2 Suka0 Tidak suka

603 tayangan477 halamanResidual stresses

Jul 03, 2015

Â© Â© All Rights Reserved

PDF, TXT atau baca online dari Scribd

Residual stresses

Â© All Rights Reserved

603 tayangan

Residual stresses

Â© All Rights Reserved

- Residual Stress Analysis of Pipeline Girth Weld Joints
- Microstructure of Metal Materials
- Metallographic Handbook
- Design of Welded Connections - Lincoln Electric
- Mechanical Relaxation of Residual Stress
- Steel Heat Treatment Handboook-Metalluurgy and Technologies
- Reducing Brittle and Fatigue Failures in Steel Structures (2010)
- ASM - Failure Analysis of Engineering Structures Methodology and Case Histories
- FEM Prediction of Welding Residual Stress and Distortion in Carbon Steel [Lit.8]
- Corrosion of Weldments-J.R. Davis-0871708418-ASM International-2006-236-$199
- Time Temperature Transformation (TTT) Diagrams.pdf
- Stress Corrosion Cracking Recommended Practices 2007
- Steel Heat Treatment: Equipment and Process Design
- Method for the Determination of Residual Stress
- Residual Stress
- A Review of Residual Stress Measurement Methods
- Cold and Hot Forging - Fundamentals and Applications
- Thesis Jiawen
- 14427990 Reduction of Residual Stress
- ferrite.Bhadeshia.PMS

Anda di halaman 1dari 477

and

Deformation of Steel

Edited by

G. Totten

M. Howes

T. Inoue

www.asminternational.org

www.asminternational.org

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

www.asminternational.org

Copyright 2002

by

ASM International

All rights reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any

means, electronic, mechanical, photocopying, recording, or otherwise, without the written permission of the

copyright owner.

First printing, March 2002

Great care is taken in the compilation and production of this book, but it should be made clear that NO

WARRANTIES, EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, WARRANTIES OF

MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE GIVEN IN CONNECTION

WITH THIS PUBLICATION. Although this information is believed to be accurate by ASM, ASM cannot

guarantee that favorable results will be obtained from the use of this publication alone. This publication is

intended for use by persons having technical skill, at their sole discretion and risk. Since the conditions of

product or material use are outside of ASMs control, ASM assumes no liability or obligation in connection

with any use of this information. No claim of any kind, whether as to products or information in this

publication, and whether or not based on negligence, shall be greater in amount than the purchase price of this

product or publication in respect of which damages are claimed. THE REMEDY HEREBY PROVIDED

SHALL BE THE EXCLUSIVE AND SOLE REMEDY OF BUYER, AND IN NO EVENT SHALL EITHER

PARTY BE LIABLE FOR SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES WHETHER OR

NOT CAUSED BY OR RESULTING FROM THE NEGLIGENCE OF SUCH PARTY. As with any material,

evaluation of the material under end-use conditions prior to specification is essential. Therefore, specific

testing under actual conditions is recommended.

Nothing contained in this book shall be construed as a grant of any right of manufacture, sale, use, or

reproduction, in connection with any method, process, apparatus, product, composition, or system, whether or

not covered by letters patent, copyright, or trademark, and nothing contained in this book shall be construed as

a defense against any alleged infringement of letters patent, copyright, or trademark, or as a defense against

liability for such infringement.

Comments, criticisms, and suggestions are invited, and should be forwarded to ASM International.

Prepared under the direction of the ASM International Technical Book Committee (20002001), Sunniva R.

Collins, Chair

ASM International staff who worked on this project included Veronica Flint, Manager of Book Acquisitions;

Bonnie Sanders, Manager of Production; Carol Terman, Production Project Manager; and Scott Henry,

Assistant Director of Reference Publications.

Library of Congress Cataloging-in-Publication Data

Handbook of residual stress and deformation of steel/[edited by] G. Totten, M. Howes, T. Inoue.

p. cm.

Includes bibliographical references and index.

1. SteelFatigueHandbooks, manuals, etc. 2. Residual stressesHandbooks, manuals, etc. I. Totten,

George E. II. Howes, Maurice A.H. III. Inoue, Tatsuo, 1939TA473 .H3215 2001

620.176dc21

2001040042

ISBN: 0-87170-729-2

SAN: 204-7586

ASM International

Materials Park, OH 44073-0002

www.asminternational.org

Printed in the United States of America

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

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

www.asminternational.org

Contributors

I. Alexandru

Faculty of Materials

Technical University of Iasi

Romania

T. Inoue

Department of Energy Conversion Science

Graduate School of Energy Science

Kyoto University, Japan

H. Bhadeshia

Department of Materials Science and

Metallurgy

University of Cambridge, UK

D.-Y. Ju

Saitama Institute of Technology

Saitama, Japan

V. Bulancea

Faculty of Materials

Technical University of Iasi

Romania

V.V. Dobrivecher

Ukraine National Academy of Science

Institute of Engineering Thermophysics

Kiev, Ukraine

T. Ericsson

Linkopings Tekniska Hogskola IKP

Linkoping, Sweden

F.D. Fischer

Vorstand des Institutes fur Mechanik

Montanuniversitat Loeben

Loeben, Germany

A.V. Fomin

Leading Research Scientist

Institute Russian Academy of Sciences

Moscow, Russia

K. Funatani

Nihon Parkerizing Co. Ltd.

Nagoya, Japan

Bo Gong

Department of Metallurgy & Materials Science

University of Toronto, Canada

J. Grum

Faculty of Mechanical Engineering

University of Ljubljana, Slovenia

A.Y. Hassan

Director/Dean, School of Mechanical

Engineering

Universiti Sains Malaysia, Malaysia

K. Heess

Karl Heess GmbH

Lampertheim, Germany

F.T. Hoffmann

ITW Bremen

Bremen, Germany

R. Hoffmann

ITW Bremen

Bremen, Germany

N.I. Kobasko

Ukraine National Academy of Science

Institute of Engineering Thermophysics

Kiev, Ukraine

Z. Kolozsvary

S.C. Plasmaterm S.A.

Tg-Mures, Romania

A.I. Kovalev

Surface Phenomena Research Group

CNIICHERMET

Moscow, Russia

J. Kritzler

Metal Improvement Company, Inc.

Unna, Germany

K.-H. Lang

Institut fuf Werkstoffkunde 1

Universitat Karlsruhe (TH), Germany

R.W. Lewis

Department of Mechanical Engineering

University College of Swansea, UK

D. Lohe

Institut fuf Werkstoffkunde 1

Universitat Karlsruhe (TH), Germany

J. Lu

Universite de Technologie de Troyes

Trayes Cedex, France

T. Lubben

ITW Bremen

Bremen, Germany

M.V. Medvedev

Research Scientist

Institute Russian Academy of Sciences

Moscow, Russia

V.P. Mishina

Surface Phenomena Research Group

CNIICHERMET

Moscow, Russia

V.S. Morganyuk

Ukraine National Academy of Science

Institute of Engineering Thermophysics

Kiev, Ukraine

iii

M. Narazaki

Utsunomiya University

Tochigi, Japan

J. Pan

School of Materials Science and Engineering

Shanghai Jiao Tong University

Shanghai, P.R. China

P. Ramakrishnan

Department of Metallurgical Engineering and

Materials Science

Indian Institute of Technology

Bombay, India

I.A. Razumovsky

Leading Research Scientist

Institute Russian Academy of Sciences

Moscow, Russia

T. Reti

Banki Donat Polytechnic

Budapest, Hungary

C. Ruud

Pennsylvania State University

University Park, PA USA

G. Schleinzer

Vorstand des Institutes fur Mechanik

Montanuniversitat Leoben

Leoben, Germany

B. Scholtes

Institut fur Werkstofftechnik

Universitat Kassel, Germany

K.N. Seetharamu

School of Mechanical Engineering

Universiti Sains Malaysia, Malayasia

G.E. Totten

G.E. Totten & Associates Inc.

Stony Point, NY USA

O.Vohringer

Institut fuf Werkstoffkunde 1

Universitat Karlsruhe (TH), Germany

D.L. Wainstein

Surface Phenomena Research Group

CNIICHERMET

Moscow, Russia

H.W. Walton

Consultant

Forest City, NC USA

Z. Wang

Department of Metallurgy & Materials Science

University of Toronto, Canada

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

G.M. Webster

Union Carbide Corporation

Tarrytown, NY USA

W. Wubbenhorst

Metal Improvement Company, Inc.

Unna, Germany

www.asminternational.org

V.V. Zabilsky

Physical Technical Institute, Ural Branch of

RAS

Ijevsk, Russia

iv

W. Zinn

Institut fur Werkstofftechnik

Universitat Kassel, Germany

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

www.asminternational.org

Contents

Preface ........................................................................... vi

and Case-Hardened Components ........................................ 189

T. Reti

H.K.D.H. Bhadeshia

Z. Kolozsvary

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

J. Lu

J. Grum

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

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

D. Lohe and O. Vohringer

T. Inoue

Control of Residual Stress Formation and Steel Deformation

during Rapid Heating and Cooling...................................... 312

N.I. Kobasko, V.S. Morganyuk and V.V. Dobrivecher

and Stress Corrosion Cracking.............................................70

A.I. Kovalev, V.P. Mishina, D.L. Wainstein, and V.V. Zabilsky

and Substructure ........................................................... 331

Ioan Alexandru and Vasile Bulancea

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

H. Walton

J. Kritzler and W. Wubbenhorst

C. Ruud

Residual Stress Formation during Casting ................................ 361

R.W. Lewis, K.N. Seetharamu and A.Y. Hassan

J.Albert Sue and Gary S. Schajer

Continuous and Centrifugal Casting Processes........................ 372

D.-Y. Ju

Stress Fields ................................................................ 125

I.A. Razumovsky, M.V. Medvedev, and A.V. Fomin

Residual Stress Formation in the Shaping of Materials

W. Zinn and B. Scholtes

Z. Wang and B. Gong

P. Ramakrishnan

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

F.D. Fischer and G. Schleinzer

T. Ericsson

K. Funatani

P. Jiansheng

Effects of Process Equipment Design...................................... 183

F.T. Hoffmann, T. Lubben, R. Hoffmann, and K. Hee

v

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

www.asminternational.org

Preface

Control of steel deformation is one of the most

common concerns within the metals processing

industry. Numerous surveys have been conducted by various organizations in recent years

to assess the critical needs of the industry. In

nearly every survey that has been conducted, distortion is either the greatest or second greatest

concern among the steel heat treating community. Steel distortion control will exhibit tremendous effects on the profitability of the commercial enterprise. Therefore, it is not surprising that

the ability to understand the overall distortion

process and to be able to design solutions to this

problem typically rank very high on these same

surveys.

In view of the enormous visibility and importance of steel deformation problems, the editors

decided to put together an engineering handbook

on steel deformation. To address this subject

properly, contributing factors to overall steel deformation problems, including material effects,

machining, heating and cooling, must be examined.

This handbook contains 27 articles, divided

into five sections: Effect of Materials and Processing, Measurement and Prediction of Residual Stress and Distortion, Residual Stress Formation in the Shaping of Materials, Residual

Stress During Hardening Processes, and Residual Stress Formation During Manufacturing Processes.

There are five articles in the section Effect of

Materials and Processing. Material Factors

discusses the effects of various material properties such as thermal properties and the interactions of residual stresses on the transformation

products formed and steel deformation during

fabrication. Transformation plasticity is discussed in some detail along with the use of modeling to better understand these processes.

Prestress Engineering of Structural Material

provides a global design approach to understanding the effects of residual stress generated during

surface engineering manufacturing processes

such as PVD and CVD on the material properties

obtained. Some of the topics discussed in this

chapter include developments in the measurements of residual stresses, advanced mechanical

surface treatments, and modeling of fatigue behavior taking residual stresses into consideration.

The effect of residual stresses on fatigue behavior is discussed in detail in the next article.

Examples of topics discussed include stability of

residual stresses, some aspects of fatigue in

steels, influence of residual stresses on cyclic deformation behavior, influence of residual stresses

on crack initiation and propagation, and effect of

residual stresses on S-N curves; an overview of

modeling of the effect of residual stresses on fatigue behavior is provided.

The next article provides an overview of the

stability and relaxation behavior of macro and

micro residual stresses in steel due to thermal

and mechanical treatments. This discussion includes relaxation of residual stresses by annealing, residual stress relaxation by uniaxial deformation, and relaxation by cyclic deformation.

Hydrogen embrittlement of metals, as well as

other types of brittle fracture, result from nucleation and development of micro-cracks caused

by internal stresses. The last article in this section provides an overview of the effect of residual stress on hydrogen embrittlement and stress

corrosion cracking (SCC) of steel. This discussion includes the effect of hydrogen on structure

and transformation of steel, types of hydrogen

embrittlement, delayed fracture in steel, crack

initiation and growth, SCC of low alloy steels,

crack initiation and growth mechanism of SCC

processes, methods of estimating sensitivity to

SCC, effect of alloying elements on resistance

to SCC, and the role of structure and thermal

processing in SCC.

In the section Measurement and Prediction of

Residual Stress and Distortion, the first article

describes a number of simple, inexpensive deflection (dissection) methods used to estimate residual stress of various types of components. The

methods include Almen strip; Navy C-Ring;

plate or bar slitting and deflection; tube slitting

and opening; and bending of bars, H-beams, and

channels.

The next article provides an overview of residual stress measurement methods. Topics include residual stresses arising from various

manufacturing processes, measurement methods

including strain measuring technique, post-stress

relaxation measurement, sectioning and material

removal methods. In addition, strain measurement methods such as x-ray and neutron diffraction, ultrasonic, birefringent and laser, optical

gages, brittle coatings, Barkhausen noise, and

chemical coatings are discussed. Semidestructive methods such as blind hole drilling and ring

coring are discussed.

Measurement of residual stresses in coatings

and thin films is important because their influence on mechanical and physical properties affect component service performance. Stress Devi

measuring residual macrostress in coatings, Specific topics include origin of residual stresses in

coatings and residual stress measurement methods including the deflection method, x-ray diffraction, and hole-drilling. A comparison of

these methods is provided.

The last article in this section provides a detailed review of methods used to measure and

subsequent data analysis of inhomogeneous residual stress fields. This discussion includes residual stress as an inverse problem of experimental mechanics, indicator crack method of

measuring residual stress, arbitrary cut-out indicator method, and experimental methods and

equipment including photoelastic coating

method, and optical interferometry. Although

this is a relatively rigorous numerical discussion,

practical examples also are provided.

Residual Stress Formation in the Shaping of

Materials contains four articles. The first article

covers residual stress in the steel forming processes. The steel forming processes included are

cold forming such as wire drawing, and hot

forming such as extrusion, rolling, and forging.

The effects of residual stresses involved in these

processes are reviewed, and specific topics include residual stress in cold metal forming such

as bending of sheet, drawing of wire, rod, and

tube, and residual stresses in deep drawn cup,

sunk tubes, and radial forging products.

The effect of final shaping prior to heat treatment on residual stress formation is discussed in

the next article. The effects of shaping processes

including grinding, milling, turning, shot peening, and straightening on residual stress are discussed. Also discussed is distortion after final

part shaping and experimental and computational studies of these processes.

The next article provides a practical overview

of the factors affecting residual stress and distortion during final part shaping. Included are

discussions of influence of component shape on

heat treatment distortion, the effect of crosssection size and asymmetry, effect of heat treating procedure and machining process on final

component shape, effect of sequence of heat

treating and machining, influence of machining

allowance and stress relieving procedure, influence of residual stresses caused by cutting, methods of manufacturing blanks and effect of original structure, hot-rolled steels or forgings and

effect of banded segregation and carbide segregation, influence of heat treating methods, the

effect of heating including the rules of heating,

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

equipment and racking.

A more focused, but practical, discussion on

the effect of process equipment design on distortion follows. Subjects that are covered include

distortion generating process equipment, methods that may be used to minimize equipmentrelated distortion, quench system design, and

press quenching.

Residual Stress During Hardening Processes

contains eight articles. The first article provides

a detailed discussion on the residual stresses in

carburized, carbonitrided, and case-hardened

components. Topics include process considerations for carburized and carbonitrided components, transformations and stress evolution in

carburized and case-hardened components, effect of heat treating operations on residual stress

distribution, relationship between residual

stresses and properties of carburized parts and

modeling and prediction of residual stress field.

The article on residual stresses in steel nitriding includes a discussion of nitrided layer structure as a function of nitriding process, residual

stresses in nitrided layers, influence of residual

stresses on fatigue behavior of nitrided steel

components, and modeling and prediction of residual stresses in nitrided steel components.

The article on residual stress formation in induction hardening processes include an overview of the induction hardening process and

steels used for this process, magnetic flux concentrators, conditions in induction heating and

quenching of machine parts, residual stress surface profiles after induction surface hardening,

stress profiles in the machine part in the loaded

state, workpiece distortion in induction surface

hardening, induction surface hardening of gear

wheels, fatigue strength of materials, and residual stresses after induction surface hardening and

finish grinding.

The next article provides an overview of residual stresses and distortion resulting from reheating and quenching. Topics include phase

transformation during heat treating including

steel transformations, TTT and CCT diagrams,

metallurgical crystal structure, estimation of volumetric change due to steel transformation upon

quenching, cooling of steel with and without

metallurgical transformation, tempering, basic

distortion mechanism, relief of residual stresses,

material movement due to thermal gradients during heating and cooling, material, component

and process effects, retained austenite, quench

severity and uniformity and process design effects on distortion, quench distortion and cracking, quenchant selection, measurement and evaluation of quenching power, estimation of heat

transfer coefficient, wetting behavior and nonuniform quenching, surface conditions, and

quench process modeling and simulation of residual stress and distortion after quenching.

A detailed approach to modeling and simulation of residual stress and distortion applied to

quench processing follows. This discussion is

based on a metallo-thermo-mechanics approach,

and topics discussed include an overview of

www.asminternational.org

In the article on the control of residual stress

formation and steel deformation during rapid

heating and cooling, a particular emphasis is on

intensive quenching. This is the first detailed,

article-length discussion of this old, but littleknown technology in the western world. Topics

include mathematical model for calculation of

thermal and stress-strain state, computation of

stress-strain state, possibility of predicting hardening cracks, predicting the deformation of bearing rings during hardening, thermal stresses

formed in carburized steel products due to excessive cooling rates, generalization of computational and experimental results for heating and

cooling of parts with different geometries and

thermal and physical fundamentals of processing

of high-strength materials.

An often contradictory subject is the cryogenic processing of steels, and the detailed overview of the effect of cryogenic cooling on residual stress is presented here. Specific topics

include role of residual stresses within martensitic transformation at cryogenic temperatures,

evaluation of residual stresses after cryogenic

cooling, influence of cryogenic cooling on residual stresses and dimensional stability of steels,

and influence of cryogenic cooling on the structure and substructure of steels.

The practical use of controlled shot peening

to induce compressive residual stresses is described in detail next. This discussion includes a

historic overview of shot peening, elementary

processes of shot peening, workpiece and material process parameters, process monitoring,

process optimization, x-ray diffraction, and industrial examples.

In Residual Stress Formation During Manufacturing Processes, the first article includes an

extensive discussion of residual stress and deformation problems arising from the casting process, and modeling of residual stress formation

during casting. Discussion includes finite element analysis of heat flow during casting, formulation of the elasto-viscoplastic stress model,

and deformation of a solidifying material.

The next article describes residual stress formation during the casting process, and it includes continuous and centrifugal casting. Topics discussed include inelastic behavior and

unified constitutive theory of metallic material

in solidification, analytical method of the

thermal-mechanical problem for the casting process, residual stress formation during semicontinuous casting, residual stress formation during

centrifugal casting, and residual stress formation

during strip casting by the twin-roll method.

The origin and assessment of residual stresses

during welding or brazing is discussed next.

Welding residual stresses are discussed including residual stresses due to shrinkage, quenching, and phase transformations. Characteristic

residual stress distributions in brazed components is also discussed.

The article Residual Stresses in Powder

Metal Processing is divided into two parts. The

vii

M parts including powder characteristics, compaction in rigid dies, isostatic compaction, sintering, heat treatment of P/M parts, hot pressing,

roll compaction, powder forging, metal injection

molding, spray forming, warm compaction, and

rapid prototyping. The second part discusses residual stresses in P/M processing including

powder production, compaction of metal powders, sintering of metal powders, pressure sintering and hot isostatic pressing, heat treatment

of P/M parts, and microstructural development

and properties.

Residual Stress Formation and Distortion of

Rail Steel covers the cooling process including

the cooling boundary conditions and heat transfer, residual stress state analysis, weight and frictionthe rail end problem, experimental results;

roller straightening including residual stresses in

unused roller-straightened rails, behavior of rail

steel under plastic deformation, simulation of

roller straightening; and rails in service including residual stresses due to welding and residual

stress formation in rolling contact.

The last article provides a detailed description

of residual stress formation during hypoid gear

manufacture. It includes an overview of residual

stress formation in carburized and hardened

work, profiles and peak magnitudes of residual

stresses, measurement methods including the

Sachs hole-drilling method, x-ray and neutron

diffraction, influence of steel properties on residual stresses, influence of carburizing process parameters on residual stress formation, benefits of

residual stresses on fatigue strength, and the effects of hardness, case depth, intergranular oxidation, influence of shot peening, change of residual stresses during fatigue, and distortion of

carburized and hardened steels.

The preparation of a text of this scope was a

tremendous task. The editors are deeply indebted

to many colleagues for their patience, support,

and assistance; without them this text would not

have been possible. Special thanks go to the

ASM staff who often labor in the background

but who are vital members of the team. Particularly, thanks go to Veronica Flint and Carol Terman of ASM International for their help and encouragement.

Very special thanks go to our families for their

seemingly unending support. Without their understanding and encouragement, this project

would never have been completed.

George E. Totten, Ph.D., FASM

Editor

G.E. Totten & Associates Inc.

Stony Point, NY USA

Prof. Maurice A.H. Howes, Ph.D. (Retired)

Editor

Worcestershire, England

Prof. Tatsuo Inoue, Ph.D., FASM

Editor

Department of Energy Conversion Science

Faculty of Energy Science

Kyoto University

Kyoto, Japan

G. Totten, M. Howes, T. Inoue, editors, p3-10

DOI: 10.1361/hrsd2002p003

All rights reserved.

www.asminternational.org

Material Factors

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

interactions among time, temperature, deformation, and microstructure (Fig. 1). Material or

material-related characteristics that influence the

development of residual stress include thermal

conductivity, heat capacity, thermal expansivity,

elastic modulus and Poissons ratio, plasticity,

thermodynamics and kinetics of transformations,

mechanisms of transformations, and transformation plasticity.

Many general statements can be made about

the role of material factors in the evolution of

residual stress. Spatial variations in temperature

give rise to nonuniform thermal strains, the effect of which becomes exaggerated when the

material is elastically stiff and has a high yield

strength. A large thermal conductivity helps reduce residual stress by reducing temperature gradients (Ref 2). The dissipation or absorption of

heat depends not only on the external environment of the component but also on internally

generated heatfor example, during adiabatic

deformation or due to the latent heat of transformation. Similarly, the plastic strain distribution

in the component depends both on the constitutive properties and on how the shape deformations due to phase transformations compensate

for the development of stress.

