the magnitude of the residual free annealing to relieve residual stresses. In the
stresses can also be illustrated by describing the case of forming, for example, the residual
r residual stress relaxation with the Norton ap- stresses can be reduced by a second forming
et constant ep ee ep (Eq 11) proach, which is known from high-temperature stage using a smaller reduction in the cross-
E
creep. The total strain rate as the sum of elastic sectional area. This can be achieved by redraw-
and plastic strain rate must vanish according to ing, restretching, rerolling, repressing, and
However, macro residual stresses are inhomo- straightening (Ref 16, 17). However, these tech-
geneously distributed over the cross section, and niques can be used only on simply shaped com-
e t e p e e 0 (Eq 13)
for the residual stress relaxation in a localized ponents with a uniform cross section. In the case
area the following relationship holds true: of welded seams, a uniaxial load is applied to
Thus, the Norton law and Eq 13
reduce or redistribute macro residual stresses.
rrs
0
rs
r When a critica1 value of the applied loading
constant ep r rs
E E e p C*(T) (rrs)n e e (Eq 14) stress is exceeded, directed dislocation move-
E ment converts the elastic strain associated with
0 r )
(rrs rs
ep (Eq 12) the macro residual stress into micro plastic
E
give a proportionality between the elastic and strain.
plastic strain rates, and between the residual re- Several typical examples will serve to illus-
Regarding real values of the residua1 stresses, laxation rate and the actual residual stress to the trate the relaxation of residua1 stresses due to
plastic strains of several tenths of a percent max- power of the Norton exponent n, respectively. joining and shot peening by uniaxial deforma-
imum are produced by complete residual stress The experimental data shown in Fig. 12 reveal tion. In an evaluation of the effect on the strength
relaxation (rrs 0); that is, the deformation is strain rates that present a strong dependency on of macro residual stresses set up during welding,
in the microcreep range. temperature and residual stress values (Ref 26). it is important to know the stability of these
Equations 10 and 11 form the basis for quan- Furthermore, the data reveal strain rates that are stresses on loading the weld. Figure 13 illustrates
titative estimates of the residual stress relaxation. typical for creep processes. This finding supports the considerable reduction in macro residual
Multiaxial and inhomogeneous residual stress the conclusion drawn above that diffusion- stresses accompanying tensile loading of joints
states are neglected or excluded. Although Eq 10 controlled dislocation creep in the residual stress
cannot be substituted directly in Eq 11 on ac- field should be the rate-controlling process for
count of the variable value of the stress (r the relaxation of residual stresses in steels.
rrs), it can be seen qualitatively from a combi-
Creep yield strength, Rce

650
200 rs+ ls
t1, 1
T( t ) 500

0 550 t2, 2
Tsalt = 450 C 400
ta = 6000 min rs
Ta = 450 C
rs, MPa

T, C
rs, MPa

200 450 Surface data 300 t2 >> t1


60 min
2 1
>>
400 200
Measured Transient relaxation 0 T 1 T2 T1
Calculated 350
6 min 100 Temperature, T
600