The fundamental material properties are, of

course, temperature dependent. Table 1 illustrates how several key properties might vary

with temperature (Ref 3). Some of these properties, which can to some extent be estimated

quantitatively, are discussed in detail in the sec-

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

Heat Capacity

The dominant contribution to specific heat capacity comes from lattice vibrations (phonons),

since the majority of free electrons are prevented

from participation in heat absorption by the Pauli

exclusion principle. However, for iron and its

alloys, a further important contribution comes

from magnetic changes. The net specific heat capacity can therefore be factorized into three components:

T C

CP{T } C LV

TD

CeT C Pl{T }

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

Temperature, C (F)

Property

Phase(a)

c

P

b

Poisson ratio

c

P

b

c

P

b

Thermal expansivity, K1

Time,

temperature

1. Thermal stress

6. Latent

heat

Stress,

strain

4. Heat of

deformation

5. Stressinduced

transformation

2. Temperaturedependent phase

transformations

3. Transformation

strain

Microstructure

Fig. 1

(Eq 1)

constant pressure. Ce is the electronic specific

heat coefficient, and C lP is the component of the

specific heat capacity due to magnetism. Figure

2 illustrates the data for ferrite and austenite in

pure iron. Whereas it is well known that ferrite

undergoes a paramagnetic to ferromagnetic transition on cooling below 1042.15 K, the magnetic

properties of austenite are seen from Fig. 2 to be

of some consequence in determining the heat capacity. There are two coexisting electron states

of austenite, one of which is ferromagnetic with

a Curie temperature of 1800 K and the other of

which is antiferromagnetic with a Neel temperature of 55 to 80 K (Ref 4). The balance between

these states changes with temperature, giving

rise to corresponding changes in heat capacity.

The data in Fig. 2 are for pure iron, but there

is now sufficient understanding of the components of heat capacity to enable similar estimates

for iron alloys, using internationally available

computer programs and thermodynamic data-

c

P

b

c

P

b

c

P

b

0 (32)

200

210

210

200

0.291

0.280

0.280

0.280

15.0

49.0

49.0

43.1

4.15

3.78

3.78

3.76

190

360

440

1600

300 (570)

175

193

193

185

600 (1110)

150

165

165

168

0.309

0.296

0.296

0.296

2.1 105

1.4 105

1.4 105

1.3 105

18.0

41.7

41.7

36.7

4.40

4.46

4.46

4.45

110

230

330

1480

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

0.327

0.310

0.310

0.310

21.7

34.3

34.3

30.1

4.67

5.09

5.09

5.07

30

140

140

1260

800 (1470)

124

120

120

...

0.345

0.325

0.325

...

25.1

27.0

27.0

...

4.90

5.74

5.74

...

20

30

30

...

bases (Ref 6). After all, changes in fundamental

thermodynamic quantities such as enthalpy and

entropy are derived from heat capacity data. It is

surprising that this capability has not yet been

exploited in any calculation of residual stress,

even though the methodology is widely available.

Table 1 shows that the expansion coefficient

of austenite is larger than that of ferrite; this

might be considered surprising given the lower

density of ferrite. However, the behavior is again

c0

c1

V cm{T } (1 y)V m

{T } y V m

{T }

30

Heat capacity, CP, J mol 1K1

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

40

20

V am{T }

7.061(1 2.043 105T 1.52 108T 2)

Austenite

10

0

75

Ferrite

50

25

0

states of austenite (c0 and c1), each with a thermal expansion coefficient that is identical to that

of ferrite. The c0 component has the lower molar

volume and is the antiferromagnetic form,

whereas the denser c1 form is ferromagnetic. The

relative proportion of atoms in the c0 and c1

states changes with temperature, so that the apparent expansion coefficient of austenite as a

whole, as detected experimentally, is much

larger than that of ferrite (Fig. 3).

The molar volumes (in cm3 /mol) of c0, c1, c,

and over the temperature range of 300 to 1775

K are:

300

600

900

1200

1500 1800

Temperature, K

the c1 state, the details of which can be found

elsewhere (Ref 4, 5).

These data are for pure iron, but thermodynamic data can be used to assess how the expansion coefficients would change with alloying,

since there are quite sophisticated treatments of

the effect of solute elements on the magnetic and

other components of the free energies of iron.

Note that the two electronic states picture of

austenite is a simplification of the real scenario,

but first-principles calculations (Ref 7), which

pure iron, as a function of temperature. The thin

lines represent the combined contributions of the phonons

and electrons, whereas the thicker lines also include the

magnetic terms. The dashed vertical lines represent the Curie, c, and c d transitions. d-ferrite is simply an

alternative historical name for high temperature . Source:

Ref 5

applicable to practical alloys.

Plastic Deformation

The familiar mechanisms of plastic deformation are slip, mechanical twinning, and creep.

Phase transformations also cause permanent deformation (Ref 811). In steels, austenite can decompose into a large variety of microstructures

that are distinguished by the atomic mechanism

of transformation (Fig. 4). In a displacive transformation, the change in crystal structure is

achieved by a deformation of the parent structure. A reconstructive transformation is one in

which the change in structure is achieved by a

flow of matter, which occurs in such a way that

strains are minimized.

All the transformations cause changes in

shape (Fig. 5a), which for reconstructive transformations simply reflects the change in density.

For displacive transformations, the shape change

is an invariant-plane strain (IPS), that is, a combination of a shear on the invariant plane and a

dilatation normal to that plane. The strain energy

associated with a constrained IPS is minimized

when the product phase has a thin-plate shape.

This is why Widmanstatten ferrite, bainite, acicular ferrite, and martensite in steels grow in the

form of plates. The distinguishing features of a

variety of deformation modes are compared in

Table 2, and Table 3 describes the shape deformations.

The permanent strain caused by any transformation is called transformation plasticity. A

Fig. 2

Displacive

Reconstructive

deformation with large shear

component. No iron or

substitutional solute diffusion.

Thin plate shape

nucleation and growth.

Sluggish below about 850 K

8.0

7.8

Allotriomorphic

ferrite

7.6

V1

7.4

Widmansttten ferrite

Carbon diffusion during

paraequilibrium nucleation

and growth

Idiomorphic

ferrite

ferrite

7.2

Massive ferrite

7.0

No change in

bulk composition

paraequilibrium nucleation.

No diffusion during growth

6.8

V0

6.6

300

600

900

1200

1500 1800

Temperature, K

Fig. 3

Source: Ref 5

Fig. 4

Pearlite

Martensite

Cooperative growth

of ferrite and

cementite

Diffusionless

nucleation and

growth

Material Factors / 5

phase change in a stress-free material is usually

triggered by heat treatment, when the parent

phase passes through an equilibrium transformation temperature. Alternatively, the application of a stress in isothermal conditions can trigger transformation in circumstances where it

would not otherwise occur. Unusual effects can

occur when stress and temperature work together. The transformation may occur at remarkably low stresses or at very small deviations

from the equilibrium temperature. This is why

even minute stresses can greatly influence the

development of microstructure, and vice versa.

It is not surprising that transformation plasticity

can be obtained at stresses that are much smaller

than the conventional yield stress of the parent

phase.

and Related Phenomena

The strains due to phase transformations can

alter the state of residual stress or strain. It is well

known that the martensitic transformation of the

carburized surface of a steel component puts the

surface under compression. It is argued that this

is because of the expansion at the surface due to

formation of the lower-density martensite from

austenite.

stress. Greenwood and Johnson (Ref 13, 14)

showed that when a phase change is accompanied by a change in volume, the tensile strain

expected when transformation occurs under the

influence of a tensile stress r is given by:

e

5 DV r

6 V rY

(Eq 2)

and DV/V is the transformation volume strain.

The role of shear strains associated with transformation has been emphasized in later work by

Magee and Paxton (Ref 15, 16), and subsequently by Fischer (Ref 17), Leblond et al. (Ref

1822), Olson (Ref 23), and Bhadeshia et al.

(Ref 24). Not only does transformation affect

stress, but the latter modifies the development of

microstructure. The microstructure tends to be

more organized when transformation occurs in a

stresss parent phase, because the stress favors

the formation of certain orientations relative to

others. This is illustrated schematically in Fig.

5(b) to (d). These aspects will now be discussed

in more detail, because transformation plasticity

can radically alter the state of residual stress.

Table 2

Reconstructive

Displacive

Single

crystal

Invariant-plane strain shape changewith a large shear component

Changes crystallographic orientation

Changes lattice type

Can lead to a density change

Table 3

Allotriomorphic ferrite

Idiomorphic ferrite

Pearlite

Widmanstatten ferrite

Bainite

Acicular ferrite

Martensite

Cementite plates

Mechanical twins ()

Annealing twins (c)

Slip

deformation

Mechanical

twinning

Displacive

transformation

Reconstructive

transformation

Yes

Yes

No

No

No

Yes

Yes

Yes

No

No

Yes

Yes

Yes

Yes

Yes

Yes

No

Yes

Yes

Yes

Transformation

(a)

modes of plastic deformation. Just as a combination of a plane and a direction constitutes a

deformation system for slip or twinning, the

habit plane and displacement vector of the

invariant-plane strain accompanying displacive

transformation completely describe the deformation system responsible for transformation

plasticity. The displacement vector describes the

sense of the macroscopic displacements accompanying transformation and, along with the habit

plane indices, also contains information about

the magnitude of the shear component and dilatational component of the displacements. Typical data for the deformation systems associated

with transformations are listed in Table 4. Note

that reconstructive transformations involve only

a volume change together with diffusional mass

flow, so it is not appropriate to regard them as

deformation systems in the present context.

Given the cubic crystal structure, and the fact

that habit planes tend to be irrational, there will

in general be 24 of these systems per austenite

grain, and they may operate simultaneously to

varying extents. Of course, unlike ordinary slip,

Characteristic

Deformation System

s(b)

d(b)

Morphology

Volume change

Volume change

Volume change

Invariant-plane strain

Invariant-plane strain

Invariant-plane strain

Invariant-plane strain

Invariant-plane strain?

Invariant-plane strain

0.00

0.00

0.00

0.36

0.22

0.22

0.24

0.21?

1/2

0.00

0.02

0.02

0.03

0.03

0.03

0.03

0.03

0.16?

0.00

0.00

Irregular

Equiaxed, faceted

Spherical colonies

Thin plates

Thin plates

Thin plates

Thin plates

Thin plates

Thin plates

Faceted

(a) An invariant-plane strain here implies a large shear component as well as a dilatational strain normal to the habit plane. (b) s and d refer to the

shear and dilatational strains, respectively. The values stated are approximate and will vary slightly as a function of lattice parameters and the details

of crystallography.

(b)

(c)

(d)

transformations. Note that the horizontal scale

bars are all the same length. (a) The two kinds of shape

changes that occur when a single crystal of austenite transforms to a single crystal of ferrite, as a function of the mechanism of transformation. (b) Polycrystalline sample of austenite. (c) Polycrystalline sample of austenite that has

partially transformed by a displacive transformation mechanism into a random set of ferrite plates. (d) Polycrystalline

sample of austenite that has partially transformed by a displacive transformation mechanism into an organized set of

ferrite plates.

Fig. 5

Table 4

Phase

Displacement vector

0.185

Bainite

0.27

Widmanstatten ferrite

0.36

Martensite

Note: Typical habit plane and displacement directions for low-alloy steels. The indices all refer to the austenite phase. Note that the indices stated are

approximate, since the habit plane and displacement direction are usually irrational. The displacement vector does not quite lie in the habit plane

because the dilatational strain is directed normal to the habit plane. The magnitude of the displacement is given by m, which is the total displacement

including the shear and the dilatational components.

the different deformation systems within an austenite grain cannot intersect, except in special

circumstances where intervariant transformations are possible, as is the case with some

shape-memory alloys. It follows that the ordiN

A (applied tensile stress)

max

stress, rN, and the shear stress, s, both act on the

habit plane. The vector d is the direction along which lie

the shear displacements of the shape deformation. smax is

the maximum shear stress on the habit plane, but s is given

by resolving smax along d. Note that d differs slightly from

the displacement vector of the IPS, which includes a dilatational component in addition to the shear.

Fig. 6

driving force coefficients

DG/r, J/(mol MPa)

Nature of stress

Uniaxial tension

Uniaxial compression

Elastic crack tip (a)

0.86

0.58

1.42

Tensile

+

Compression

0

The interaction of an applied elastic stress

with a phase change can occur in two ways:

1. The stress can alter the driving force for the

transformation.

2. The stress can change the appearance of the

microstructure by favoring the formation of

those variants which best comply with the applied stress.

For reconstructive transformations, only the

hydrostatic component of stress can interact with

Hydrostatic

compression

Stress (below austenite yield strength)

Indication of how the transformation-start temperature (for Widmanstatten ferrite, bainite, acicular ferrite, or martensite) should vary as a function of the

nature and magnitude of an applied stress whose magnitude is less than that of the yield stress.

Fig. 7

temperatures in steels to applied stress

Phase

Nature of stress

Martensite

Bainite

Eutectoid

Martensite

Pressure

Pressure

Pressure

Tensile

Source: Ref 32

Sensitivity, K/MPa

0.06

0.09

0.011

0.06

1000

Free energy, J mol1

(a) The stress state for the crack tip is multiaxial, but the coefficient is

calculated by expressing the stress in terms of the von Mises equivalent

tensile stress. Source: Ref 32

Work hardening nevertheless manifests itself via

a different mechanism, in which the stability of

the austenite increases as it becomes ever more

finely divided.

The Taylor/von Mises criterion (Ref 25, 26)

states that in any given crystal, a minimum of

five independent slip systems is necessary to

produce an arbitrary shape change. A crystal in

a polycrystalline aggregate has to accommodate

the arbitrary deformations of neighboring grains.

Therefore, a polycrystalline material is brittle

unless each grain contains at least five independent slip systems. Similar logic can be applied

to the crystallographic variants of a phase generated by displacive transformation. The habit

plane is predicted theoretically (Ref 27, 28) and

found experimentally (Ref 29) to have irrational

indices. This means that there exist, in principle,

24 possible variants of the habit plane per grain

of austenite (that is, 24 independent deformation

systems). Given this large number of transformation variants available per grain, the Taylor

criterion leads to the conclusion that transformation plasticity can cause, or accommodate,

any externally imposed, arbitrary shape

changeassuming that a sufficient quantity of

parent phase is available. It follows that polycrystalline samples can remain intact at grain

boundaries when transformation plasticity is the

sole mode of deformation.

the volume change. The corresponding interaction with displacive transformations is much

larger because of the shear component of the

IPS.

For displacive transformations, the influence

of stress on the transformation can be expressed

as a mechanical driving force (DGmech), which is

the work done by the external stress in producing

the macroscopic shape deformation (Ref 30, 31):

DGmech rNd ss

(Eq 3)

and s is the component of the shear stress on the

habit plane that is parallel to the direction along

which the shear displacements of the shape deformation occur (Fig. 6). The strains d and s are

the dilatational and shear components, respectively, of the shape deformation. Some typical

values of the mechanical driving force terms are

given in Table 5. Given a free choice of some

12 to 24 crystallographic variants of the transformation product in each grain of austenite, the

work done by the shear stress is always expected

to be positive, whereas that due to the dilatational component depends on the sign of rN. For

steels, this latter component is relatively small.

Any observed consequences of stress must therefore reflect the dominant role of the shear component unless the stress is purely hydrostatic.

Since the shear stress remains positive irrespective of whether the sample is pulled in tension or uniaxially compressed, and since the

shear component of the shape change is large, a

uniaxial stress will always cause a temperature

increase for displacive transformations in steels.

Hydrostatic stress, on the other hand, has no deviatoric components and consequently interacts

only with the dilatational component of the

shape change. Thus, hydrostatic compression is

expected and found to lead to a decrease in the

transformation temperature (Fig. 7); some data

(Ref 32) on the sensitivity of the transformation

temperature to applied stress are presented in Table 6.

Limits to Stress-Assisted

Transformation

Chemical

0 Mechanical

1000

Total

2000

300

400

500

600

700

800

900

Temperature, C

Typical magnitudes of the chemical and mechanical driving forces for stress-affected transformation. The mechanical driving force is estimated for an

applied stress that is equal to the yield stress of austenite.

Since this yield stress becomes small at high temperatures,

the contribution of the mechanical driving force also decreases. Therefore, transformation becomes impossible as

the temperature exceeds about 700 C (1290 F).

Fig. 8

equilibrium transformation occurs, an applied

stress can assist reaction when the chemical driving force is insufficient to achieve the change on

its own. There must exist a point, however, when

the applied stress simply cannot provide enough

mechanical driving force to complement the

chemical term to give a driving force large

enough to induce transformation. After all, the

magnitude of the stress that can be applied is

limited by the yield point of the parent phase.

Thus, there are limits to what can be achieved

by the application of stress as a stimulus to transformation (Fig. 8).

Material Factors / 7

during welding or heat treatment. A few elegant

experiments illustrate how phase transformations interact with the buildup of residual stress.

Using bainitic, martensitic, and stable austenitic steels, Jones and Alberry (Ref 33, 34) demostrated that transformation plasticity during

the cooling of a uniaxially constrained sample

from the austenite phase field acts to relieve the

buildup of thermal stress as the sample cools. By

contrast, the nontransforming austenitic steel exhibited a continuous increase in residual stress

with decreasing temperature, as might be expected from the thermal contraction of a constrained sample.

When the steels were transformed to bainite

or martensite, the transformation strain compen-

500

AISI 316

9CrMo

2CrMo

Austenite YS

400

Stress, MPa

300

200

100

0

100

200

400

600

800

Temperature, C

(a)

Stress

Plastic

strain in

austenite

Transformation

finished

Stress due to

thermal

contraction

of austenite

Stress due to

thermal contraction

of ferrite

Transformation

begins

arose during cooling. Significant residual

stresses were therefore found to build up only

after transformation was completed and the specimens approached ambient temperature (Fig. 9).

The experiments contain other revealing features. The thermal expansion coefficient of austenite (1.8 106 /K) is much larger than that

of ferrite (1.18 106 /K), and yet the slope of

the line prior to transformation is smaller when

compared with that after transformation is complete (Fig. 9). This is because the austenite yields

to accommodate the thermal contraction, which

is possible because the yield strength of the austenite is reduced at elevated temperatures. Ferrite

is strong at low temperatures, so the slope of the

stress/temperature curve (after transformation is

complete) is steeper and consistent with the magnitude of thermal contraction strains.

Interpretation of experimental data of the kind

illustrated in Fig. 9 is difficult in the region of

the stress/temperature curve where transformation occurs. The popular view that the volume

change due to transformation is the major component of transformation plasticity is probably

incorrect for displacive transformations such as

bainite or martensite. The shape change due to

transformation has a shear component that is

much larger than the dilatational term (Table 3).

Admittedly, this shear component should, on average, cancel out in a fine-grained polycrystalline sample containing plates in many orientations (Fig. 5c). However, the very nature of the

stress effect is to favor the formation of selected

variants, in which case the shear component rapidly begins to dominate the transformation plasticity (Fig. 5d).

Bulk transformation strain, %

Residual Stress

3

75 MPa compressive stress

2

1

0

2

100

200

300

400

500

600

Transformation time, s

(b)

Temperature

a martensitic (9CrMo), bainitic (2CrMo), and austenitic steel (AISI 316). Adapted from Ref 33, 34. (b) Interpretation of the Jones and Alberry experiments. The thermal

expansion coefficient of austenite is much larger than that

of ferrite.

Fig. 9

Transformation Plasticity

When an unstressed polycrystalline sample of

austenite is transformed to plates of ferrite, the

shear caused as each randomly oriented plate

forms is canceled on a macroscopic scale; only

the volume expansion is observed experimentally. However, if the plates do not form at randomfor example, when certain variants are favored because they comply better with the

external stressthe shear strains are no longer

canceled out. Transformation will then lead to

highly anisotropic strains, as illustrated in Fig.

11. Naturally, any anisotropy will be greatest for

displacive rather than reconstructive transformations, given that the former involve large

shear strains.

Modeling Anisotropic

Transformation Strains

1

larger when the austenite finishes transformation

at a high temperature. This is because thermal

contraction strains can no longer be compensated by transformation plasticity once the austenite has decomposed. Low transformation temperatures help minimize residual stresses.

High-strength welding alloys used for making

submarine hulls therefore have transformation

temperatures of less than about 250 C (480 F).

Figure 10 illustrates one kind of distortion

found in welds, measured in terms of the angle

h through which the unconstrained plates rotate

as they cool. Table 7 shows how the distortion

depends on the temperature at which the majority of the transformation is completed, for two

manual metal arc welds deposited with a 60 Vjoint preparation in a multipass fabrication involving about 11 layers, with two beads per layer

to complete the joint. The distortion is clearly

larger for the case where the transformation is

exhausted at the higher temperature.

strain when bainite forms under the influence

of a constant, elastic applied compressive stress. Note that

the shear strain associated with the formation of one plate

is about 26%, with a volume change of about 3%. The

potential for anisotropy is therefore much greater than illustrated here.

Fig. 11

along all radial directions in a circle with the

compression axis as its diameter (Ref 35, 36).

The circle is divided into 18 equal segments (i

1 18), each segment representing a particular orientation of bainite habit plane. The

choice of 18 segments is convenient and arbitrary. The compression axis of the sample is

taken to be the z direction, the x and y directions

measured distortion (h) for two manual metal arc, multipass weld deposits

Composition, wt%

C

Fig. 10

were originally flat

0.06

0.06

Si

Mn

Ni

Mo

Cr

DT, C(F)

0.5

0.3

0.9

1.6

...

1.7

...

0.4

...

0.35

802400 (1476750)

422350 (792660)

14.5

8

being radially orientated; the unit vectors x, y,

and z define the orthonormal basis X of the sample, giving a corresponding reciprocal basis X *.

The shear and dilatational components of the IPS

accompanying the growth of bainite are approximately s 0.22 and d 0.03. Thus, the 3

3 deformation matrix describing the shape deformation is given by:

P

1 fse1 p1 f dp1 p1

fse2 p1 f dp2 p1

fse3 p1 f dp3 p1

fse1 p2 f dp1 p2

1 fse2 p2 f dp2 p2

fse3 p2 f dp3 p2

e1

p1

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

e3

p3

reduction of notation is achieved using the

MacKenzie and Bowles notation (Ref 27):

(XPiX ) I fis[X;ei]( pi ;X *)

fid[X;pi]( pi; X *)

(Eq 4)

where the subscript i identifies a particular segment of interest and X and X *, respectively, represent the real and reciprocal bases of the coordinate system in which the deformation is

described. The notation due to MacKenzie and

Bowles (Ref 27) is discussed in detail in Ref 35.

The components of the shear direction and the

dilatation direction are given by:

[X;ei] fi[cos(hi) 0 sin(hi)]

[X; pi] fi[sin(hi) 0 cos(hi)]

120

180

210

(a)

Ui

330

and the direction of the shear component of the

applied stress as resolved onto the habit plane.

To facilitate a two-dimensional analysis, the

value of i is taken to be zero. A positive value

of Ui adds to the chemical driving force (DG ac

G c G a) for transformation; a negative

value thus opposes transformation. Using these

values of interaction energies, the model can be

modified so that the segments transform in an

order of decreasing Ui. There is, however, a further complication. The effect of stress should be

largest when the interaction energy is large compared with the chemical driving force. To allow

for this, the volume fraction fi of each segment

90

60

150

30

180

210

300

270

r

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

2

120

30

(X Pi X ) [1 0 0]

i1

along the x and z directions. These are assumed

to be equal to radial and longitudinal strains eR

and eL, respectively.

It is expected that those segments that comply

best with the applied stress transform most rapidly, whereas the others do so at a lower rate, or

not at all. This can to some extent be incorporated into the model by calculating the energy

change Ui as the stress interacts with the shape

deformation of a particular variant (i). Patel and

Cohens method (Ref 30) gives:

60

150

240

(X Pi X ) [0 0 1]

i1

18

[X; x]

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

90

18

[X; z]

fse1 p3 f dp1 p3

fse2 p3 f dp2 p3

1 fse3 p3 f dp3 p3

plane of variant i and fi is the volume fraction of

bainite located in segment i.

A unit vector along the z direction changes to

a new vector z given by:

330

240

(b)

300

270

Transformation behavior inherent in the model (Ref 36). The dots illustrate the area fraction of each segment,

which in the calculations is scaled according to the value of Ui DG c. (a) Zero stress. All segments have

equal area fraction, and the order in which they transform is irrelevant. (b) Applied stress of 40 MPa and DG c 400 J/

mol. The area fractions of the segments are no longer equal. The segments in which the distance of the dot from the origin

is largest transform first.

Fig. 12

DG ac.

Note that for the model calculations, the transformation occurs with the most favored variants

growing first (Fig. 12). The model thus exaggerates the effect of stress, since in reality, for the

sort of stress levels considered experimentally,

no variant is likely to be entirely suppressed. In

addition, the grains in a polycrystalline sample

are randomly oriented, so that perfect compliance with the applied stress is impossible. Nevertheless, the trends revealed by the model are

expected to be correct.

The experimental data that need explaining,

and their interpretation in terms of the model are

summarized in Fig. 13 and may be stated as follows:

Without any stress, in a random polycrystal-

isotropic. This is easily understood since the

shear components of randomly oriented

plates tend to cancel out (Fig. 13a).

The application of the small stress at a high

transformation temperature (that is, a small

chemical driving force) causes the development of anisotropic strains, the transverse

strain first being negative and then positive

(Fig. 13b). The same effect is observed for a

large stress and low temperature (that is, a

large driving force). The model explains this

effect when it is assumed that the favored variants form first, but that the stress is not large

enough to suppress the eventual formation of

other variants. The signs of eL and eR are always opposite for the favored variants, but

are identical for the rest of the variants.

Therefore, the transverse strain is initially

negative but then becomes positive as transformation progresses. The low-stress/hightemperature situation is equivalent to the

high-stress/low-temperature case because in

both of these circumstances, variants that are

not favored cannot be suppressed. In the former case the stress is too small for suppression, whereas in the latter case the chemical

driving force is too large to permit suppression.

When a large stress is applied at a high temperature, the favored variants dominate.

Therefore, the strains are always of opposite

sign (Fig. 13c).

The model is thus capable of qualitatively explaining all the essential features of the formation of bainite under the influence of a small tensile stress. A uniaxial compressive stress (as

used in the experiments described below) simply

causes a reversal of the signs of the longitudinal

and transverse stresses; there is also a minor effect from the unfavorable interaction between

the compressive stress and the dilatational component of the IPS shape deformation.

The most interesting conclusion to emerge

from comparison of the model with experimental

data is that transformation under the influence of

a mild stress occurs sequentially. Variants that

comply with the applied stress grow first, fol-

Material Factors / 9

lowed by those that do not. This also carries the

implication that the interaction of the stress is

with the growth process (that is, the IPS shape

deformation) rather than the strain field of the

nucleus, which is likely to be different. It is

worth noting that there are similar results for

martensite: most favored variants grow first in

the sequence of transformation under stress (Ref

15, 24).

Summary

Many of the thermal properties of steelsfor

example, heat capacity, thermal expansion coefficients, and latent heats of transformation

are remarkably well understood. Indeed, commercially available thermodynamic databases

and programs can be used to estimate these

quantities as a function of temperature and

chemical composition. This capability has not

been exploited in the analysis of residual

stresses, even though phase diagram calculations

using the same software are now routine in industry and academia.

Other properties, such as elastic modulus, are

the case that they are insensitive to alloying, but

that remains to be demonstrated in the context

of residual stress analysis.

There is little doubt that transformations in

steel play a major role in the development of

residual stresses. For reconstructive transformations (for example, pearlite), it is the difference

in density between the parent and product phases

that contributes to transformation plasticity. The

plasticity can be much larger for displacive

transformations (Widmanstatten ferrite, bainite,

martensite) because of the large shear component of the shape deformation when these transformation products form. These are quite sophisticated effects which, with few exceptions,

are not incorporated in most residual stress analyses.

3.

4.

5.

6.

7.

8.

REFERENCES

9.

1. T. Inoue and Z. Wang, Mater. Sci. Technol.,

Vol 1, 1985, p 845850

2. P.W. Fuerschbach, in The Metal Science of

Joining, M.J. Cieslak, J.H. Perepezko, S.

10.

11.

12.

L

T

+

L, T

Strain

Strain

13.

14.

Transformation time

15.

16.

Equivalent time

(a)

17.

L

18.

Strain

Strain

L

0

19.

20.

21.

Equivalent time

Transformation time

(b)

22.

L

23.

+

Strain

Strain

0

Transformation time

24.

0

T

25.

Equivalent time

26.

(c)

27.

Schematic of the reported variations (Ref 24) in longitudinal and radial strains during the isothermal formation

of bainite under the influence of a tensile load, presented alongside predictions (Ref 36) from the crystallographic/thermodynamic model. The stresses are all intended to be below the austenite yield strength, and the data in this

case refer to uniaxial tension. (a) Zero stress, any temperature. (b) Small stress, low temperature. (c) Small stress, high

temperature; or large stress, low temperature.

Fig. 13

28.

Metals and Materials Society, 1992, p 21

30

R. Schroder, Mater. Sci. Technol., Vol 1,

1985, p 754764

L. Kaufman, E.V. Clougherty, and R.J.

Weiss, Acta Metall., Vol 11, 1963, p 323

335

L. Kaufman, in Energetics in Metallurgical

Phenomenon, Vol III, W.M. Mueller, Ed.,

Gordon and Breach, 1967, p 5584

K. Hack, Ed., The SGTE Casebook: Thermodynamics at Work, Institute of Materials,

1996

D.G. Pettifor and A.H. Cottrell, Ed., Electron Theory in Alloy Design, Institute of

Materials, 1992

J.W. Christian, in Decomposition of Austenite by Diffusional Processes, V.F. Zackay

and H.I. Aaronson, Ed., Interscience, 1962,

p 371386

H.M. Clark and C.M. Wayman, Phase

Transformations, American Society for

Metals, 1970, p 59114

J.D. Watson and P.G. McDougall, Acta Metall., Vol 21, 1973, p 961973

J.W. Christian, Physical Properties of Martensite and Bainite, Special Report 93, Iron

and Steel Institute, 1965, p 119

H.K.D.H. Bhadeshia, Bainite in Steels, 2nd

ed., Institute of Materials, 2001, p 1453

G.W. Greenwood and R.H. Johnson, Proc.

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

R.H. Johnson and G.W. Greenwood, Nature, Vol 195, 1962, p 138139

C.L. Magee, Ph.D. thesis, Carnegie Mellon

University, 1966

C.L. Magee and H.W. Paxton, Trans. Met.

Soc. AIME, Vol 242, 1968, p 17411749

F.D. Fischer, Acta Metall. Mater., Vol 38,

1990, p 15351546

J.B. Leblond, G. Mottet, and J.C. Devaux,

J. Mech. Phys. Solids, Vol 34, 1986, p 395

409, 411432

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

Int. J. Plasticity, Vol 5, 1989, p 551572

J.B. Leblond, Int. J. Plasticity, Vol 5, 1989,

p 573591

J.B. Leblond and J. Devaux, Residual

Stresses, Elseviers, 1989, p 16

J.B. Leblond, Internal Report CSS/L/NT/

90/4022, FRAMASOFT, 1990, p 112

G.B. Olson, Deformation, Processing and

Structure, American Society for Metals,

1982, p 391424

H.K.D.H. Bhadeshia, S.A. David, J.M. Vitek, and R.W. Reed, Mater. Sci. Technol.,

Vol 7, 1991, p 686698

R. Von Mises, Z. Angew. Math. Mech., Vol

8, 1928, p 161

G.I. Taylor, J. Inst. Met., Vol 62, 1928, p

307

J.K. MacKenzie and J.S. Bowles, Acta Metall., Vol 2, 1954, p 138147

M.S. Wechsler, D.S. Lieberman, and T.A.

Read, Trans. Amer. Inst. Min. Metall. Eng.,

Vol 197, 1953, p 15031515

29. A.B. Greninger and A.R. Troiano, Trans.

Amer. Inst. Min. Metall. Eng., Vol 140,

1940, p 307336

30. J.R. Patel and M. Cohen, Acta Metall., Vol

1, 1953, p 531538

31. L. Delaey and H. Warlimont, in Shape

Memory Effects in Alloys, J. Perkins, Ed.,

TMS-AIME, Plenum Press, 1975, p 89114

Beck, Mater. Sci. Technol., Vol 1, 1985, p

805814

33. W.K.C. Jones and P.J. Alberry, Ferritic

Steels for Fast Reactor Steam Generators,

British Nuclear Engineering Society, 1977,

p 14

34. W.K.C. Jones and P.J. Alberry, Residual

Institute, 1977, Paper 2

35. H.K.D.H. Bhadeshia, Worked Examples in

the Geometry of Crystals, Institute of Metals, 1987

36. A. Matsuzaki, H.K.D.H. Bhadeshia, and H.

Harada, Acta Metall. Mater., Vol 42, 1994,

p 10811090

G. Totten, M. Howes, T. Inoue, editors, p11-26

DOI: 10.1361/hrsd2002p011

All rights reserved.

www.asminternational.org

Material: A Global Design Approach

to the Residual Stress Problem

J. Lu, LASMIS, Universite de Technologie de Troyes, France

which influences its fatigue behavior, fracture

strength, and even its corrosion resistance. Few

metalworking methods exist that do not produce

new stresses. The role of residual stress is, therefore, very important when designing mechanical

parts. Over the last few years, an increasing number of studies have been carried out to understand the effects of residual stress on mechanical

performance. This article attempts to present a

global approach to including residual stress in

expected fatigue life calculations, and the possibility of introducing it into mechanical engineering design offices. The definitions and origins of residual stress according to production

methods are first presented. Then, shown are the

problems involved in correctly adapting these

modeling techniques for use in design offices

and the industrial consequences of taking residual stress into account on quality assurance control procedures. This article deals mainly with

residual stress measurement techniques and the

overall necessity to combine destructive (incremental hole-drilling method) and nondestructive

(x-ray and neutron diffraction) methods in order

to precisely evaluate the residual stress distribution. Some new results concerning optical

methods are also discussed. Shown, too, are the

beneficial and harmful effects of residual stress

on the resistance of structures or industrial components, depending on whether they are tensile

or compressive. The different models used to

predict residual stress induced by different types

of processing are then presented. The last section

shows the effect of residual stress on fatigue behavior. A model based on the finite element

method (FEM) for predicting the relaxation of

residual stresses is presented. Finally, prediction

of fatigue life, taking residual stress into account,

using FEM is discussed. A new approach to concurrent engineering applied to the design of mechanical components with residual and applied

stress consideration is presented.

that remains in mechanical parts that are not subjected to any outside stresses. Residual stress exists in practically all rigid parts, whether metallic

or not (wood, polymer, glass, ceramic, etc). It is

the result of the metallurgical and mechanical

history of each point in the part and the part as

a whole during its manufacture.

In the case of structural materials, surface engineering can lead indirectly to the innovation of

genuinely new materials based on conventional

materials. The frustrating slowness of some of

these developments reflects not only difficulties

in scaling up laboratory techniques, but also general conservation in the engineering industries

and a reluctance to change established habits.

One of the factors that contributes to slowing the

pace is that the current approach to surface engineering directly correlates performance to processing parameters; designers are left without the

tools needed to optimize and take into account

modified or prestressed surfaces for a particular

application, and the processors are left without

appropriate goals for surface properties.

The recent European Network of Surface and

Prestress Engineering and Design (ENSPED)

project led by the University of Technology of

Troyes and funded by the European Union is one

of the contributions currently being made to the

development of such tools. The aim is to develop

a project with a hard core of about 18 partners

representing a balanced selection of Europes

major industries, such as SNECMA; European

Aeronautic, Defense and Space (EADS); ABB;

Siemens; Volvo; Fiat; British Aerospace; Robert

Bosch; Hydro Aluminum; Wartisla; and so on.

It is based on building a bridge between the

fields of surface modification and prestress processing in materials engineering and computeraided design in mechanical engineering in order

to find an appropriate way of approaching this

interdisciplinary area. The members of the project will focus on the field of prestress processing. The main goal of the prestress engineering

approach is the optimization of residual stress for

and structure. Under this approach, the residual

stress must not be considered as a parameter that

only depends on the material processing conditions, but must also be considered as a parameter

that can be optimized.

There is an increasing interest in how the state

of residual stress affects the mechanical properties of a material and its structure. The failure of

a structure or a mechanical component is not

only due to external loads. Residual stress is an

important parameter in this respect. All manufacturing processes, for example, introduce a

new state of residual stress. This can have a positive effect, such as increasing the fatigue limit in

the case of surface compressive stress, or it can

have a negative effect, such as decreasing the

stress corrosion behavior of a material in the case

of tensile residual stress.

Basic and applied research in the field of residual stress has been stepped up in the last few

years. Residual stress is taken into account in

advanced design in the aerospace, automotive,

and nuclear industries. Even the microelectronics industry is starting to take residual stress into

account for the dimensional stability of electronic packaging.

The introduction of advanced materials has

also contributed to the development of knowledge in the field of residual stress. In fact, many

new materials are multimaterials, for example,

metal-matrix composites, plasma-sprayed coating, physical vapor deposition (PVD), and chemical vapor deposition (CVD) coatings, which

contain residual stress as a result of the thermal

and mechanical incompatibilities of the different

phases of the material or structure.

Figure 1 shows the different fields of research

in which residual stress is taken into account and

its relevance for industrial applications. Three

main fields must be developed for a global approach of prestress engineering: measurement

techniques for the quality control and processing

analysis, processing parameter optimization and

processing modeling, and modern design tool for

the life cycle simulation with residual stress consideration.

In general, macroscopic residual stress can be

induced due to:

Nonhomogeneous plastic flow under the ac-

tion of external treatment (shot peening, autofretting, roller burnishing, hammer peening,

shock laser treatment)

Nonhomogeneous plastic deformation during

nonuniform heating or cooling (ordinary

quenching, molding of plastics)

Structural deformation from metalworking

(heat treatment)

Heterogeneity of a chemical or crystallographic order (nitriding or case hardening)

Various surface treatments (enameling,

nickel plating, chrome plating, PVD and

CVD coating)

Differences in expansion coefficients and mechanical incompatibility of the different components of composites (composites with a

metallic and organic matrix, ceramic coatings)

stress for metalworking operations usually carried out in the industry. To produce an industrial

part, one or several of the techniques listed in the

table can be used. To calculate the residual stress

existing in a part, the source of the stress must

be identified first.

Residual Stress Measurement Techniques

in Global Approach and Quality Assurance.

Over the last few decades, various quantitative

and qualitative techniques have been developed

(Ref 1). These techniques are used for the processing optimization and quality control of material. In general, a distinction must be made between destructive and nondestructive methods.

The first series of methods is based on destroying the state of equilibrium in the mechanical component. The residual stress is then evaluated from its relaxation. However, it is only

possible to measure the consequences of stress

relaxation and not the relaxation itself (displacement, fracture, strain). In most cases, the change

in strain is selected as the parameter to be studied. The following procedure is used:

1. Creation of a new stress state by machining

or layer removal

2. Detection of the local change in stress by

measuring the strain or displacement

3. Calculation of the residual stress as a function

of the strain measured using the elastic theory

(analytical approach or numerical calculations such as FEM)

During the recent years, the incremental holedrilling method is extensively used. It is sensitive to the first kind of residual stress, that is, the

macroscopic residual stress. The principle of this

technique is simple. It involves monitoring the

change in strain when a hole is drilled in a component with residual stress. These strain measurements can be related to the original residual

stress distribution in the analyzed sample at the

hole location. The relationship between the

strain and the residual stress can be calculated

with the calibration coefficients Ain and Bin. The

general approach used to determine the Ain and

Bin FEM is detailed in Ref 2. Recently, the high

sensitivity moire interferometer and incremental

hole-drilling method for residual stress measurement has been developed. The theoretical development of a combined method is introduced in

Ref 3 and 4. The relationship between the threedimensional surface displacements produced by

introducing a blind hole and the corresponding

residual stress is established by employing the

existing theoretical solution containing a set of

undetermined coefficients. The coefficients are

calibrated by the three-dimensional finite ele-

Quality control

Manufacturing

Measurement

technique

Prestress process

development and

modeling

account in integrated design

Relaxation of

residual stress

Dimensional

stability

Fig. 1

Effect of residual

stress on behaviors

New design

Mechanical

design

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

ment method. The in-plane surface displacements Ux and Uy and out-of-plane surface displacement Uz produced by the relaxation of

residual stresses are obtained using moire interferometry and Twyman-Green interferometry,

respectively. Figure 2 shows three-dimensional

displacement as a function of the drilling depth.

The main advantage of this technique is the possibility of studying the residual stress with an inplane stress gradient (Ref 5). It is also possible

to study composite materials (Ref 6). In this

case, the through thickness residual stress distribution was evaluated ply by ply in a carbon fiber/epoxy composite. Figure 3 shows an example of residual stress determined by using this

technique.

Recently, this method was used (Ref 7) to

study the Plastic Ball Grid Array (PBGA) package, which is a cost-effective surface-mounting

package with a high-density interconnection,

low profiles, and light weight. It is currently used

in many electronic products, including portable

telecommunication and computing products. A

typical structure in a PBGA package consists of

four layers: a plastic molding compound, a silicon chip, a chip-attach adhesive layer, and an

organic chip carrier. Due to the coefficient of

thermal expansion mismatch between the silicon

chip, the plastic compound, and the organic chip

carrier, considerable residual stresses are developed in the package during the assembly process. The process-induced residual stress can

play a significant role in the reliability of electronic components and packages. Since a PBGA

package is small and the surface layer is made

of a plastic material, it has proved very difficult

to use other existing methods of residual stress

measurement. In this research work, a practical

method has been developed to determine residual stress for electronic packaging. In this

method, blind holes are drilled into the specimens, and relationships are established between

the released surface displacement and the corresponding residual stress by introducing a set

of calibration coefficients. A multilayer three-dimensional FEM is established to determine the

relevant calibration coefficients. The surface displacements are measured accurately in a small

region around the hole. For a practical PBGA

package, the tensile residual stress is determined

in both the plastic molding compound and the

glass/epoxy laminate chip carrier. The method

is accurate, simple, convenient, and practical.

More applications in the field of electronic products are anticipated.

The x-ray diffraction and neutron diffraction

methods are based on the measurement of lattice

strains by studying variations in the lattice spacing of the polycrystalline material. The first

method measures the residual strain on the surface of the material, and the second measures the

residual strain within a volume of the sample.

Diffraction techniques can be used to study all

three types. The peak shift method is sensitive

to the first two, whereas line broadening is sensitive to the second and third types. These techniques have been used to measure the first type

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 13

of residual stress in different phases of advanced

materials, such as metal-matrix composites (in

matrix and in reinforcement) (Ref 8, 9).

Residual stress can be incorporated into the

design of mechanical components. Although this

leads to a better knowledge of the fatigue life of

parts and reduces the safety coefficient at the design stage, it also poses a host of new problems

on a quality assurance level. All statistical controls are only applied today to a few critical components in the aeronautical and nuclear industries; this practice could easily become

Fig. 2

stress must therefore be developed. The methods

used industrially (x-ray diffraction and the incremental hole method) will not be sufficient in the

future. Other nondestructive testing (NDT) techniques (ultrasound, magnetic methods, acoustic

emission) are presently being developed. But as

they currently stand, these techniques use physical parameters that depend not only on the residual stress present in the parts, but also on microstructural changes. In the near future, NDT

techniques will be applied at the same time as

Three-dimensional displacements measured for different hole depth by a combination of optical methods and

the incremental hole-drilling method in a shot peened sample. Champ, field

Table 1

the reference techniques. Figure 4 proposes a residual stress inspection plan. Other new techniques, such as neutron diffraction and optical

method, will be introduced in industry. A detailed analysis of different techniques of measurement is out of the scope of this article.

Modeling of Process

The experimental techniques developed previously can contribute to the development of a

residual stress prediction model. The results calculated by the model make it easier to take the

residual stress into account during the mechanical design. The residual stress induced by the

thermal processing is extensively treated in the

other articles of this Handbook. Only some examples concerning the mechanical surface treatments and the comparison between the numerical simulation and experimental validation are

presented briefly.

Two kinds of models can be mentioned: analytical models and numerical models. In this

case, mechanical surface treatment models were

developed for shot peening (Ref 10) and cold

rolling (Ref 11), while several finite element

codes were used or developed for welding,

grinding, heat treatment (quenching), and thermal cutting (Ref 12, 13). The results show good

correlation between the prediction of the model

and the experimental results. But it can also be

seen that three-dimensional calculations are necessary to obtain good results in all directions. If

a two-dimensional calculation is used, the residual stress evaluation correlates well in one direction only. So, in the future, three-dimensional

calculations will be of greater significance for

real case modeling (Ref 13).

In the case of shot peening, a three-dimensional finite element dynamic model was developed to obtain a better description of the shotpeening process and to introduce this approach

into the component design (Ref 14). With the

Process

Casting

Shot peening, hammer peening, roller burnishing, laser

shock treatment, bending, rolling, chasing, forging,

straightening, extrusion

Grinding, turning, milling, drilling, boring

Mechanical

Thermal

Structural

No

Heterogeneous plastic deformation

between the core and surface of the

part

Plastic deformation due to the removal

of chips

No

No

No

Phase transformation

Depends on the material

during machining

Temperature gradient

Temperature gradient

the temperature is sufficiently high

No

Change of volume due to a phase change

No

Thermal incompatibility

Welding

Brazing

Electroplating

Shrinkage

Mechanical incompatibility

Mechanical incompatibility

Temperature gradient

Thermal incompatibility

Mechanical incompatibility

Mechanical incompatibility,

microcracking

Mechanical incompatibility

Mechanical incompatibility

gradient

Mechanical incompatibility

Mechanical incompatibility

modification

Microstructural change (HAZ)

New phase at interface

Composition of plating depending on bath

used

Change of phase in plating

Surface quenching with a phase change (induction, EB,

laser, plasma, classical methods)

Case hardening, nitriding

PVD, CVD

Composite

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

Change of phase

No

help of three-dimensional modeling, the enormous influence of shot interaction is verified by

simulating simultaneous impacts. This simulation is very similar to the industrial process. Results of the residual stress obtained by the simulation are closer to the experimental results, if a

three-dimensional FEM is used (Fig. 5). The

higher the coverage rate, the lower the intensity

of the stress. This study shows how FEM can be

used to model shot peening and determine the

associated residual stress field.

The results obtained for both a single impact

and several impacts follow the general distribution of measured residual stress fields. This type

of analysis seems promising for studying the repercussions of a large number of parameters.

However, these methods are only efficient when

a maximum number of experimentally measured

parameters can be introduced. They also need to

be improved if they are to take most phenomena

into account. The proposed models could be

completed by analyzing the influence of the type

of contact, the direction of the shot during the

impact, behavior of the material, and so on. It

could then be possible to introduce the residual

stress field into metal parts in order to study their

behavior, using numerical simulation, for several

types of mechanical and fatigue life tests.

70

60

Average in two laminas of fibers

Average in two laminas of epoxy

50

40

Residual stress, MPa

30

20

10

0

112.5

225

337.5

450

562.5

675

787.5

900

10

20

Mechanical Strength of Materials

Generality. When a part is subjected to a field

of elastic residual stresses characterized by a tensor rR, on which is superposed a field of service

stresses defined by the tensor rS, the real stress

to which the part is subjected is characterized by

the tensor rR rS (Fig. 6). If the residual

stresses are added to the service stresses (residual tensile stress, for example), the part is locally

overloaded due to residual stress. If, on the contrary, an appropriate finishing operation (shot

peening or roller burnishing, for example) is

used to introduce residual compressive stress,

the part is relieved of some of the load locally

and the mechanical performance of the materials

is increased as a result.

Figure 7 shows the properties of materials that

are influenced by residual stress. In the subsequent section, several quantitative examples of

the effect of residual stress are given.

Influence on the Fatigue Strength (Initiation Crack Phase). Residual stress plays an extremely important role with respect to the fatigue

strength of materials. It can be considered to be

a mean or static stress superimposed on the cyclic stress. As the mean stress rm increases, the

fatigue strength decreases. This is demonstrated

in the Haigh and Goodman diagrams.

30

y

40

50

Hole depth, m

Fig. 3

S

R + S

I M

Mechanical components

manufacturing lines

NDE techniques (US,

magnetic, AE)

Fig. 6

100

Good

quality

50

Defect

Assembly

of parts

Statistical analysis

of results

0

Measurement of the residual

stress distribution through

the thickness of the part

using conventional techniques

(x-ray, hole drilling)

0.10

0.20

0.30

Fatigue

0.40

50

100

Corrosion

under tensile

stress

150

200

Residual

stress

250

300

Breaking

srr L = 0.3 mm

szz L = 0.3 mm

Friction

Tensile

strength

350

400

Improvement of

process

Residual stress inspection plan for the purposes

of quality assurance. NDE, nondestructive evaluation; US, ultrasonic; AE, acoustic emission

Fig. 4

450

Depth, mm

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

Fig. 5

Dimensional

stability

Fig. 7

Interfacial

bond strength

of coating

materials

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 15

leads to a tangential residual stress equal to or

slightly greater than the longitudinal stress. The

thickness of the material subjected to residual

compressive stress is in the same order of magnitude as the layer transformed during treatment

(Ref 15). Fatigue tests were carried out by the

French Technical Center for Mechanical Industry (CETIM) on 36 mm diameter XC42 steel cy-

introduces very high residual compressive stress

into the hardened layer, which results from the

increase in volume of the martensitic structure

with respect to the ferrito-pearlitic structure (this

applied to the treatment of annealed steel, for

example). In induction-quenched cylindrical

bars, the residual stress on the surface usually

Table 2

at 45 HRC, mm (in.)

Surface

hardness, HRC

at the fatigue limit, MPa (ksi)

Fatigue limit

after 5 106

cycles, MPa (ksi)

rm

ra

5556

596 (87)

584 (85)

5556

623 (90)

610 (88)

D, water quenched after through

heating without stress-relieving

annealing 3.5 (0.14)

5459

6061

670 (97)

780 (113)

660 (96)

750 (109)

Longitudinal

stress

Transverse

stress

128 (19)

243 (35)

273 (40)

341 (49)

655 (95)

863 (125)

777 (113)

468 (68)

571 (83)

583 (85)

676 (98)

603 (87)

1132 (164)

1156 (168)

Source: Ref 16

Sample type

2.10, MPa

State

Crack

initiation site

BX, MPa

As welded

207

40

207

37

Shot peened

392

519

Weld toe

1

Multipass

Fig. 8

Effect of residual stress on the fatigue strength of E690 welded joints. rRX, residual stress perpendicular to the

fillet (rRX); rRI, principal maximum residual stress evaluated at a depth of 0.1 mm below the surface. Source:

Ref 17

r rR rf*

106

ZAT D.T.

(40-75 mm)

ZAT B.S.

(40 mm)

Nontreated metal

da/dN, m/s

107

108

Rs = 0.1

E36Z

109

10

20

30

40

50

60

70

80

K, MPa m

Fig. 9

Ref 18

and subjected to repeated bending stress (Ref

16). The results obtained are presented in Table

2. It can be seen that the higher the residual compressive stress, the greater the fatigue strength.

The resulting gain in fatigue strength produced

by the residual stress can be as much as 50% of

the fatigue strength of the base material treated.

Figure 8 shows the effect of residual stress on

the fatigue strength of welded HLE (E690) steel

joints (Ref 17). Three cases are show: as-welded

(residual tensile stress), stress relieved (no residual stress), and shot peened (residual compressive stress). A marked increase in the fatigue

strength was observed in the case of shot peening.

Influence on Fatigue Failure (Propagation

Phase) and Brittle Fracture (Ref 13, 18, 19).

In the case of welded assemblies, the presence

of welding defects at the weld toe and the geometric profile of the latter generally lead to a

limited period of crack initiation. The cracking

phase must be considered by taking into account

the residual stress field induced by the welding

operation.

The decisive influence of the residual stress

field on the crack propagation speed has been

demonstrated (Ref 18). Figure 9 shows the results of cracking as a function of the residual

stress. Relieving residual stress by heat treatment

changes the crack propagation speed considerably when the stress is high.

In the case of a brittle fracture, cleavage starts

in a grain when the local stress reaches a critical

value of rf* and it generally propagates without

difficulty in the adjacent grains by producing a

brittle fracture. The tensile residual stress rR, in

addition to the applied stress r, initiates this type

of failure for low loads, such that:

Effect of a residual stress-relieving treatment on the cracking speed in the HAZ (butt-welded assembly of an

E36Z steel). ZAT D.T., HAZ of the heat treated sample; ZAT D.T., HAZ of the as-welded sample. Source:

stress alone can be enough to allow propagation

to continue at a high speed. Failure is therefore

very sudden. Residual stresses that facilitate the

initiation of brittle fracture by cleavage are therefore very dangerous for steels under load at low

temperature. This is why the stress relieving of

welded joints is also recommended.

Grain slips come up against inclusions and

create concentrated stresses at their interface that

lead to fracture of either the interface or the inclusion. Cavities then appear for a critical initiation stress and grow by plastic deformation of

the matrix until their coalescence leads to ductile

fracture at least on a microscopic level. The

speed at which the cavity grows is not only proportional to the plastic deformation speed but

also to the degree of triaxial state of the stresses

and to the ratio of the mean stress to the ultimate

stress. Coalescence is a plastic instability phenomenon that no doubt occurs for a critical cavity size. Tensile residual stress not only facilitates the initiation of cavities but, by increasing

the mean stress, also accelerates growth. These

two effects combine to decrease the critical elon-

gation of ductile fracture. However, this is only

important if the ductility is already very low in

the absence of residual stresses, since plastic deformation can eliminate them.

Effect on Stress Corrosion (Ref 20, 21).

Stress corrosion is a mechanical and chemical

cracking phenomenon that can lead to failure under the combined effect of tensile stress and a

corrosive environment. Cracking is generally

transcrystalline and can appear on all types of

materials, such as aluminum alloys, steels, copper, titanium and magnesium. The introduction

of residual compressive stress can considerably

increase the fatigue life of parts subjected to

stress corrosion. Tests carried out on magnesium

test specimens placed under stress in a salt solution gave the following results:

used, can remain in the coatings and in the substrates. They are of several types: microstresses

in the grain, produced during cooling, and

macrostresses affecting the entire coating.

Macrostresses are created not only by cooling

but also by the difference in temperature between the substrate, the sprayed layer, and the

outside surface. The differential contraction thus

produced between the various materials, due to

the difference in physical and mechanical properties, determines the stresses in the coating and

the coating-substrate interface. These stresses

therefore influence the mechanical and thermomechanical behavior of the coated parts.

In order to appreciate the quality of a coating,

three types of damage to parts in service can be

considered:

The properties of the substrate are modified

utes

Shot peened test specimen: no cracking after

12 days under the same conditions

The tests conducted by W.H. Friske show that

the fatigue life is 1000 times greater for a shotpeened 304 grade stainless steel part than it is

for a non-shot-peened part (Ref 20). Tests carried out by CETIM on Z6CN18.9 stainless steel

produced similar results (Ref 21).

Effect on Adhesion of Coatings (Ref 22

25). Most coatings are produced for a specific

reason, particularly to improve the corrosion and

wear resistance of the base material, or to provide a thermal barrier for use at high temperature. But this is only achieved if the coating adheres to the substrate correctly. Adhesion

therefore indicates correct preparation of the surfaces to be coated and the quality of the coating

operation. The last few years have seen the appearance of plasma-spraying techniques, both at

atmospheric pressure and at reduced pressure.

These processes offer a high degree of flexibility

for coatings in critical areas. However, high residual stress, inherent to the coating method

R

Kc

7

6

300

200

4

3

100

2

1

0

Kc, MPa m

1

2

TT

STT

100

stress and the toughness of the interface: case

of plasma-sprayed coatings at atmospheric pressure. TT,

heat treated; STT, as-sprayed

Fig. 10

by the coating.

the adhesion and the fatigue life.

C. Richard et al. have shown that decreasing

the residual stress by thermal treatment of the

coating considerably improves adhesion at the

interface (Ref 23). Figure 10 illustrates the effect

of residual stress. It can be seen that the apparent

toughness of the interface is improved by 100%

when heat treatment is applied. There is a high

level of residual tensile stress in the test specimen without heat treatment. When the level of

residual tensile stress increases, the true toughness of the coating decreases. An increase in the

residual compressive stress produces the opposite effect.

Influence of Residual Stress on the Tensile

Strength, Friction, Wear and Dimensional

Stability. The effect of residual stress on the tensile strength is obvious, particularly in structures

made of composite materials or when the prestressed layer is very thick compared with the

thickness of the parts. In composites, residual

stress is produced as a result of the thermal and

mechanical incompatibility of the reinforcements and matrix. This can influence the macroscopic properties of composites under tensile

or compressive stress (Ref 26).

Little research has been carried out on the effect of residual stress on friction and wear properties. Their role is often masked by other parameters. The increase of hardness during treatment

and changes in the toughness and adhesion of

antiwear coatings due to residual stress can considerably affect the resistance to friction. Up until the present, this effect has been integrated into

the global parameter of adhesion. In the future,

work will be carried out to try to determine the

real effect of residual stress.

The problem of dimensional stability has been

known for a long time. When a part is machined

that contains residual stress produced by heat

treatment or welding, the shape of the part can

change after operation due to the relaxation of

residual stress. This is why stress-relieving treatments are frequently used to avoid this type of

defect. Reference 27 gives a very methodical approach to defining the criteria and processes relating to relieving stress in welded structures.

The same type of reflection can be applied to

other types of parts.

when Calculating the Expected

Fatigue Life

In the previous section, the different effects of

residual stress on the mechanical strength of

structures and materials were mentioned. Although the ability to quantitatively estimate the

fatigue life taking residual stress into account is

just beginning, it is still too early to extend these

predictions to other types of stress that are far

more complex and involve physical and chemical phenomena. Statistics show that failures of a

purely mechanical origin are mainly due to fatigue. It is for the reasons indicated previously

that this article only addresses problems concerning the prediction of fatigue life. Two articles (H. P. Lieurade and A. Pellissier-Tanon) in

Ref 13 deal with the question of predicting the

effect of residual stress on crack propagation

phase. Although, they concern welded structures, the concepts developed in these two references can be applied to other types of structures. By limiting the approach to prediction of

the fatigue life to the fatigue cracking initiation

stage, the problem of predicting the fatigue life

of mechanical components subjected to a high

cycle fatigue can be analyzed.

Calculating the Effect of Residual Stress on

the Fatigue Strength. Based on the experimental results mentioned previously, it would seem

that a linear relationship of the Goodman type

can be used to take residual stress into account:

ra rD

rD

(rm rR)

Rm

(Eq 1)

rm is the mean fatigue stress, rD is the purely

reverse tensile fatigue limit, Rm is the true rupture strength, and rR is the residual stress measured in the direction of the applied service

stress. The numerous studies mentioned in Ref

28 show that the effect of residual stress is

greater when the properties of the materials are

high.

If we try to represent the development of ra

according to the residual stress rR by an equation

of the following type:

ra rD rR

(Eq 2)

increases with the strength of the material; for

example, in the case of machining stresses in an

XC38 grade steel, Syren et al. found:

0 in the annealed state

0.27 when quenched and tempered

0.4 when quenched

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 17

Unfortunately, these results are in contradiction with an equation of the Goodman type. In

Eq 2, the coefficient is none other than what

is usually called the endurance ratio:

rD

Rm

of steels increases.

This apparent contradiction is probably explained by the fact that the residual stress relaxation phenomenon has not been taken into account. The value of the residual stress rR to be

introduced into equation of type 1 or 2 mentioned previously, must correspond to the stabilized fatigue stress, or the coefficient of influence

will include the relaxation process. The article

by D. Lohe and O. Voehinger in this Handbook

presents a large number of results concerning the

relaxation of residual stress under mechanical

and thermal loading. References 29 and 30 provide the information on the relaxation mechanism of residual fatigue stress as a function of

800

A

B Repeated

C bending tests

D

Repeated tension test

700

a, MPa

600

500

400

300

200

100

0

0

1000

m, MPa

2000

Use of Haigh diagrams to take longitudinal residual stress into account (XC42 steel

quenched after induction heating). A through D represent

results of repeated bending tests; X, repeated tension test.

Table 2 presents fatigue test results that correspond to A to

D.

Fig. 11

a, MPa

A

B

Repeated

C

bending tests

D

Repeated tension test

400

200

100

100

200

300

400

Pmax, MPa

Use of the Dang Van criterion to take residual

stress into account (XC42 grade steel quenched

after induction heating). A through D represent repeated

bending test; X, repeated tension test.

Fig. 12

is the amplitude of octahedral shearing or amplitude of the maximum shearing, and Palt is the

amplitude of hydrostatic pressure.

An example can be given as follows:

E

salt A B PD

mhp C Palt

300

200 100

Syrens results show that relaxation is much

greater when the mechanical properties are

lower.

When these experimental results are used with

the residual stress measured after carrying out a

fatigue test, and therefore stabilized, it is sometimes possible to use an equation such as Eq 2.

In the case of the fatigue bending test on cylindrical XC42 steel bars quenched after induction

heating (Table 2), the fatigue test results for the

different treatments correspond perfectly to the

Haigh diagram, provided any possible influence

of transverse residual stress on the fatigue stress

is ignored (Fig. 11).

It is not possible, however, to extend these

results to all materials and to the different manufacturing processes that introduce residual stress.

Also, preliminary tests are needed to validate the

methodology. The use of residual stress in calculations based on endurance diagrams of the

Haigh or Goodman type usually only allows for

an estimation of the increase in fatigue strength

as a function of the residual stress. Secondly, this

approach only allows for the combination of uniaxial stresses. Yet the residual stresses produced

by the various manufacturing methods used to

make the part are always multiaxial. The stresses

on the surface are biaxial while those inside the

part are triaxial. Depending on the area in which

the fatigue crack is initiated (on or below the

surface), the biaxial or triaxial stresses need to

be included when calculating the fatigue life.

This raises the problem of choosing a multiaxial

fatigue stress criterion. A simplified approach

based on an endurance diagram can therefore

only be an approximation.

The test carried out by the CETIM (Ref 16)

shows that the traditional Mises and Tresca criteria can only be used in the presence of higher

mean or residual stress. In this case, it is preferable to use criteria (Ref 31) that include the

amplitude of octahedral shearing (socta) or the

maximum shearing (sa) and maximum hydrostatic pressure (Pmax), as indicated here:

(Eq 3)

If Dsalt is taken on the maximum shearing

plane, D E l and B C, the result is the

Dang Van criterion (Ref 32).

If Dsalt is taken on the octahedral shearing

plane, when D E 1 and B C, the Crossland criterion results (Ref 33); when D E

1 and C 0, the Sines criterion (Ref 34); and

when D E 1 and B C, the Kakuno criterion (Ref 35).

This type of development can be continued to

invent new criteria, but it leads to complications because of the increasing number of parameters that need to be determined. Even with a

linear relationship of the Dang Van type, two

Wohler curves have to be determined to obtain

at least the two points needed to produce the diagram. If additional constants are added, the test

plane will be even greater, which means that the

criterion cannot be used in industry. As a result,

the criterion to be used must be simplified as

much as possible. This case deals with radial

loading problems (r1 K1 r2 K2 r3), and a

relationship of Eq 3 is sufficient. To simplify

matters further, the Crossland or Dang Van criterion can be used. In the case of combined and

out-of-phase loading, new criteria have been developed to take the out-of-phase effect into account (Ref 3639).

But as yet, these criteria have not been validated in a study in which combined and out-ofphase residual stresses have been taken into account. When the fatigue stress is complex, it is

also very difficult to calculate the expected residual stress relaxation.

When fatigue cracks are initiated on the surface, the stresses to be taken into account are

biaxial; this gives the following for the Crossland or Dang Van criterion:

socta

sa

Pmax

2

3

r 1a r 2a r r

2

1a 2a

r1a

2

1

(r1a r2a r1m r2m r1R r2R)

3

main reversed fatigue stresses (r1a r2a); r1m,

r2m represent the average value of the main fatigue stresses; and r1R, r2R are the residual stress

values measured in the two main directions (stabilized values).

To use the multiaxial fatigue criteria, the reference curve for the material being considered is

needed, just as it is when using the Goodman or

Haigh diagram. Reference 16 shows that the use

of Crossland or Dang Van criteria takes the increase in the bending fatigue strength into account perfectly as a function of the residual

stress introduced by the various treatment conditions (Fig. 12).

When the multiaxial aspect is brought into the

picture, the method that consists in introducing

residual stress into the calculation in the same

way as a mean stress, therefore, seems to give

satisfaction. The whole problem lies in defining

the residual stresses to be included in the calculation.

Taking residual stress into account is essential

for correct prediction of the fatigue limit. Figure

13 shows the important role played by compressive residual stress. If it is not taken into account,

the fatigue strength is underestimated (Fig. 13a).

If the residual stress measured or calculated is

used without taking relaxation of the residual

stress into account, the fatigue strength is over-

estimated (Fig. 13b). The correct method consists in calculating the fatigue strength after taking relaxation into account (Fig. 13c).

In order to correctly evaluate the effect of residual stress, various problems must be solved:

Measuring methods must be available to de-

a, MPa

a, MPa

(a)

a, MPa

Pmax

Fig. 13

large range of measuring methods currently

exists and it is possible to take measurements

(Ref 1) in most of the cases studied, particularly as a result of development of the x-ray

method and the incremental hole method.

When a complete measurement profile is

Relaxation

des R

Pmax

Initial

(b)

Pmax

(c)

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

a, MPa (Surface)

Xian

CETIM

300

200

100

100

200

300

400

500

Fig. 14

Fatigue results for a 45SCD6 steel treated under different conditions in a Dang Van diagram, taking the residual

stress on the surface into account. Xian, results obtained by Xian Jiaotong University; CETIM, results obtained

by CETIM

a, MPa

layer in which the fatigue crack is initiated,

the profile sometimes has high stress gradients. It is then difficult to know what stress

value to choosethe surface stress or the

stress peak that is often slightly below the surface (in the case of shot peening, for example).

To make a correct calculation, it would be

necessary to use calculation methods that take

the stress gradient into account and make the calculation not only for a single point, but for a

sufficient thickness of the material (thickness of

critical layer) for it to be representative of the

basic volume in which the fatigue damage process occurred (Ref 31). Figures 14 and 15 give

an example of processing of our results (details

can be found in Ref 29) in the case of the fatigue

of a shot-peened spring steel. Figure 14 shows

the fatigue results on a Dang Van diagram, taking both the residual stress and its relaxation into

account. A fairly good correlation can be observed. This indicates that a multiaxial fatigue

criterion taking the hydrostatic pressure into account can be used to predict the fatigue strength

in the presence of residual stress. Since, in this

case, the crack initiation zones are below the surface, calculations were made for different critical

layer depths. Figure 15 shows the results obtained for a critical layer depth of 100 lm. Better

alignment of the experimental points was observed. This example illustrates the possibility

of improving the calculation precision by using

the critical layer thickness approach. It is particularly relevant in the case of notched parts.

It has long been known that residual stresses

are not stable when they are subjected to fatigue

loading. To calculate the expected fatigue life,

precise information is therefore needed in order

to introduce stable residual stress values into the

calculation presented previously, that is, the

stress values that are really likely to be present

in the part during the best part of its lifetime.

The stress must therefore be measured on a

part already under cyclic loading or relaxation of

the residual stress estimated according to expe-

600

500

350

With

Without

400

A Consideration

600 MPa

200

a, MPa

300

Xian

CETIM

300

550 MPa

2 106

5 106

100

250

300

200

100

100

200

300

400

500

Fatigue results for a 45SCD6 steel treated under different conditions in a Dang Van diagram, taking the residual

stress into account using the critical layer approach. Xian, results obtained by Xian Jiaotong University;

CETIM, results obtained by CETIM

50

100

150

200

Pmax, MPa

Pmax, MPa

Fig. 15

107

240

Van criterion for a shot-peened 35NCD16

steel, without taking residual stress into account

Fig. 16

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 19

rience or modeling. In Ref 29 and 30, a complete

model was presented using FEM to determine

the stabilized residual stress after fatigue loading. This estimation of the residual stress can

then be used to calculate the fatigue life of a part,

taking residual stress into account. Despite the

initial definition of the Dang Van criterion,

which proposes that it only be used in cases of

fatigue strength with an unlimited number of cycles, an attempt was made to extend this criterion

to include a limited fatigue life with a very large

number of cycles (more than 2 106) (Ref 40).

Figure 16 shows an example of the fatigue life

estimated by calculation. First, a fatigue strength

diagram of the Dang Van type was defined according to the fatigue life obtained from a series

of fatigue life contours. Then the residual stress

relaxation was calculated using FEM (Ref 29).

Finally, the stabilized residual stress was introduced into the diagram. In this example, in the

case of a loading of 550 MPa, the point corresponding to loading is inside the limit of the

fatigue life at 107 cycles. No failure occurs. For

a loading of 600 MPa, the point corresponding

to loading including the residual stress is between the line corresponding to 5 106 cycles

and that of 2 106 cycles. Failure therefore

alt

2 Hardening effect

3 Roughness increase

1

3

Pmax

Fig. 17

strain hardening on the fatigue strength

Role of hardening

and residual stress

20

Role of

residual stress

10

0

Role of superficial hardening

10

20

30

40

50

HRC

Effect of the resistance of the base metal on the

increase in the fatigue strength after shotpeening treatment, distinguishing between the effect of

strain hardening and that of residual stress

Fig. 18

concentration

to plasticizing

layers

Figure 17 shows the effect of the surface finish

and strain hardening on the fatigue strength of

materials. It can be seen that an increase in the

roughness decreases the safety area, and strain

hardening increases the safety area, provided it

does not damage the material.

In the case of thermal or thermochemical surface treatments (induction quenching, case hardening, etc.), for example, it is necessary to take

the new fatigue strength of the treated layer into

account in the calculation.

The problem is more complex in the case of

residual stress introduced by plastic deformation

(pre-straining, machining, shot peening, roller

burnishing), since it is more difficult to distinguish between the influence of residual stresses

and residual microstresses present in the grains

of the deformed material and that of strain hardening of the material.

Evans (Ref 41) made this distinction in the

case of shot peening; to do so, he carried out

three types of fatigue tests on materials with

various mechanical properties:

Fatigue tests on a non-shot-peened material

Fatigue tests on a shot-peened material

Fatigue tests on a shot-peened material, but

with a mean test stress rm, which compensated for the surface residual stress. In this

type of test, the effect of the macroscopic residual stress is cancelled out and the fatigue

strength obtained only depends on the increase in the mechanical properties of the

plastically deformed material and the residual

microstresses distributed throughout the material. The results obtained are presented in

Fig. 18.

40

30

give an average fatigue life for the previously

mentioned loading in the order of 3.5 million

cycles. This example shows that, if the cyclic

properties of the material are correctly known, it

is possible to predict the fatigue strength of the

material in the presence of residual stress.

However, it should not be forgotten that other

factors must also be taken into account in calculating the fatigue lifethe introduction of residual stress is often accompanied by other

changes in parameters that have an influence on

the fatigue strength. In particular, these include:

resistance, the increase in the fatigue strength is

mainly due to surface strain hardening. On the

other hand, for highly resistant materials, it is

mainly the influence of the residual stress that

governs the fatigue strength. When materials

have low elastic limits, the stresses introduced

by shot peening relax much more easily than

they do when the elastic limit is high. This test

only shows a general tendency and does not ex-

actly show what the author is trying to demonstrate, since in the two types of tests for the same

material, the ratio of R(rmin /rmax) is modified.

As has been shown, the residual stress relaxation

changes when the R ratio is varied (Ref 29). The

real contribution of each factor will therefore be

different from that indicated in Fig. 18.

It can thus be seen that taking residual stress

into account in the calculation requires a serious

examination of the different parameters involved. When reliable results are needed, fatigue

tests will no doubt need to be carried out on the

part or the structure concerned. However, modeling enables the variation in the different parameters to be rapidly simulated in order to find

an optimal solution.

Partial Summary. The previous results show

that it is now possible to take residual stress into

account in calculations designed to predict the

fatigue life using a global approach. This must

take the relaxation of residual fatigue stress into

account, as well as the other effects (strain hardening, hardness) introduced by the manufacturing method used. A multiaxial fatigue criterion

that can integrate both the problem of residual

stress and the effect of the stress gradient applied

to a zone in the presence of stress concentration

has been developed, that is, the Crossland or

Dang Van criterion. It is used for a stabilized

state of residual stress, averaged out for a basic

volume of damage (thickness of critical layer),

and applied within a network of contours that

represents the fatigue life. In the future, tests will

be carried out to validate this type of criterion in

the case of combined stresses on notched parts

in the presence of residual stress.

Residual Stress into the Design

Office

Incorporating the notion of residual stress into

the design office must be gradual and can be divided up into several phases.

Today, very few industrial sectors consider

the residual stress parameter directly. In technical specifications, requirements are included that

are often closely related to residual stress without actually naming it. An Almen intensity must

be guaranteed in the case of shot peening, for

example, a roller-burnishing load, a machining

procedure or a minimum treated thickness in the

case of thermal or thermochemical treatment,

and a maximum dimensioning tolerance in the

case of a machined or welded part.

In the first phase of incorporation, a semiquantitative notion can be used to evaluate the

increase in performance in terms of fatigue life

or fatigue strength. A few examples can be presented. Table 3 gives an example of the effectiveness of shot peening in increasing the fatigue

life of different types of mechanical parts, and

Fig. 19 shows the beneficial role played by roller

burnishing on the fatigue strength of spherical

graphite cast iron crankshafts. Figure 20 shows

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

peening

Type of part

Type of stress

Spindles

Shafts

Gear box

Crankshafts

Aircraft coupling rods

Driving rods

Cam springs

Helical springs

Torque rods

Universal joint shaft

Gear wheel

Tank chain

Weld

Valve

Rocker arm

Reverse bending

Torsional

Fatigue life tests in service

Fatigue life tests in service

Tensile compression

Tensile compression

Dynamic stress

Fatigue life in service

Dynamic stress

Reverse bending

Fatigue life tests

Fatigue life tests

Fatigue life tests

Fatigue life tests

Fatigue life tests

4001900

700

80

3000, but highly variable

105

45

100340

3500

140600

350

130

1100

200

700

320

250

200

Alternated

stress

150

0

100

0

50

Repeated

stress

0

70

S-

G

70

S-

G

70

S-

G

40

S-

G

40

70

S-

S-

llin

g

en

ng

in

ro

en

io

ce

rfa

su

ct

di

at

re

tri

pe

fe

ot

du

in

sh

ni

tt

ou

ni

ni

te

io

ith

simplified method to calculate residual stress relaxation has been proposed in the first section.

Secondly, a method to predict fatigue life by taking the stabilized residual stresses into account

is presented. This design tool is based on the

FEM. It has been applied to shot-peened

35NCD16 grade steel. The different fatigue parameters often used in material research are studied. On the other hand, an experimental investigation about this material had been done by

Bignonnet (Ref 42). The results of the study

show that this design tool on fatigue developed

by the Laboratory of Mechanical Systems Engineering (LASMIS) (Ref 43) is able to take into

account different loading parameters.

The residual stresses, however, can be relaxed

by the deliberate application of thermal or mechanical energy. They will be especially relaxed

when the structure is subject to cyclic loading.

The relaxation phenomenon depends on a complex interaction of a number of factors, such as

the applied stress amplitude, the number of cycles of loading, the state of initial residual

stresses, and the nature, origin, and mechanical

properties of the material. In this article, only

relaxation during cyclic loading and the influence of the stable residual stresses on the fatigue

life are studied.

In only a few cases are the residual stresses

systematically analyzed using measurement of

the residual stress state during and after fatigue

testing. This is usually a difficult, time-consuming task. Several fatigue tests under tensile and

torsion loading with different stress amplitudes

have been done.

A numerical method for the prediction of the

residual stress distribution during and after fatigue has been developed. Finite element software is used for incorporating cyclic plasticity

into the calculation. A simplified method to calculate stabilized residual stresses was proposed.

of cycles up to a stabilized state of cyclic

properties.

make the following general predictions (Ref 30):

Fig. 19

TT by CO2 laser, R = 0.1

TT by CO2 laser, R = 1

TT by laser shock, R = 0.1

Treatment or process

in terms of fatigue strength from various surface

treatments. The results presented here are not at

all exhaustive and are taken from a limited bibliography. However, this figure should not be

taken as a reference, since the geometry of the

test specimens differs for each type of treatment.

In certain cases, this parameter can have an important effect on the gain achieved. Each industrial sector must carry out this type of comparison for the treatments and materials used in order

to help engineers design their products more effectively.

The second phase consists of predicting the

fatigue life using the notions developed in the

previous paragraph.

The third phase is the development of integrated tools for taking the residual stress into

account. The following paragraph presents an

example of such a design tool on fatigue for

three-dimensional components.

ST by nitriding R = 0.1

TT by induction R = 0.1

TS carburizing R = 0.1

ST by surface rolling R = 0.1

200%

ST by surface rolling R = 0.1

ST by shot peening R = 0.1

0

Fig. 20

25

50

Increase, %

75

Beneficial effect of various surface treatments on the fatigue strength (maximum gain reported in the literature).

TT, thermal treatment; ST, surface treatment

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 21

portant in the prediction of residual stress relaxation. They make it possible to calculate the residual stress state for a corresponded number of

cycles.

Then, a method to predict fatigue life while

taking these stabilized residual stresses into account is presented. The fatigue criterion under

multiaxial stresses is used. After the relaxation,

the residual stresses can be combined directly

with the cyclic loading. They change the value

of the mean stresses and influence the fatigue

life.

As a result, a design tool on fatigue for threedimensional components, FATIGUE3D, has

been developed in the laboratory of LASMIS

Calculation of

residual stresses

Outside solicitation,

cyclic loading

account when calculating the fatigue life of a

structure. It is based on the FEM. With this tool,

an iso-colored image of distribution of fatigue

life or safety factor can be obtained.

These methods mentioned above are applied

to shot-peened 35NCD16 grade steel. The different fatigue parameters often used in material

research are studied. A comparison with the experimental results has been carried out (Ref 43).

Basic Steps for the Calculation. Figure 21

shows the elementary steps of the calculation of

fatigue life while taking into account the residual

stresses.

First, the initial state of the residual stresses

of residual stresses

Measurement of

residual stresses

Prediction of

residual stress

Cyclic behaviors

of material

Fatigue, 3D

Mechanical properties

of material

safety factor, and

admissible residual stress

Fig. 21

Basic steps of calculation while taking into account the residual stresses

N1

N1

Fig. 22

s = OM/OM

M

N-cycles

M

Fig. 23

by the simulation of surface treatment. Then, this

initial field of residual stresses can be introduced

by FEM using ABAQUS (Ref 44, 45). Furthermore, a stabilized field of residual stresses can

be predicted after an elastic-plastic calculation,

in which the cyclic behaviors of material are

used. This stabilized field of residual stresses after relaxation must be taken into account when

calculating the fatigue life. At last, with the help

of a design tool on fatigue, FATIGUE3D, the

distribution of fatigue life and that of the safety

factor on fatigue can be obtained.

Principle and Basic Function. According to

the criteria of fatigue under multiaxial loading,

the fatigue life changes along with two parameters. One of them is maximum static pressure;

another is equivalent octahedral stress. Therefore, two stress-number of cycles (S-N) curves

are needed under simple cases (for example,

bending or torsion). With these two curves, a line

can be determined for a certain number of cycles

(life) in the stress plane P s (Fig. 22).

A different fatigue life is represented by a different line. If there is a complex stress state, it

gives a point in the plane P s. Respectively,

the line to which this point is nearest represents

the fatigue life of that stress state. In this way,

the fatigue life for any complex stress state and

for any part of a component can be predicted. A

distribution of the fatigue life is helpful for the

designer.

On the other hand, the safety factor on fatigue

is another interesting parameter for the designer.

FATIGUE3D provides a distribution of the

safety factor. In Fig. 23, N-cycles are the designed life and M is the point given by the stress

state. The safety factor on fatigue is defined as S

OM/OM.

P

Principle of calculating the safety factor on fatigue

Fig. 24

Three-dimensional mesh

process is a widely used technique because it can

produce a field of residual stress on the surface

of a part. These stresses are compressive and will

improve the properties on fatigue. As an example, a shot-peened part made of 35NCD16 grade

steel (0.35% carbon, 3.7% chromium, 0.3% molybdenum and having tensile properties of: a

yield strength of 1000 MPa, an ultimate tensile

strength of 1100 MPa, and an elongation of

17.5%) is studied. Many results of fatigue experiments are available regarding tension compression and torsion loading (Ref 44).

The geometry of the part is defined from the

real geometry so that the testing results can be

compared. It is similar to the fatigue test sample,

only the numerical values (radius, length) vary

for tension compression or torsion loading. The

difficulty is to define a thin mesh near the surface

(0.20.4 mm, or 0.0080.016 in.) where the residual stresses are introduced while maintaining

a normal mesh in other zones. Figure 24 shows

the three-dimensional mesh of the part.

Introduction of the Initial Field of Residual

100

100

Axial stresses

under tensile loading

100

Measurement

Calculated

200

Tangential stresses

under tensile loading

0

Residual stresses, MPa

300

400

500

Measurement

Calculated

100

200

300

400

500

600

600

0.05

0.10

0.15

0.20

0.25

0.30

0.05

0.10

0.20

0.25

0.30

0.4

0.5

0.6

100

100

Axial stresses

under torsion loading

100

0

Residual stresses, MPa

0

Residual stresses, MPa

0.15

Depth, mm

Depth, mm

Measurement

Calculated

200

300

400

500

600

Tangential stresses

under torsion loading

Measurement

Calculated

100

200

300

400

500

600

700

0

0.1

0.2

0.3

0.4

0.5

0.6

0.1

Depth, mm

Fig. 25

0.3

Depth, mm

calculated from the residual stresses measured

by the x-ray diffraction method in the upper

layer where initial stresses are introduced by surface treatment. In depth, the stresses are calculated for the structural equilibrium. Figure 25

shows the comparison of initial residual stresses

between the calculation and the experiment.

Prediction of the Residual Stress Relaxation. In order to predict the relaxation of the residual stresses, a simplified method proposed by

Lu et al. has been used (Ref 30). It supposes that

the relaxation of the residual stresses depends on

plastic deformation is the main cause of the relaxation. So, the cyclic behaviors of material are

very important for the prediction of the residual

stresses. However, they can be measured by experiment.

Figure 26 shows the cyclic behaviors of

35NCD16 grade steel, a cyclic softening material.

When the stabilized residual stresses are calculated, the behavior corresponding to NR /2 is

used. After an elastic-plastic calculation with

FEM, the relaxation of the residual stresses un-

1 cycle

2

3

15

50

150

/2, MPa

1000

900

NR/2

800

700

600

35 NCD 16

(Rm (UTS) = 110 MPa)

0.5

0.75

N,

cycles

1

2

3

15

50

150

NR/2

Re(Y.S.),

MPa

940

920

910

870

830

800

660

t/2, %

Fig. 26

0.2

The cyclic behavior of 35NCD16 grade steel. Re, yield stress; R ec, cyclic yield stress as a function of number

of cycles; Det, axial plastic strain range during cyclic loading

the results and their comparison with the test values have been shown in Fig. 27. In the case of

the traction-compression loading, two loading

levels were calculated. The same level for torsion loading with two different cycle numbers

was also analyzed. Figure 28 shows the stabilized residual stresses for different traction-compression load levels that are used in the prediction of the fatigue life. Unfortunately, it is

impossible to compare with the experiment results.

Prediction of the Fatigue Life. For predicting the fatigue life under a complex stress state,

it is necessary to have two S-N curves. One is

for simple traction or traction compression; another is for alternated torsion. These curves are

basic data of a material, and they can be obtained

by experimental method.

Using the program FATIGUE3D, a design

tool on fatigue developed in the LASMIS laboratory, a distribution of fatigue life and a distribution of safety factor for a structure can be obtained. In this example, the fatigue life under

traction-compression cyclic loading and torsion

loading has been calculated. The stabilized residual stresses play a role as static load. Figure

29 shows the results of fatigue life and their comparison with the experimental results. It is very

clear that the proposed method is available.

Prediction of Admissible Residual Stress.

In this approach, a calculation method has been

developed that can predict the admissible residual stress for a given fatigue life. This tool can

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

100

100

103 cycles, traction compression 550 MPa

100

Axial measurement

Axial calculated

Tangential measurement

Tangential calculated

200

300

400

Axial measurement

Axial calculated

Tangential measurement

Tangential calculated

100

200

300

400

500

500

0

0.10

0.05

0.15

0.20

0.25

0.30

0.05

0.10

0.15

Depth, mm

100

200

300

400

Axial measurement

Axial calculated

Tangential measurement

Tangential calculated

0

100

200

300

400

500

600

500

0

0.1

0.2

0.3

0.4

0.5

0.1

Fig. 27

The fatigue life can then be used to deduce a

stabilized residual stress. An example of defining

the quenching conditions according to the residual stress field obtained before relaxation is analyzed.

An inverse technique designed to obtain the

quenching conditions is then developed (Ref 46,

47). In this technique, it is necessary to simulate

quenching using the FEM. To simulate quench-

100

Tangential residual stress, MPa

100

Traction-compression

550 MPa

600 MPa

620 MPa

650 MPa

700 MPa

200

300

Traction-compression

550 MPa

600 MPa

620 MPa

650 MPa

700 MPa

0

100

200

300

400

500

0.05

0.10

0.15

Depth, mm

Fig. 28

0.3

0.4

0.5

whether or not residual compressive stress needs

to be introduced. The computer code also indicates the level of the residual stress and the zone

in which it needs to be introduced. In this way,

the treatment condition during the mechanical

design phase can be defined. Figure 30 shows a

map of the prestress zone indicated for a notch

sample. It can be seen that in the high stress concentration area, residual compressive stress must

be introduced to obtain a predetermined fatigue

0.2

Depth, mm

Depth, mm

0.30

100

Axial measurement

Axial calculated

Tangential measurement

Tangential calculated

100

0.25

200

100

0.20

Depth, mm

0.20

0.25

0.30

0.05

0.10

0.15

Depth, mm

0.20

0.25

0.30

mechanical data of the material need to be

known. Generally speaking, these parameters

depend on the temperature, but the surface heattransfer coefficient depends not only on the temperature, but also the sample geometry, the quenchant, and the quenching temperature. In order

to solve the problem, it is considered that there

is no change in any of the quenching parameters

or the sample geometry. Only the quenching

temperature changes. Quenching at different

temperatures was carried out. This gives the variation of the surface heat-transfer coefficient as a

function of the temperature for each quenching

temperature. It should be mentioned that numerical methods alone are not sufficient to obtain

the heat-transfer coefficient. The temperature

must be measured during quenching. Since the

heat-transfer-coefficient curve is known, the parameters of these curves as a function of the

quenching temperature can be defined. Once this

coefficient is known, quenching can be modeled

at different temperatures. The quenching temperature in the case of water quenching varies

between 20 and 80 C (68 and 176 F). The lower

the quenching temperature, the higher the level

of the residual stress field introduced. Initially,

the residual stress field obtained was found to be

lower than at 20 C (68 F) and higher than at

750

Without residual stresses

Experimental

Numerical

700

650

600

550

500

Experimental

Numerical

420

400

380

360

340

320

450

105

Fig. 29

440

Torsion load, MPa

106

107

Fatigue life N cycles

300

105

106

107

Fatigue life N cycles

Summary

S11

Value

2.24E+02

4.24E+01

+1.39E+02

+3.20E+02

+5.01E+02

+6.82E+02

1

3

2

Fig. 30

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

using the finite element method

Mechanical component

development

Introduction of a surface

and prestress treatment

Optimization of geometry

according to the choice

of manufacturing process

No

Stress analysis by

a finite element method

Adjustment of a

finite element model

Is fatigue life

in accordance

with

specifications?

Yes

Stop

Experimental checking of

stress: rapid prototyping

No

Fig. 31

Are results

coherent?

Yes

verified, the quenching temperature can be determined by varying the quenching temperature

during simulation. After each simulation, the

calculated residual stress field is compared to

that obtained from the fatigue life. A limit condition can be defined to improve the precision of

the results. When comparing the residual stress

fields, the hydrostatic pressure was used at each

point.

Fatigue life

estimation

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

with respect to the different properties of materials. The gain obtained from the presence of residual stress can be enormous. This article attempts to show the effects of residual stress

through the example of fatigue strength. Here, it

has been shown that it is now possible to predict

the fatigue strength of materials, taking residual

stress into account. The results of this study

show that it is possible to predict the residual

stress relaxation and fatigue life, with consideration of the influence of residual stress by the

FEM. It has been found that the calculated results of fatigue life at surface agree very well

with experimental results.

Although the author is not in a position to provide the same type of calculation tools for other

properties, such as corrosion resistance and the

adhesion of coatings, it is now reasonable to expect that the notion of residual stress will be

gradually introduced into the design stage of mechanical parts. Numerical modeling of the behavior beforehand saves a considerable amount

of time because of the reduction in the number

of experimental tests required. These tests are

often very long and costly, but they have proved

to be indispensable. The problem of taking residual stress into account at the design stage will

become more and more critical with the development of new materials (multimaterials, etc.)

and new treatments (combined treatments, etc.)

With the development of different experimental and numerical techniques, it is now possible

to introduce residual stress into the design office

for the integrated design of mechanical components, thus offering a new concurrent engineering approach applied to the design of mechanical

components taking residual and applied stress

into consideration. Figure 31 shows the different

connections between residual stress-integrated

design and other sectors that use the concurrent

engineering approach. A mechanical component

designer can simulate dynamic characteristics,

material processing, and product life.

Consideration of residual stress is becoming

increasingly important for two reasons: the introduction of multimaterials that induce residual

stress, and the need for the designer to reduce

the weight of components in order to remain

competitive. Basic research has brought a better

understanding of the phenomena relating to residual stress.

The main aim is to develop an integrated qual-

Prestress Engineering of Structural Material: A Global Design Approach to the Residual Stress Problem / 25

ity control tool. For industrial applications, future developments are necessary:

Measurement techniques

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

measurement methods

Integration of the portable optical method for

strain measurement in destructive techniques

Processing and materials

Development for the industrialization of new

laser shock, surface nanocrystallization, and

so on.

Optimization of residual stress in advanced

materials: functional gradient coating system,

for example, plasma spray coating, PVD,

CVD, or diamond-like carbon (DLC) coating;

metal-matrix composite; electronic packaging; intermetallic materials

Modeling

Development of advanced tools for the pre-

diction of initial residual stress in various processes: forging, casting, machining, welding,

plasma spray coating, composite processing,

electronic packaging manufacturing, mechanical surface treatment

Integration of the residual stress modeling

method into a global life design (3-D) and the

creation of optimization codes for the design

of mechanical components using the concurrent engineering approach

With the recent progress in the field of research and development, it is now possible to

introduce a new area of research: prestress engineering of structural materials (PESM or

PRESMA). In the near future, the mechanical

engineer will be able to use these tools in a prestressed concrete approach to civil engineering.

The ENSPED program in Europe is an excellent

example of the cooperation of industrial and research partners. Using this new approach, the fatigue behavior and mechanical behavior of new

materials can be considerably improved, increasing two- or threefold in some cases. In fact, many

failures occur in the area close to the surface.

Improving the mechanical behavior is often only

necessary on the surface (fatigue, fretting fatigue, wear, stress corrosion). As a result, this

type of approach can be expected to develop

very quickly in order to introduce adequate prestressing treatment in the area indicated by the

advanced design tool.

ACKNOWLEDGMENTS

The author is grateful to all the people at LASMIS working in the field of residual stress for

their help during the preparation of this paper:

Prof. J.L. Chaboche, Dr. D. Retraint, Dr. E. Rouhaud, Dr. X.L. Gong, Dr. L. Couturier, Dr. R.

Akrache, Dr. Z. Wu, Dr. S. Rasouli Yazdi, Mr.

A. Milley, Mr. D. Deslaf, Mr. B. Guelorget, Mr.

A. Voinier, Mr. F. Belahcene, Mr. K. Gong. Financial support from the European Union (under

12.

REFERENCES

1. J. Lu, Ed., Handbook of Measurement of Residual Stresses, Society for Experimental

Mechanics, USA, 1996

2. J. Lu, A. Niku-Lari, and J.F. Flavenot, Latest Development of Residual Stresses Measurement by the Hole Drilling Method, Materiaux et Techniques, Dec 1985, p 709718

3. Z. Wu, J. Lu, and B. Han, Study of Residual

Stress Distribution by a Combined Method

of Moire Interferometry and Incremental

Hole Drilling. Part I: Theory, J. Appl.

Mech., ASME, Sept 1998

4. Z. Wu, J. Lu, and B. Han, Study of Residual

Stress Distribution by a Combined Method

of Moire Interferometry and Incremental

Hole Drilling. Part II: Implementation of

Applied Mechanics, J. Appl. Mech., ASME,

Sept 1998

5. Z. Wu and J. Lu, Residual Stresses by Moire

Interferometry and Incremental Hole Drilling, Experimental Mechanics, Advances in

Design, Testing and Analysis, Vol 2, I.M.

Allison, Ed., Conf. ICEM 98, 2428 Aug

1998, (Oxford), p 13191324

6. Z. Wu, X.K. Niu, J. Lu, and P.G. Ifju, Study

of Process-Induced Stress in Orthotropic

Composite LaminateCarbon/Epoxy [O2 /

902]2s, Proceedings of the SEM Spring

Conference on Experimental and Applied

Mechanics, 13 June 1998, (Houston,

Texas), p 179181

7. Z. Wu, J. Lu, and Y. Guo, A New Method

of Residual Stress Measurement on Electronic Packaging, Experimental Mechanics,

Advances in Design, Testing and Analysis,

Vol 2, I.M. Allison, Ed., Conf. ICEM 98,

2428 Aug 1998, (Oxford), p 987991

8. J. Lu, J.F. Flavenot, and S. Thery, Study on

the Effect of the Finishing Treatment on the

Residual Stress Gradient in SiC Reinforced

Aluminum Metal Matrix Composite, J.

Compos. Technol. Res., ASTM, Vol 12 (No.

4), 1990, p 232238

9. J. Li, J. Lu, M. Perrin, M. Ceretti, and A.

Lodini, Study on the Residual Stress in

Cold-Rolled 7075 Al-SiC Whisker Reinforced Composites by X-ray and Neutron

Diffraction, J. Compos. Technol. Res., July

1995, p 194198

10. J. Lu, P. Peyre, C. Omam Nonga, A. Benamar, and J.F. Flavenot, Mechanical Surface

Treatments, Current Trends, and Future

Prospects, Surface Modification Technologies VIII, T. S. Sudarshan and M. Jeandin,

Ed., TMS, Sept 1994, p 589601

11. A. Benamar, J. Lu, J.F. Flavenot, P. Barbarin, G. Chalant, and G. Inglebert, Modeling Residual Stresses by Cold Rolling an

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

V. Hauk, H.P. Hougardy, E. Macherauch,

and H.D. Tietz, Ed., DGM, Verlag, 1992, p

891900

J.B. Roelens, F. Maltrud, and J. Lu, Determination of Residual Stresses in Submerged

Arc Multi-Pass Welds by Means of Numerical Simulation and Comparison with Experimental Measurements, Welding in the

World, Vol 33 (No. 3), 1994, p 152159

Proceedings of the National Conference on

the Residual Stress and Design Office, CETIM, (Senlis), 1991

D. Deslaef, E. Rouhaud, and J. Lu, Finite

Element Modeling of Shot-Peening Processes, Proceedings of the SEM Spring Conference on Experimental and Applied Mechanics, 13 June 1998, (Houston, Texas),

p 419422

A. Chabenat, and J.F. Flavenot, Measurement of Residual Stress on the Quenched

Axles after Heating by Induction, Traitement thermique/Revue de metallurgie,

(Heat treatment/Journal of Metallurgy), No.

124, 1978, p 233239 (in French)

N. Skalli, and J.F. Flavenot, Prediction of

Fatigue Life by Simulation with the Residual Stress Considerations, Conference proceedings, Spring Conference, SF2M

(French Society for Metal and Materials),

2223 May 1984, p 98117 (in French)

H.P. Lieurade, P. Castelucci, J.F. Flavenot,

and J. Lu, Efficiency of Improvement Techniques on the Fatigue Strength as a Function

of the Type of Welded Joint, Welding in the

World, Vol 31 (No. 4), 1991, p 268271 (in

French)

H.P. Lieurade, C. Maillard-Salin, and M.

Truchon, Fatigue Cracking of Welded Joints

of High Strength Steel, Proceedings of

IABSE, Colloquium, Lausanne, 1982, p

137144 (in French)

D. Francois, Role of the Residual Stress on

the Fracture, Conference Proceedings, Residual Stress in Welding Structure, CETIM,

SENLIS, Dec 1987, p 8797 (in French)

W.H. Friske, Shot Peening to Prevent the

Corrosion of Austenitic Stainless Steels,

Rockwell International Cooperation, 1975

A. Niku-Lan`, M. Menier, and G. Beranger,

Effect of the Surface State on the Stress

Corrosion Behaviour of a Stainless Steel,

CETIM Informations, No. 78, Dec 1982

C. Richard, J. Lu, J.F. Flavenot, and G. Beranger, Study of Residual Stress in the

Plasma Sprayed NiCrAlY Coating and

Characterization of Interface Toughness,

Memoires et Etudes Scientifiques, Revue de

metallurgie, May 1991, p 295306 (in

French)

C.S. Richard, G. Beranger, J. Lu, and J.F.

Flavenot, The Influence of Heat Treatments

on the Adhesion of Plasma Sprayed NiCrAlY Coatings, Surf. Coat. Technol., Vol

82, 1996, p 99109

24. C.S. Richard, G. Beranger, J. Lu, J.F. Flavenot, and T. Gregoire, Four-Point Bending

Tests of Thermally Produced WC-Co Coating, Surface and Coatings Technology, (No.

78), 1996, p 284294

25. C.S. Richard, J. Lu, G. Beranger, and F. Decomps, Study of Cr2O3 Coating Materials,

Part II: Adhesion to a Cast Iron Substrate,

J. Therm. Spray Technol., ASM International, Dec 1995, Vol 5 (No. 4), p 347352

26. R.J. Arsenault, Tensile and Compressive

Properties of Metal Matrix Composites,

Metal Matrix Composites: Mechanisms and

Properties, R.K. Evertt and R.J. Arsenault,

Ed., p 133167

27. P. Balladon, Stress Relieve After Welding,

Criterion and Process, Conference Proceedings, Residual Stress in Welding Structure,

CETIM, Senlis, Dec 1987, p 4569 (in

French)

28. R. Lematre, J.L. Lebrun, and J. Maeder,

Residual Stress and Fatigue, Materiaux et

Techniques, Sept 1982 (in French)

29. J. Lu, J.F. Flavenot, A. Turbat, and D. Francois, Modeling of the Relaxation of Residual Stress under Fatigue Loading, Conference proceedings, Residual Stress at Design

Office, CETIM, Senlis, Nov 1991, p 7198

(in French)

30. J. Lu, J.F. Flavenot, and A. Turbat, Prediction of the Residual Stress Relaxation During Fatigue, Mechanical Relaxation of Residual Stresses, ASTM STP993, 1998, p

7590

31. J. Lu, J.F. Flavenot, A. Diboine, S. Lasserre,

C. Froustey, M. Bennebach, and T. PalinLuc, Development of a General Multiaxial

Fatigue Criterion for High Cycles of Fatigue

Behavior Prediction, Multiaxial Fatigue

and Design, ESIS 21, A. Pineau, G. Cail-

32.

33.

34.

35.

36.

37.

38.

39.

40.

Engineering Publications, London, 1996, p

477487

K. Dang Van, On the Fatigue of Metals, Science Technique Armement, Memorial de

lartillerie francaise, 3ie`me Fascicule, 1973,

p 647722 (in French)

B. Crossland, Effect of Large Hydrostatic

Pressure on the Torsional Fatigue Strength

of an Alloy Steel, Int. Conf. on Fatigue of

Metals, IME/ASME, (London), 1956, p

138149

G. Sines, Fatigue Criteria under Combined

Stress or Strains, Transactions ASME, J.

Eng. Mater. Technol., Vol 13, April 1981,

p 8290

H. Kakuno and Y. Kawada, A New Criterion of Fatigue Strength of a Round Bar

Subjected to Combined Static and Repeated

Bending and Torsion, Fatigue of Engineering Materials and Structures, Vol 2, 1979,

p 229236

I.V. Papadopoulos, Cyclic Fatigue of

Metal, a New Approach, Ph.D. dissertation, Ecole Nationale des Ponts et Chausse`es, France, Dec 1987 (in French)

A. Galtier and J. Seguret, Multiaxial Fatigue

Criterion, Revue Francaise de Mecanique,

No. 4, 1990, p 291299 (in French)

A. Deperrois, On the Fatigue Limit Calculation, Ph.D. dissertation, Ecole Polytechnique, France, June 1991 (in French)

R. Akrache and J. Lu, 3D Calculation of

High Cycle Fatigue Life under Out-of-Phase

Multiaxial Loading, Fatigue and Fracture

of Engineering Materials and Structures,

Vol 22, July 1999, p 527534

J. Lu and J.F. Flavenot, Prediction of the

Residual Stress Relaxation During Fatigue

Loading and Taking the Residual Stress in

41.

42.

43.

44.

45.

46.

47.

Proceedings of 2nd International Conference on Residual Stress, (Nancy, France),

Elsevier Applied Science, Beck et al., Ed.,

Nov 1988, p 784790

W.P. Evans, R.E. Ricklefs, and J.F. Millian,

Local Atomic Arrangements Studied by Xray Diffraction, J.B. Cohen and J.E. Hillard,

Ed., Proc. of a Symposium, (Chicago), Feb

1965, Chap. 11

A. Bignonnet, Fatigue Behavior of a Shot

Peened 35NCD16 Grade Steel, Evolution of

Residual Stress Due to the Shot Peening as

a Function of the Type of the Loading, Proceedings of Ninth Conference on Shot Peening, CETIM, Senlis, France, Nov 1985, p

107116 (in French)

FATIGUE3D, User Manual, UTT-LASMIS, Jan 2000

ABAQUS Theory Manual, Hibitt, Karlsson,

and Sorensen, Inc., 1996

A. Milley, E. Rouhaud, and J. Lu, Prediction

of Residual Stress Relaxation in Three-Dimensional Structures under Tensile and

Torsional Loadings, Fifth International

Conference on Residual Stress, Vol 1, T. Ericsson, Ed., (Linkoping), 1618 June 1997,

p 424429

S. Rasouli Yazdi and J. Lu, Simulation of

Quenching and Fatigue Relaxation of Residual Stresses in Aluminium Parts, Fifth International Conference on Residual Stress,

Vol 1, T. Ericsson, Ed., (Linkoping), 1618

June 1997, p 490495

S. Rasouli Yazdi and J. Lu, Inverse Method

for Determining the Quenching Conditions

Using a Residual Stress Field, Proceedings

of the Third Int. Conf. on Quenching Control and Distorsion, ASM, (Prague), March

1999

G. Totten, M. Howes, T. Inoue, editors, p27-53

DOI: 10.1361/hrsd2002p027

All rights reserved.

www.asminternational.org

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

University of Karlsruhe, Germany

COMPONENTS MAY FAIL DURING SERVICE in very different ways. Important aspects

are:

Onset of plastic deformation or exceeding of

crack length

In practice, failure of a component is almost

always determined by the interaction of several

external parameters, such as unexpected high

static and/or cyclic loads, friction, thermal energy, oxidation, corrosion, and so forth. On the

other hand, the component itself may contribute

to premature failure, if there is confusion of the

component material, wrong heat treatment, surface decarburization, and so forth.

The residual stress state in a component may

be one of the most important parameters influencing its service behavior, particularly regarding high-strength material states. Residual

stresses may reduce yield or elastic collapse

loads and may promote corrosion cracking (Ref

1, 2). With respect to fatigue, which is the topic

of this article, residual stresses may alter the cyclic deformation behavior, promote or retard

crack initiation, accelerate or decelerate crack

propagation, and may be beneficial or detrimental to finite fatigue life and the endurance limit.

It is important to realize that in a component locally very high amounts of residual stresses may

exist. For example, in a spring steel up to 2000

MPa compressive residual stresses may be produced by a combined warm and stress peening

process (Ref 3).

In practice, no component is free of micro residual stresses. Almost all components have

macro residual stresses to a certain extent. Only

a small number of components exist where

macro residual stresses are negligible, for example after a suitable stress relaxation heat treatment. The reasons for residual stress formation

in components are treated in other articles in this

Handbook. The most important ones are the various manufacturing processes. The fundamental

mechanisms of residual stress formation are inhomogeneous plastic deformation caused by mechanical, thermomechanical, or thermal attack

and the development of constraint between different constituents of a component such as

phases or coatings during the formation of the

constituents (e.g., during the coating process)

and/or during cooling from a higher process

temperature (e.g., the sintering temperature).

Manufacturing processes not only result in the

formation of definite macro residual stress states,

but may also effect other changes of the component state close to its surface:

Formation of a specific topography

Work-hardening or (in very hard material

alteration of the micro residual stress state

Phase transformation

Crack initiation

All these processes may change the fatigue behavior of a given component. It is hardly possible to completely separate these influences, even

though this is very important for the understanding of the existing relationships.

Residual Stress

Stability of Residual Stresses. For the assessment of the influence of residual stresses on

fatigue behavior, the stability of the residual

stress state is of utmost importance. The reader

is referred to the article Stability of Residual

Stresses in this Handbook, which deals with the

relaxation of residual stresses due to monotonic

or cycling loading or due to thermal energy. It

is shown there that the micro residual stress state

is more stable against mechanically and thermally induced relaxation than the macro residual

stress state. On the other hand, it is also proved

that during fatigue the micro residual stress state

of a given steel may be changed by cyclic hardening and/or cyclic softening processes, which

are closely related to cyclic plastic deformations.

Hence, during fatigue loading of a component

with locally varying macro as well as micro residual stress states, complex interactions of the

macro residual stress state with the cyclic loading stresses and the micro residual stress state

processes occur. In both cases, the amount of

cyclic plastic deformation is the most important

parameter. Since during fatigue loadings that result in technically relevant lifetimes or in infinite

life, the cyclic plastic deformations decrease

with increasing hardness, it is generally expected

that the influence of macro residual stresses is

lower in low-strength steels than in high-strength

steels.

Comparison between Loading Stresses and

Residual Stresses. Since the alteration of the

micro residual stress state caused by manufacturing processes changes the local hardness

within a component and hence the local strength,

it is convenient to comprise the corresponding

influence on the fatigue behavior of the component as a change of the local fatigue resistance

of the steel used for the component.

On the other hand, stable macro residual

stresses are static stresses and may be regarded

as locally varying mean stresses. However, it is

important to realize that there are definite differences between loading mean stresses and residual stresses:

Loading mean stresses exist as a consequence

and/or internal pressure, and eventually temperature gradients. Residual stresses exist as

a consequence of inhomogeneous plastic deformation and/or as a consequence of constraints between different constituents of a

component.

Distribution and sign of loading mean

stresses depend on the external loading, the

geometry of the component, the elastic-plastic deformation behavior of the material (if

maximum loading stresses exceed the yield

strength or the cyclic yield strength), and on

thermophysical material properties (if temperature gradients exist). Distribution and

sign of residual stresses depend on the

amount and extension of inhomogeneous

plastic deformation in relation to the whole

component volume, on the interaction of mechanically and/or thermally driven deformation during manufacturing processes, on the

strength of the material state, and/or on the

volume fraction, the elastic as well as the

of the component that impose constraint on

each other.

Loading mean stresses are in equilibrium

with external forces and moments. Residual

stresses are in equilibrium with themselves

regarding balance of forces and moments

with relation to any sectional area and axis,

respectively. The latter is also true for thermally induced loading stresses.

Loading mean stresses may be changed and

finally disappear, if external loadings and/or

Nf

Ni

Ns

Number of cycles, N

Fig. 1

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

Ref 5

0.4

0.3

0.2

0.1

0

1

10

102

103

104

105

of normalized AISI 4140 steel. Source: Ref 68

characterized by different processes. Generally,

it contains four successive stages A, B, C, and

D:

0.15

Inhomogeneous

softening

Homogeneous

hardening

0.10

0.05

01

10

102

Number of cycles, N

inhomogeneous strain distribution during cyclic

softening of normalized AISI 4140 steel. Source: Ref 68

Fig. 3

the cyclic loading stresses determines the fatigue

behavior of components. The following sections

of this article discuss the consequences of this

interaction on cyclic deformation behavior, on

crack initiation and crack propagation, as well as

on the fatigue life and on the endurance limit.

Throughout this article, no special account is

given for multiaxial loading or residual stress

states and equivalent stresses are not used. Instead, this article considers simple loading states

and those components of the residual stress

statewhich is of course always multiaxial

that are relevant for the interaction of the cyclic

loading and the residual stress state under consideration. With regard to notched specimens

in which a multiaxial stress state always exists

upon mechanical loadingthe treatment is

based almost entirely on nominal stresses and

stress amplitudes. Hence, the fatigue strengths

and the cyclic loadings corresponding to finite

fatigue lives given in this article are nominal

stress amplitudes. First of all, the section Some

Aspects of Fatigue of Steels introduces some

basic definitions and relations of fatigue. This is

done here rather briefly. The phenomenon and

the processes of the fatigue of steels are presented in general and in detail in Volume 19 of

the ASM Handbook (Ref 4).

106

Number of cycles, N

Fig. 2

and finally removed. Residual stresses may be

changed and eventually disappear by thermally and/or mechanically induced relaxation.

Loading mean stresses are not influenced by

cyclic plastic deformation in stress-controlled

loading, unless stress gradients become very

highfor example in the root of sharp

notches, where mean stress redistribution

may occur. Residual stresses always relax, if

the cyclic loading exceeds certain threshold

values (see the article Stability of Residual

Stresses in this Handbook).

ening or/and work-softening effects in the total volume (homogeneous loading such as

push-pull of smooth specimens) or in the

highest loaded regions (inhomogeneous loading such as bending, torsion, or any loading

of notched components) and development of

persistent slip bands at the surface

Stage B: Microcrack initiation and propagation. Normal evolution of a macrocrack in the

surface region

Stage C: Stable macrocrack propagation connected with changes in the state of the material in the crack tip plastic zone

ure

An example of the consequences of these fatigue

stages is shown in Fig. 1 in an S-N diagram

(Wohler diagram) for the plain carbon steel SAE

1020 (German grade C 20), which presents experimental results for persistent slip band formation (ra /log Ns curve), microcrack initiation

(ra /log Ni curve), and failure (ra /log Nf curve)

(Ref 5).

Cyclic Deformation Behavior. Investigations on the cyclic deformation behavior in stage

A using computerized servohydraulic test systems enable the determination of the complete

stress/total-strain response at distinct numbers

of cycles that yields to hysteresis loops. In a

stress-controlled test, the plastic strain amplitude

ea,p plotted versus log N results in a cyclic deformation curve that corresponds to the microstructural changes in the material during cyclic

loading. Decreasing (increasing) ea,p values with

an increasing number of cycles N are typical for

a cyclic work-hardening (work-softening) behavior.

Characteristic cyclic deformation curves of

the low-alloy steel AISI 4140 (German grade 42

CrMo 4) in a normalized state are shown in Fig.

2. During stress-controlled push-pull tests with

R rmin/rmax 1 and ra RP0.2, the plastic

strain amplitudes systematically change with the

stress amplitudes applied. In the initial stage of

cyclic loading, a quasi-linear elastic behavior is

found. This stage is followed by a period of increasing ea,p values due to cyclically induced

work-softening effects that are connected with

inhomogeneous deformation processes caused

by fatigue Luders bands. This can be proved with

photoelastic investigations. A typical result obtained for this steel is presented in Fig. 3. Plastically deformed areas appertaining to the individual stages of inhomogeneous cyclic

deformation were registered photographically

and marked as hatched areas on the specimens

that are allied to the cyclic deformation curve.

Similar macroscopic softening effects occur in

all normalized states of plain carbon steels and

low-alloy steels with carbon contents below 0.6

wt% (Ref 68).

The quasi-linear elastic period shortens and

the extent of cyclic work softening grows simultaneously with increasing stress amplitude

ra. After reaching a maximum of softening, the

materials behavior is mainly characterized by

work hardening, and the ea,p values decrease

with increasing number of cycles. At the end of

the specimens life, the plastic strain amplitude

increases again, however fictitiously as a consequence of changes in the compliance of the

specimens due to crack propagation. The plot of

ra versus ea,t at N Nf /2, the so-called cyclic

stress-strain curve, of this steel is presented in

Fig. 4 together with the monotonic-stress/totalstrain curve at nearly the same strain rate e . In the

elastic-plastic range, the cyclic curve is lower

than the monotonic curve due to the work softening. The cyclic stress-strain curves of plain

Monotonic:

0.4 101s1

400

300

Cyclic:

0.3 101s1

N = Nf/2

200

100

0

0

0.1

0.3

0.2

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

Monotonic stress-strain curve and cyclic stressstrain curve at Nf /2 of normalized AISI 4140 steel

at nearly the same strain rates. Source: Ref 68

Fig. 4

0.8 wt% C

0.45

400

0.22

300

0.01

200

100

0

0.1

0.2

0.3

810

750

0.2

0.1

700

600

500

Fig. 6

68

a = 850 MPa

0.3

10

102

103

104

Number of cycles, N

105

(one) 101s1

800

Cyclic:

0.6 101s1

N = Nf/2

600

400

200

0

10

Monotonic stress-strain curve and cyclic stressstrain curve at Nf /2 of quenched-and-tempered

AISI 4140 steel at nearly the same strain rates. Source: Ref

68

Fig. 7

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

0.4

Monotonic:

1000

0.4

0.5

500

of the plastic strain amplitude that lasts up to

failure. The cyclic softening sets in after an incubation number of cycles Ni, which depend on

the applied loading amplitude. Figure 6 shows

the cyclic deformation curves ea,p versus log N

at different stress amplitudes for AISI 4140 tempered at 570 C. As a consequence of this cyclic

deformation behavior, the cyclic stress-strain

curves of quenched-and-tempered steels proceed

below the monotonic curves (see Fig. 7). In this

material state, the cyclic softening is caused by

an inhomogeneous deformation behavior. The

plastic deformation develops localized in distinct

zones and remains localized in these zones up to

failure. Figure 8 shows results from photoelastic

investigation carried out on quenched-and-tempered AISI 4140. In the plastically deformed

zones, the cyclic softening is caused by a rearrangement of the dislocations to energetically

more favorable dislocation structures such as cell

structures. As a consequence, the stored strain

energy and the hardness decrease. Within these

plastically deformed zones, microcrack initiation

and propagation occur (Ref 68).

Crack Initiation. The plastic strains generated by a cyclical loading lead to the formation

of microcracks in the material after a certain

0.3

0.2

a = 750 MPa

Inhomogeneous

softening

0.1

0

10

106

of quenched-and-tempered AISI 4140 steel. Ref

develop at the surface. Only in special cases

for example, very heterogeneous materials such

as case-hardened steelor at special loading

conditionssuch as Hertzian pressurefatigue

cracks may initiate below the surface. The first

microscopic cracks can appear in a very early

stage of the cyclic loading. Depending on the

loading conditions and the material, microcracks

can already be detected after 0.1 to 10% of the

lifetime and number of cycles to failure, respectively.

In ductile materials, changes of the surface

structure precede fatigue crack initiation. Fatigue

slip bands (persistent slip bands, PSBs), which

often are connected to extrusions and intrusions,

arise from irreversible gliding. These persistent

slip bands are often places in which first fatigue

cracks initiate. The crack initiation at persistent

slip bands often is typically for lower to middle

loading amplitudes. At higher amplitudes, cracks

may also initiate at grain boundaries. Crack initiation and the first phase of crack propagation

in ductile materials are controlled by shear

stresses. Within the first grain layers, the cracks

grow at an angle of 45 to the load axis. This

phase is called stage I crack growth. After reaching a certain length, the cracks become the socalled stage II cracks. Further propagation of theses cracks is controlled by normal stresses. Stage

II cracks grow approximately rectangular to the

load axis. In brittle materials, crack initiation is

often observed at inclusions and other defects.

Crack Propagation. To describe the stress

and strain field in front of a crack tip the loading

details and the geometry of a cracked specimen

or component must be considered. If the plastic

zone in front of a crack tip is small in comparison

with the dimensions of the specimens or components, the linear elastic fracture mechanics are

applicable, and the local loading conditions in

front of the crack tip can be described by the

stress intensity factor, K. Therefore, Paris,

Gomez, and Anderson (Ref 11) considered the

stress intensity factor as a driving force for the

propagation of a crack and suggested a crack

growth equation in the form:

da

C (dK)m

dN

C (German grade Ck 01, Ck 22, Ck 45, and C

80) in normalized conditions, determined from

cyclic deformation curves at N Nf /2, are

drawn up in Fig. 5. With increasing carbon content and pearlite content, respectively, an increasing stress amplitude at ea,p constant occurs due to a phase-hardening mechanism (Ref

810).

The cyclic deformation behavior of quenchedand-tempered steels in stress-controlled experi-

102

103

104

Number of cycles, N

Cyclic deformation curve and development of

inhomogeneous strain distribution during cyclic

softening of quenched-and-tempered AISI 4140 steel.

Source: Ref 68

Fig. 8

(Eq 1)

range of the mode I stress intensity factor, and

C and m are constants (Ref 12). With this equation, which is the well-known Paris law, it is

possible to describe the crack propagation in the

midrange of the crack growth rate. Figure 9

shows the qualitative features of the crack

growth behavior in a double logarithmic plot of

the growth rate versus the range of the stress

intensity factor for experiments with constant

stress range Dr rmax rmin. Distinct curves

are observed for different mean stresses rm or

values of R rmin /rmax, respectively. The

lower parts of the curves exhibit a threshold

value of DK below which crack growth is essentially arrested. Typically, the crack growth

R1

R2

R3

Stress intensity factor range, log K

Fig. 9

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

upper parts of the curves represent the crack

growth behavior as the Kmax values approach the

critical stress intensity factor KIc, which marks

the onset of unstable crack propagation. Then the

crack growth rate increases boundlessly.

The Paris law does not represent the complete

crack growth curves shown in Fig. 9. The upper

part of the curves depends on KIc, and the lower

part on the threshold value DKth. Therefore, several complementations were suggested. An equation that incorporates all of the features of crack

growth curves has been used by NASA (Ref 13).

It has the form:

valid:

ea,p ef N a

f

and

ea,e rf /E N b

f

(Eq 6)

fatigue strength coefficient, the fatigue ductility exponent, b the fatigue strength exponent,

and E the Youngs modulus. Rearrangement of

both equations yields to:

ea,p N fa constant

da

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

dN

[(1 R) KIc DK]q

(Eq 5)

(Eq 7)

and

(Eq 2)

ea,e E N b

ra N b

constant

f

f

log a

log f

log Rm

log Rf

log Ns

Fig. 10

log Ne log Nf

f/ E

E N

N

Number of cycles to failure, Nf

Fig. 11

effect, the R ratio effect, and the final instability

effect. By suitable choice of the exponents, this

equation can accept a number of different variants of crack propagation laws. For m p q

0, for example, the Paris law is obtained.

The concepts of linear elastic fracture mechanics and the use of the stress intensity factor

are limited to cases in which small-scale yielding

develops at the crack tip. Moreover, the applicability of Eq 1 and 2 is restricted to sufficiently

long macroscopic cracks. Short cracks can grow

at smaller ranges of the stress intensity factor as

DKth. At comparable DK values, they can exhibit

clearly higher crack growth rates than long

cracks (Ref 14).

Lifetime Behavior. The lifetime behavior of

steels is commonly described in an S-N diagram

(Wohler diagram) (Ref 15), where the loading

amplitude is plotted versus the number of cycles

to failure in double logarithmic scaling. For

stress-controlled fatigue tests, the S-N curve is

frequently described by three straight lines (see

Fig. 10). In the range of the quasi-static strength

Rm (Nf Ns) and the fatigue strength Rf(N

N0f ), the endurable stress amplitude is constant

and given by the ultimate strength Rm and the

fatigue strength Rf, respectively. In the range of

finite life (Ns Nf Ne), the Basquin relation

(Ref 16):

Fatigue strength, Rf

ra rf N b

f

often describes the lifetime behavior satisfactorily. rf is the fatigue strength coefficient and b

the fatigue strength exponent.

In the case of total-strain-controlled fatigue

tests the ea,t values in the total strain Wohler

curve (Fig. 11) can be separated into the elastic

and plastic parts:

Re

Goodman

0

Rf

Gerber

Re

Mean stress m

Fig. 12

(Eq 3)

Rm

(Eq 4)

From experience it is known that both parts extend exponentially from the number of cycles to

failure Nf, which means that the plots log ea,e

versus log Nf and log ea,p versus log Nf are linear.

(Eq 8)

Coffin-Manson (Ref 17, 18) (Eq 7) and Basquin

(Eq 8).

Tensile mean stresses promote crack initiation

and accelerate crack propagation. Therefore, the

fatigue strength Rf decreases with increasing

mean stresses. This correlation is normally represented in a Smith (Ref 19) or in a Haigh (Ref

20) diagram. As an example of such fatigue

strength diagrams, Fig. 12 shows a Haigh diagram. The fatigue strength Rf is plotted versus

the mean stress rm. The Goodman approximation (Ref 21):

Rf R0f (1 rm /Rm)

(Eq 9)

which represents a straight line between the fatigue strength R0f at rm 0 and the tensile

strength Rm, is frequently used to estimate fatigue strength. For practical use, the allowable

stresses are restricted by the demand that macroscopically no plastic deformation should occur. This restriction leads to a straight line connecting the yield strength Re at both axes.

Through this, the allowable stress amplitudes lie

in the hatched area in Fig. 12. The slope of the

Goodman straight line is often also described as

mean stress sensitivity M R0f rm /Rm. The

Goodman approximation normally leads to conservative estimations. Often, the Gerber parabola

(Ref 22):

Rf R0f [1 (rm /Rm)2]

(Eq 10)

fatigue strength. Mostly, experimental results

range between the Goodman and the Gerber approximations.

In the range of infinite lifetime, the influence

of mean stresses can be described by damage

parameters. Two frequently used damage parameters are the one proposed by Smith, Watson,

and Topper (Ref 23):

PSWT (ra rm) ea,t E

(Eq 11)

(Eq 12)

mean stress, respectively, ea,t and ea,p are the resulting total strain amplitude and plastic strain

amplitude, respectively, and E is the Youngs

modulus. Normally, for ea,t and ea,p the values at

half of the lifetime are used.

the Cyclic Deformation Behavior

0.4

a in MPa

Unpeened

Shot peened

0.3

380

0.15

10

102

103

104

105

106

Number of cycles, N

0.4

a in MPa

450

Unpeened

Shot peened

0.3

550

600

0.05

400

375

350

350

400

275

0

1

10

(b)

102

103

104

Number of cycles, N

1.0

900

0.8

105

102

103

104

Number of cycles, N

10

a in MPa

Unpeened

800

Shot peened

750

700

0.2

500

400

550

600

(c)

10

102

103

104

Number of cycles, N

105

a in MPa

Unrolled

0.03

Deep rolled

600

650

550

600

0.02

500

0.01

550

500

0

102

103

104

Number of cycles, N

10

(b)

105

106

push-pull loading of (a) normalized and (b) nitrocarburized specimens of the SAE 1045 steel in unpeened

and deep-rolled conditions. Source: Ref 27

Fig. 14

0.10

Untreated

0.08

Shot peened

0.06

Deep rolled

0.04

0.02

0

106

push-pull loading of (a) normalized, (b)

quenched-and-tempered (730 C/2 h), and (c) quenchedand-tempered (570 C/2 h) smooth specimens of the steel

AISI 4140 in unpeened and shot peened conditions.

Source: Ref 25, 26

Fig. 13

106

0.04

1

1

105

106

0.6

0.4

500

450

1

450

550

0.2

0.1

a in MPa

Unrolled

Deep rolled

510

0.10

(a)

Plastic strain amplitude (a,p), %

(a)

Plastic strain amplitude (a,p), %

600

325

300

250

0.1

0.20

350

0.2

macro and micro residual stresses on the cyclic

deformation behavior can be studied very well

after mechanical surface treatments, for example

cyclic deformation curves for stress-controlled

push-pull loading of different heat treated

smooth specimens of the steel AISI 4140 (German grade 42 CrMo 4) are compared in Fig. 13

in unpeened and in shot peened conditions with

compressive residual stresses at the surface (Ref

25, 26). In the normalized state (Fig. 13a), the

onset of cyclic deformation is different in both

conditions, since the shot peened specimens with

surface compressive residual stresses rrs

10

100

103

Number of cycles, N

104

105

Cyclic deformation curves of different mechanically surface treated materials for stress-controlled push-pull loading in the low-cycle fatigue (LCF) regime of the plain carbon SAE 1045 steel. Stress amplitude

ra 350 MPa, Almen intensity I 0.120 mm A, rolling

pressure p 150 bar; surface residual stresses: shot peening rrs 500 MPa, deep rolling, approximately 600/

350 MPa. Source: Ref 28, 29

Fig. 15

cycle and higher plastic strain amplitudes during

the first cycles for stress amplitudes ra between

250 and 350 MPa. After a certain number of cycles, the opposite tendency can be detected and

the plastic strain amplitudes of the shot peened

conditions are smaller than those of the unpeened

material states. However, for the same ra values

the plastic strain amplitudes of both conditions

approach another at relatively high numbers of

cycles. Corresponding results for a quenchedand-tempered (730 C/2 h) AISI 4140 steel are

presented in Fig. 13(b). In the unpeened condition, the characteristic cyclic deformation behavior of quenched-and-tempered steels occurs with

a quasi-elastic incubation period, which is followed by cyclic softening until crack initiation.

After shot peening that generates surface compressive residual stresses rrs 410 MPa, the

onset of cyclic softening is shifted to smaller

numbers of cycles. Furthermore, it is interesting

to note that for identical stress amplitudes and

comparable numbers of cycles, the higher plastic

strain amplitudes are always measured for the

shot peened specimens. Figure 13(c) shows a

compilation of cyclic deformation curves for another quenched-and-tempered (570 C/2 h) AISI

4140 steel with a higher strength compared with

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

shot peened condition that has surface compressive residual stresses rrs 530 MPa is characterized for all investigated ra values by small

measurable plastic strain amplitudes during the

first cycle that diminish or disappear first of all

with an increasing number of cycles. After a subsequent regime of quasi-elastic behavior, cyclic

softening, which yields to lower plastic strain

amplitudes and larger numbers of cycles to failure in comparison with the unpeened conditions,

is dominant.

The effect of deep rolling on the behavior of

the cyclic deformation curves is presented in Fig.

14(a) for the normalized plain carbon steel SAE

1045 (German grade Ck 45) (Ref 27). At comparable stress amplitudes essentially smaller

plastic strain amplitudes are measured in the

deep-rolled condition. In this materials state, the

cyclic softening effects are extremely restricted,

but continue during the whole lifetime until

crack initiation. As a consequence of the smaller

plastic strain amplitudes at identical stress amplitudes, higher numbers of cycles to failure are

observed for the deep-rolled condition compared

with the normalized ones. Figure 14(b) shows

the influence of deep rolling in the case of a nitrocarburized SAE 1045 steel. Due to the different microstructures in the near-surface areas of

the specimens, the same stress amplitudes yield

plastic strain amplitudes that are five times

smaller than those of the normalized state. The

deep-rolling process modifies the cyclic deformation curves of the nitrocarburized specimens

in a manner similar to the normalized specimens.

However, the reduction of the plastic strain amplitudes is much less pronounced.

Figure 15 shows plastic strain amplitudes as a

function of the number of cycles plotted for

304 (rolling pressure 150 bar, surface residual

stresses r 350 MPa) using compact as well as hollow

specimens with different percentages of strain-hardened

layers compared with an untreated state and a shot peened

condition (rrs 450 MPa). Source: Ref 28, 29

amplitude, the benefit of mechanical surface

treatments becomes obvious. The influence of

mechanical surface treatments on cyclic plasticity can be summarized in cyclic stress-strain

curves, which correlate stress amplitudes and

plastic strain amplitudes for, for example, half

the number of cycles to failure (see the section

Cyclic Deformation Behavior). Figure 16, as

an example, shows data for an austenitic AISI

304 steel in shot peened as well as deep-rolled

conditions compared with untreated specimens

(Ref 28, 29). Effects of mechanical surface treatments are more distinct with a higher ratio between the area of affected surface layer and the

cross section of the specimen. In the case presented here, hollow thin-walled specimens were

prepared from compact specimens for analysis.

As shown in Fig. 16, cyclic yield strength considerably increases compared with compact

specimens, if only the fatigue behavior of the

surface layers of mechanically surface treated

components is investigated.

If residual stresses are present in the root of

notched specimen, marked changes always occur

in the initial parts of the notch root cyclic deformation curves (ea,p /log N curves) and in the

notch root cyclic mean strain curves (em /log N

curves) that depend on the sign and the magnitude of the residual stresses. Characteristic examples are presented in Fig. 17 for normalized

push-pull loading with the stress amplitudes indicated for a normalized SAE 1045 steel (Ref

28, 29). Results for annealed as well as for shot

peened or deep-rolled conditions are shown. One

can clearly see that both mechanical surface

treatments considerably diminish the plastic

strain amplitudes. Due to the thicker affected

surface layer in the case mentioned first, the effect for deep-rolled states is more pronounced

than for shot peened states. Because fatigue

700

600

500

Deep rolled (0.5 mm wall thickness)

Deep rolled (0.3 mm wall thickness)

400

300

200

Untreated

Deep rolled

Shot peened

100

0

0

0.1

0.2

0.3

0.4

Plastic strain amplitude (a,p), %

0.5

Fig. 16

rs

l

0.4

n,a in MPa

rs = 15 MPa

rs = 490 MPa

0.3

220

0.2

240

200

220

175

0.1

150

0

1

10

102

170 150

190

103

104

105

106

0.06

n,a = 200 MPa

0.04

rs = 710 MPa

460

0.02

320

0

1

10

102

103

104

105

106

Number of cycles, N

(a)

Number of cycles, N

0.06

0.30

n,a in MPa

220

190

170

150

0.15

0.00

(a)

0.04

0.02

460

rs = 710 MPa

0.2

1

0.15

1

(b)

10

102

103

104

105

106

(b)

10

102

103

104

105

106

Number of cycles, N

Number of cycles, N

mean strain versus number of cycles (b) for

stress-controlled push-pull loading of normalized notched

specimens (notch factor Kt 3.0) of the SAE 1045 steel in

unpeened and shot peened conditions. Source: Ref 30, 31

Fig. 17

320

mean strain versus number of cycles (b) for

stress-controlled push-pull loading of quenched-and-tempered (400 C/2 h) notched specimens (notch factor Kt

3.0) of the SAE 1045 steel in cut-milled and shot peened

conditions with different surface residual stresses. Source:

Ref 30, 31

Fig. 18

specimens and in Fig. 18 for quenched-and-tempered specimens of SAE 1045 with a notch factor Kt 3.0 that were investigated in push-pull

tests (Ref 30, 31). Figure 17(a) compares the cyclic deformation curves of material states with

very low residual stresses (produced by milling

and subsequent annealing) with rather high residual stresses (produced by shot peening). In

both conditions plastic deformation occurs in the

first cycle at nominal stress amplitudes rn,a

above 150 MPa. The annealed state yields continuous cyclic softening for all investigated rn,a

values. Contrarily, the initial plastic strain amplitudes of the shot peened condition that are

much less dependent on the stress amplitude

compared to the milled and annealed state are

reduced during the first cycles. This is a result

of residual stress relaxation that is the more pronounced the higher the nominal stress amplitude

is (see the following section Evaluation of Experimental Results). The consequences are

smaller effective stress amplitudes and smaller

ea,p values in the following cycles. Hence, the

observed cyclic work hardening is fictitious.

However, at higher numbers of cycles, cyclic

softening occurs with lower ea,p values than in

the unpeened condition.

On the other hand, as can be seen from Fig.

17(b), the shot peening notch root residual

stresses have an influence on the initial parts of

the mean strain curves that were determined for

different nominal stress amplitudes (Ref 30, 31).

As a consequence of the cyclic deformation induced relaxation of the compressive residual

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

log N curves show negative plastic mean strains,

where the magnitudes of em increase with increasing nominal stress amplitude. The minima

of em are shifted to lower numbers of cycles with

increasing rn,a values. However, from the minima of em at approximately 5% of the number of

cycles to failure, the em /log N curves show increasing mean strains for all nominal stress amplitudes. Simultaneously, a decrease of the tensile compliance in the hysteresis loops that

intensifies with an increasing number of cycles

is observed. By that, the compressive compliance is approximately constant. This finding

proves that this increase of the mean strains with

the number of cycles is caused by an increase of

the number of microcracks and/or the growth of

microcracks. These sections are indicated by

dashed lines in the em /log N curves of Fig. 17(b).

The true plastic strain amplitudes in the cyclic

deformation curves of the quenched-and-tempered (400 C/2 h) specimens in Fig. 18(a) are

influenced by the magnitude of the notch root

residual stresses, but not by their sign (Ref 30,

31). All investigated conditions in Fig. 18(a)

show for the initial cycles fictitious cyclic-workhardening effects due to the relaxation of residual stresses rrs of the surface layers (rrs 460

MPa and 320 MPa produced by upcut milling

as well as 710 MPa by shot peening). Up to

N 20 cycles, increasing magnitudes of the

compressive residual stresses cause higher negative mean strains. However, positive milling re-

0

n,a = 220 MPa

200

n,a = 150 MPa

400

0.20

250

a in MPa

0.16

125

0.12

150

0.08

350

0.04

200

100

0

1

102

10

(a)

3

10

102

103

104

105

106

Number of cycles, N

(a)

105

106

125

250

9

350

15

a in MPa

102

10

400

450

103

104

105

106

Number of cycles, N

strain versus number of cycles (b) for axial

stress-controlled cyclic loading with mean stress rm 300

MPa for specimens of the normalized SAE 1045 steel.

Source: Ref 32

Fig. 20

rs

300

100

300

600

900

0

(b)

10

102

103

104

105

106

Number of cycles, N

(notch factor Kt 3.0) for (a) normalized SAE 1045 in shot

peened conditions and (b) quenched-and-tempered (400

C/2 h) SAE 1045 in cut-milled and shot peened conditions.

Source: Ref 30, 31

Fig. 19

A Untreated

B Deep rolled

80

60

40

20

0

C Shot peened

Macrocracks

(>750 m)

Short cracks

(200-750 m)

Microcracks

(>200 m)

No damage

marks

the damage evolution of the push-pull loaded

austenitic steel AISI 304. ra 320 MPa; R 1; A, Nf

3859; B, Nf 4445; C, Nf 20,265. Source: Ref 28

Fig. 21

18b).

A study of the residual stress effects on the

cyclic deformation behavior seems to be possible by their simulation by applied mean stresses,

which are homogeneously distributed over the

cross section of the specimen. An example is

presented in Fig. 20 for smooth, normalized SAE

1045 steel specimens that were stress-controlled

loaded in push-pull tests with a constant mean

stress of rm 300 MPa (Ref 32). The cyclic

deformation curves at stress amplitudes ra

125 and 150 MPa in Fig. 20(a) show first of all

a quasi-elastic incubation interval, followed by

considerable cyclic softening within some cycles. Subsequently, within some further cycles,

cyclic work hardening occurs to such an extent

that extremely low plastic strain amplitudes result. For ra 200 MPa, the compressive stress

peak | rm ra |, which is higher than the yield

strength, induces plastic deformation in the first

cycle. This procedure is followed by a rapid cyclic work hardening in such a manner that at cycles above 100 extremely low ea,p values are observed. As a consequence of the constant

compressive mean stress, the specimens shorten

for all investigated stress amplitudes, as proved

by Fig. 20(b). For ra 200 MPa, the mean

strains increase considerably and continuously.

However, for ra 200 MPa the ea,p values saturate above 20 cycles.

200

3

(b)

1

104

150

100

21

600

103

Number of cycles, N

in the work-hardened surface layers.

In hardened as well as in quenched-and-tempered at low temperature conditions, the changes

in the cyclic deformation behavior result not

only from surface hardening or softening effects,

but also from the more stable residual stresses as

can be seen in Fig. 19(b) for notch root residual

stresses of the quenched and tempered (400 C/

2 h) SAE 1045 steel under the loading conditions of Fig. 18(a) (Ref 30, 31). The consequences are a considerable influence of the

residual stresses on the cyclic behavior, especially with regard to the development of mean

rs

increases in the plastic mean strains that appear

for the numbers of cycles above the minimum of

em are caused also by microcracks as described

previously.

Evaluation of Experimental Results. In relatively soft material states, as for example in normalized as well as in quenched-and-tempered at

high-temperature conditions, the consequences

of mechanical surface treatments by shot peening or deep rolling on the cyclic deformation behavior are mainly caused by near-surface micro

residual stresses, that is, work hardening of the

surface layers, because the macro residual

stresses are relaxed very soon by cyclic plastic

deformation (see the article Stability of Residual Stresses in this Handbook). This relaxation

of macro residual stresses is shown in Fig. 19(a)

for shot peening residual stresses in the notch

root of a normalized SAE 1045 steel (Ref 30,

31) under the loading conditions of Fig. 17(a).

The dislocation structures in the ferrite after the

mechanical surface treatment are not stable and

change during cyclic loading in energetically

more favorable arrangements. The formation of

typical fatigue-induced dislocation structures is

combined with cyclic softening effects as presented in Fig. 13(a), 14(a), and 17(a). The small

plastic strain amplitudes of the shot peened or

deep-rolled conditions and the resulting increase in fatigue life are caused by the re-

the Crack Initiation

Characteristic Examples. Crack initiation

occurs as a consequence of microstructural

changes in metallic materials during cyclic loading. Different mechanisms are responsible for

their formation (e.g., Ref 33). If it is accepted

that for given materials states at comparable load

amplitudes increasing amounts of plastic strain

amplitudes lead to decreasing numbers of cycles

to crack initiation Ni, it follows that residual

stresses may extend, shorten, or leave unchanged

the number of cycles to crack initiation. However, experimental investigations concerning the

influence of macro and micro residual stresses

on crack initiation are scarce. This is due to the

difficulties connected with the observation of the

formation and the propagation of small cracks.

A recently published investigation gives a report on the influence of mechanical surface treatments on crack initiation and crack propagation

in push-pull loading of steels (Ref 28). In untreated materials, crack initiation normally takes

place at positions of high localized slip, for example, at extrusions and intrusions that are connected with persistent slip bands. However, as

shown in Fig. 21 in shot peened and deep-rolled

conditions of the austenitic steel AISI 304, crack

formation occurs later than in the untreated state

due to the consequence of numerous obstacles

for slip (dislocations, grain and twin boundaries)

in the work-hardened surface layer that impede

localized slip. In these surface-work-hardened

conditions, no persistent slip band is observed at

all. Furthermore, crack propagation is slower

than in the untreated state due to the effect of

microstructure and compressive residual

stresses. Similar results are found in Ref 28 for

the normalized steel SAE 1045, in Ref 34 for the

plain carbon steel SAE 1080, and in Ref 35 and

36 for a high-strength spring steel AISI 6150

(German grade 50 CrV 4).

However, as reported in Ref 1, the crack initiation time of shot peened specimens is sometimes shorter than that of unpeened specimens

despite increased lifetimes. One example is

given in Fig. 22 for the quenched-and-tempered

steel SAE 1045 in the case of bending fatigue

tests in seawater (Ref 37). With the exception of

high stress amplitudes, the cracks are formed

earlier in shot peened specimens than in ground

ones. This finding is attributed to an enhanced

crack initiation at micronotches resulting from

shot peening, which is obviously supported by

corrosion pittings in the case of seawater environment. However, the number of cycles to failure of the shot peened conditions are higher than

that of ground conditions.

Residual stresses may have a remarkable influence on the location of crack initiation. In the

case of bending fatigue, for example, cracks may

start below the surface, if the magnitude of the

800

high enough. Characteristic results are presented

in Fig. 23 for shot peened bending specimens in

a hardened state of the steel SAE 1045 with the

bending fatigue strength Rf 960 MPa after

shot peening with the shot size d 0.3 mm and

Rf 1050 MPa with d 0.6 mm, respectively

(Ref 38, 39). For stress amplitudes ra Rf, crack

initiation was observed directly at the surface.

However for ra Rf, subsurface cracks occurred

as shown in Fig. 24. If the centers of the rosettes

on the scanning electron micrographs (SEMs)

are considered to be the crack initiation points,

it is obvious that the crack initiation depth increases with decreasing stress amplitude (Ref

39). At stress amplitudes that result in a fatigue

life Nf above 107 cycles, crack initiation is located approximately 0.3 mm below the surface.

These findings stimulated the development of a

concept of the local fatigue strength (see the paragraphs on experimental results in the section

Influence of Residual Stresses on S-N curves

in this article).

Evaluation of Experimental Results. The definition of the number of cycles to crack initiation

Ni and particularly the establishment of the affiliated crack length is of decisive meaning for

the evaluation of the influence of different parameters on Ni from experimental results from

(a)

Fracture

Crack initiation

600

200 m

(b)

400

200 m

200

Shot peened

(c)

Ground

0

104

105

106

Number of cycles, N

107

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

200 m

(d)

200 m

200 m

0.4

200 m

0.2

Site of

macrocrack

initiation

if in the surface area:

The roughness is increased, for example, by

shot peening

second phases.

In this context, an investigation of the fatigue

crack initiation in double-edge notch specimens

of SAE 1080 steel with a fine-grained, spheroidized microstructure shows considerable effects

of the macro residual stresses (Ref 34). In a

press-fitted condition, large compressive residual macro stresses occur that remained relatively

stable throughout the fatigue life and thus greatly

increased the numbers of cycles to crack initiation Ni. With reference to the Basquin-relationship, Ni is estimated from the following relationship:

r (r

2ra

a

1/b;

r )

rs

(Eq 13)

crack length of a 0.1 mm, rrs is the macro

residual stress, ra is the applied stress amplitude,

and ri 2315 MPa and bi 0.197 are material constants.

200 m

Shot size 0.6 mm

1000

1200

Stress amplitude (a), MPa

the Crack Propagation

(g)

1400

a hardened SAE 1045 steel in different shot

peened conditions. Source: Ref 38, 39

Fig. 23

stresses

(f)

0.3

800

face treatment

(e)

Ni

Fig. 22

0.1

amounts to only some grain diameters, then Ni

is determined by shear-stress-controlled processes in stage I of crack propagation. However,

if crack lengths of several hundred microns appear (stage II of crack propagation), Ni will be

determined by normal-stress-controlled processes. In the last case, stable macro residual

stresses may have a considerable influence on

the Ni values.

The experimental results achieved until now

for crack formation directly at the surface with

an appropriate consideration of the Ni definition

show increasing numbers of cycles to crack initiation, if the surface area:

specimens of hardened and shot peened SAE

1045 at different stress amplitudes. (a) ra 1300 MPa. (b)

ra 1250 MPa. (c) ra 1200 MPa. (d) ra 1100 MPa.

(e) ra 1000 MPa. (f) ra 950 MPa. (g) ra 900 MPa.

Source: Ref 39

Fig. 24

loading, the lifetime of components is determined significantly by the stage of crack propagation, especially in the case of existing macro

residual stresses. Crack propagation as a consequence of fatigue loading without mean stresses

can be described, with the exception of the crack

initiation and the near-fracture stage, by the Paris

equation (see Eq 1):

da

C(DK)m

dN

(Eq 14)

m are constants. Mean stresses are taken into account by the Foreman equation:

da

C(DK)m

dN

(1 R)Kc DK

200

rs

200

(Eq 15)

intensity factor for plane stress. Both relationships are valid for crack propagation in materials

states without macro residual stresses. However,

itself is surrounded by a typical residual stress

field as shown in Fig. 25(a) for a high-strength

structural steel of the European grade S690QL1

(Ref 40, 41). A crack was produced by cyclic

loading up to a stress intensity range of DKI

47.4 MPa m. The distribution of the macro re-

104

As received

105

106

Autofrettaged

0.2

0.4

0.6

0.8

1.0

Ratio a/W

W in an autofrettaged and an untreated tube of

the SAE 4337 steel. Source: Ref 43

Fig. 27

400

12

8

4

0

4

8

Distance from crack tip, mm

(a)

12

10

150 bar

Crack propagation rate

(da/dN), mm/cycle

rs

200

200

8

4

0

4

8

Distance from crack tip, mm

from crack tip of the steel S690QL1 after a

mode I base load of DK 47.4 MPa m (a) and an overload of DK 94.8 MPa m (b). Source: Ref 40, 41

102

=2

103

104

=3

105

8

4

75 bar

225 bar

4

2

300 bar

0

1

2

Crack length (a), mm

Fatigue crack propagation rates da/dN in annealed and with different rolling pressures

deep-rolled steel AISI 304. Stress amplitude ra 320 MPa,

surface residual stresses rrs 200 MPa (75 bar), 350

MPa (150 bar), 400 MPa (225 bar), and 300 MPa (300

bar). Source: Ref 29

Fig. 28

Fig. 25

12

103

104

Stress relieved

As welded

105

106

10

20

30 40

60 80

length a0l for overloads with k 2 and 3.

Source: Ref 40, 41

Fig. 26

Untreated

12

12

(b)

106

400

103

Crack propagation rate da/dN versus stress intensity range DK in the heat-affected zone of

EN S355 steel in an as-welded condition and after a stressrelief heat treatment. Source: Ref 44

Fig. 29

y , which acts perpendicular to the crack flanks, shows maximum

compressive residual stresses of approximately

350 MPa at the crack tip. The alteration of this

distribution after application of 20 overload cycles with an overload ratio k 2 (DKI 94.8

MPa m) is given in Fig. 25(b). In front of the

crack tip, a larger maximum value and a larger

area with compressive residual stresses compared with Fig. 25(a) are developed. The influence of different overload cycles on the crack

propagation rate is shown in Fig. 26 for k 2

and 3. For both overload ratios, a delayed retardation of crack propagation occurs that is more

pronounced for k 3 than for k 2 due to the

effect of overload-induced compressive residual

stresses. Thus, by sufficiently high overloads,

crack arrest can be produced (Ref 42).

If a crack propagates into a macro residual

stress field, the crack propagation behavior can

be considerably influenced by magnitude and

distribution of the residual stresses. This is

shown in Fig. 27, in which crack propagation

rates within an autofrettaged tube of SAE 4337

steel are compared with those in the untreated

material (Ref 43). As a consequence of the autofrettaging process, a triaxial residual stress

state is created with very high compressive residual stresses near the inner surface of the tube.

In agreement with Eq 14, loading with a constant

stress intensity range results in a constant crack

propagation rate in residual-stress-free tubes.

However, if the crack propagates through the

compressive residual stress field at the inner surface of the autofrettaged tube, considerably

smaller crack propagation rates occur.

Crack velocity is influenced by work hardening as well as by residual stresses in mechanically treated surface layers. This is demonstrated

in Fig. 28 for differently surface-rolled AISI 304

stainless steel. The crack propagation rate da/

dN, which was determined by analyzing striations on cracked surfaces, increases with increasing rolling pressure and is considerably

diminished compared with untreated specimens

(Ref 29).

Finally, the influence of welding residual

stresses at the steel European grade EN S355 is

compared with the behavior of the same steel in

a stress-relieved condition in Fig. 29 in a da/dN

versus DK diagram (Ref 44). The crack propagation rate in the heat-affected zone is considerably lower in the as-welded state than in the

stress-relieved state due to the distribution of

welding compressive residual stresses.

Evaluation of Experimental Results. For

practical purposes, it is very important to know

that crack propagation through residual stress

fields can be modeled quantitatively by introducing an effective stress intensity range DKeff.

This can be seen in Fig. 30, which shows crack

propagation rates da/dN in welded and unwelded specimens of SAE 1019 steel (Ref 45).

Exactly at the welding seam, tensile residual

stresses of about 340 MPa exist that disappear

in a distance of 10 mm. Figure 30 presents crack

propagation rates in the unwelded base material

103

0.1 0 0.5 1.0

104

0.2

R = 0.4

2.0

105

106

10

20

30

40 50 60

(a)

103

R=0

0.5

104

105

106

1.0

2.0

6

10

20

30

40 50 60

(Eq 16)

on Kop is known, DKeff can be estimated and used

to predict the fatigue crack growth. Thus, the

double logarithmical plot of the crack propagation rate da/dN versus the effective stress intensity range DKeff in Fig. 30(c) shows that all measured data points form a narrow scatter band. In

welded specimens, the (da/dN)/DK relations are

completely changed (e.g., Fig. 30a and b). Obviously, for small DK values the crack propagation behavior is entirely controlled by the

welding residual stress state, which leads to

identical crack propagation rates irrespective of

the R ratio. However, outside the volume areas

(b)

103

104

105

106

10

20

30

300

Upcut milling

rs = 234 MPa

200

Downcut milling

rs = 242 MPa

kt = 2.5

=5

0

104

105

106

107

Number of cycles to failure, Nf

specimens of normalized plain carbon SAE

1045 steel after annealing, downcut milling and upcut milling. Source: Ref 4749

Fig. 31

103

104

105

106

10

20

30

40 50 60

(d)

Crack propagation rate da/dN versus stress intensity range DK in the base material (a) and (c)

and versus effective stress intensity range DKeff (b) and (d)

in the welded zone of the SAE 1019 steel. Source: Ref 45

Fig. 30

300

200

=1

= 0.4

= 0.1

kt = 1.7, = 2

kt = 2.5, = 5

100

kt = 4.4, = 15

kt = 2.5, = 2

Smooth

Notched

0

750 500 250

250

500

750

rs

Alternating bending fatigue strength of milled

smooth and notched specimens of normalized

plain carbon SAE 1045 steel versus surface residual stress.

Source: Ref 4752

Fig. 32

curves for each R value used are observed. For

R (pulsating compression), first increasing and then decreasing crack propagation rates

are observed indicating the influence of the tensile residual stress field. However, also in this

case, the crack propagation behavior can be described by a single, relatively narrow scatter

band if the influence of the welding residual

stresses on the DK values appearing at the crack

tip is taken into account (see Fig. 30d). In this

way, the crack propagation behavior in macro

residual stress fields can be quantitatively described if crack propagation data of macro residual-stress-free materials are available and the residual stress distribution is known.

S-N Curves

Characteristic Examples

Low-Strength Steel. Figure 31 shows S-N

curves for alternating bending of normalized

SAE 1045 steel (German grade Ck 45) (Ref 47

49). The notched specimens had a stress-concentration factor kt 2.5. The stress gradient at

the notch root dr/dz related to the maximum

stress r (normalized stress gradient g):

g

Annealed

rs = 0 MPa

100

40 50 60

(c)

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

As expected, with increasing R values, higher

crack propagation rates are observed. It is important to note that a mode I crack can only grow

during that portion of loading cycle where the

crack is open. This portion is influenced by the

R value itself and the near-crack-tip residual

stress distribution. The effective stress intensity

range can be determined quantitatively as the

difference between the maximum stress intensity

and the stress intensity where the crack opens:

1 dr

r dz

z0

(Eq 17)

was 5 mm1. All data are nominal stress amplitudes and are valid for a failure probability of

50%. The bending fatigue strength was evaluated at an ultimate number of cycles Nu 107.

By downcut milling and upcut milling, surface

residual stresses of 242 and 234 MPa, respectively, were generated. The corresponding S-N

curves are almost identical. A third batch of

specimens was annealed 2 h at 700 C after

downcut milling. The annealing results in a reduction of bending fatigue life and bending fatigue strength (Ref 4749).

The alternating bending fatigue strengths of

milled smooth and notched specimens with different geometries are plotted in Fig. 32 as a function of the surface residual stresses (Ref 4652).

Again, all data are given for a failure probability

of 50%, and the bending fatigue strengths are

nominal stress amplitudes at Nu 107. With

increasing stress-concentration factor and decreasing stress gradient, the bending fatigue

strength decreases. The influence of the stressconcentration factor is clearly visible from a

comparison of the specimens with the same

value g 2 mm1, but different values kt

1.7 and 2.5, respectively. On the other hand, the

increase of g from 2 to 5 mm1 at specimens

with kt 2.5 results in a significant increase of

bending fatigue strength. It is also interesting to

note that specimens with kt 4.4, g 15 have

a somewhat higher strength than specimens with

300

280

260

Deep rolled

240

220

200

180

104

105

106

107

Number of cycles to failure, Nf

320

280

rs

260

240

220

200

180

104

1000

Deep rolled

300

(a)

2 20 m, 30 m/s

500

2 6 m, 15 m/s

0

2 3 m, 15 m/s

500

105

106

107

Number of cycles to failure, Nf

0.1

0.2

Distance from surface, mm

(b)

600

500

400

rs = 221 MPa

300

200

100

0

104

19

kt = 1.7

=2

602

105

106

107

Number of cycles to failure, Nf

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

Fig. 34

notched specimens made from quenched-andtempered (600 C/2 h) plain carbon SAE 1045 steel by different grinding processes with the indicated two steps of

final feed and cutting speed. Source: Ref 4749

Fig. 35

S-N curves of specimens made from normalized plain carbon SAE 1015 steel in the asheat-treated state and after an additional deep rolling for

(a) push-pull loading and (b) rotating bending. After Ref 53

Fig. 33

residual stresses yield identical fatigue behavior.

It should be noted that the surface layer bearing

compressive residual stresses is rather small.

Contrarily, the generation of tensile residual

stresses causes a significant decrease of bending

fatigue strength and a rather small decrease of

finite fatigue life, which obviously vanishes at

high stress amplitudes.

The alternating bending fatigue strength of

ground smooth and notched specimen with different geometries made of the same material

state is plotted in Fig. 36 as a function of the

surface residual stresses (Ref 4752). Similar to

Fig. 32, the influences of the stress-concentration

factor and the normalized stress gradient are

clearly discernible. However, contrary to the

normalized steel there is distinct reduction of the

bending fatigue strength with increasing tensile

residual stress, this being more pronounced in

the comparison of smooth specimens with

notched specimens. In the range of compressive

residual stresses covered, the influence of residual stress on bending fatigue strength is rather

small.

The S-N curves of smooth specimens in the

ground state and after an additional shot peening

are compared in Fig. 37(a) (Ref 4749). There

is a distinct increase of the bending fatigue

strength by shot peening, but a rather small in-

400

= 0.4

Smooth

300

= 0.1

kt = 1.7

Notched

=2

200

100

0

500

=5

kt = 2.5

250

=2

0

250

500

750

rs

Surface residual stress (l ), MPa

smooth and notched specimens made from

quenched-and-tempered plain carbon SAE 1045 steel versus surface residual stress. Source: Ref 4752

Fig. 36

320

data points given in Fig. 32 to the same hardness,

it turns out that the bending fatigue strength is

hardly changed or slightly diminished at most,

if the residual stresses change from compressive

to tensile (Ref 49).

S-N curves for push-pull loading of smooth

specimens made from normalized SAE 1015

steel (German grade Ck 15) in the as-heat-treated

state and after an additional deep rolling are

shown in Fig. 33(a) (Ref 53). Again, the fatigue

behavior in the range of finite fatigue life and the

fatigue strength do not differ much. Figure 33(b)

shows a plot of S-N curves for rotating bending

of the same material states. Now, by deep rolling

finite fatigue life is increased by one order of

magnitude or more, and the bending fatigue

strength is increased significantly.

Medium-Strength Steel. Figure 34 shows SN curves that were determined in alternating

bending on notched specimens of quenched-andtempered (600 C/2 h) SAE 1045 steel (Ref 47

49). All data are valid for a failure probability

of 50% and Nu 107. Three batches of specimens were manufactured by grinding. The grinding parameters (final feed, cutting speed) and the

resulting depth distributions of residual stresses

are given in Fig. 35 (Ref 4749). Specimens

600

500

Shot peened

400

300

Ground

200

100

0

104

105

106

107

Number of cycles to failure, Nf

(a)

Nominal stress amplitude (n,a), MPa

to the residual stress state, there is no significant

influence on the bending fatigue strength, even

though the range of residual stresses covered

comes to more than 1000 MPa regarding specimens with kt 4.4, g 15 mm1. Careful

inspection of the hardness of the specimens

tested shows that a positive slope of the lines in

Fig. 32 is not related to the changing (macro)

residual stress state, but to different hardness of

the specimen and, hence, differences in the micro residual stress state produced by different

600

500

Shot peened

400

Milled

300

200

Ground

100

0

104

(b)

105

106

107

Number of cycles to failure, Nf

made from quenched-and-tempered (600 C/2

h) plain carbon SAE 1045 steel. (a) Smooth specimens after

grinding and after additional shot peening. (b) Notched

specimens after milling, grinding, and grinding with additional shot peening. Source: Ref 4749

Fig. 37

400

Smooth

= 0.4

300

the influence of the different manufacturing processes is almost negligible. The relative increase

of the bending fatigue strength by shot peening

is more pronounced compared with smooth

specimens. It is interesting to note that milled

specimens have a higher bending fatigue

strength than ground ones, even though they

have lower compressive residual stresses (159

200

Notched

kt = 1.7

=2

100

0

750 500 250

250

500

750

rs

Alternating bending fatigue strength of

quenched-and-tempered (600 C/2 h) plain

carbon SAE 1045 steel versus surface residual stress evaluated from Fig. 36 and 37

Fig. 38

fluence on finite fatigue life. Figure 37(b) compares S-N curves of notched specimens which

were milled, ground, and shot peened after

800

600

400

Notched

Smooth

200

1000

2000

3000

4000

Deep-rolling force, N

5000

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

57

specimens made from blank-hardened AISI

5115 steel in the as-blank-hardened and with different conditions shot peened states including one with a subsequently electropolished surface. 1, as-blank-hardened; 2,

shot velocity v 23 m/s, coverage c 100%, mean diameter of the shot d 0.6 mm; 3, v 53 m/s, c 100%,

d 0.3 mm; 4, v 53 m/s, c 100%, d 0.6 mm; 5,

v 81 m/s, c 600%, d 0.6 mm; 6, v 53 m/s, c

100%, d 0.6 mm, 100 lm surface layer electrolytically removed. After Ref 5456

Fig. 39

Fig. 41

600

rs = 400 MPa

500

400

300

200

923

100

0

104

kt = 1.7

=2

105

106

107

Number of cycles to failure, Nf

specimens made from quenched plain carbon

SAE 1045 steel after different grinding processes. Source:

Ref 4749

Fig. 42

1000

200

0

200

rs

rs

400

3

4

5

400

600

800

1000

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

Depth distribution of the residual stress in specimens made from blank-hardened AISI 5115

steel in the as-blank-hardened (1) and in different conditions of the shot peened state (3), (4), and (5) corresponding

to Fig. 39. After Ref 5456

Fig. 40

2 15 m, 30 m/s

500

2 9 m, 15 m/s

0

2 3 m, 15 m/s

500

0.1

0.2

Distance from surface, mm

notched specimens made from quenched plain

carbon SAE 1045 steel and ground with the two steps of

final feed and cutting speed indicated. Source: Ref 4749

Fig. 43

MPa).

Figure 38 shows plots of the alternating bending fatigue strengths evaluated from Fig. 37 as a

function of the surface residual stresses. The arrows mark the shift in bending fatigue strengths

and surface residual stresses produced by shot

peening. Additionally, data points of ground

specimens with negligible or tensile residual

stresses shown in Fig. 36 are included. In the

case of notched specimens, all data points lie on

a common line with the slope 0.154 except for

ground specimens with compressive residual

stresses at the surface. Regarding smooth specimens, the influence of tensile residual stresses

on the bending fatigue strength is much more

pronounced than the influence of compressive

residual stresses.

Figure 39 shows S-N curves for alternating

bending of smooth specimens (g 1 mm1) of

blank-hardened AISI 5115 steel (German grade

16 MnCr 5) determined in the unpeened and

various shot peened states including one with

electrolytically removed surface layer (Ref 54

56). The corresponding depth distributions of residual stresses are given in Fig. 40. From the

comparison of both figures, it becomes clear that

the surface residual stress is not a suitable parameter for the assessment of the influence of the

various treatments on the fatigue behavior.

The influence of the deep-rolling force on the

rotating bending fatigue strength of smooth and

notched specimens made from quenched-andtempered SAE 5135 steel (German grade 37 CrS

4) is shown in Fig. 41 (Ref 57). In both cases, a

maximum of the fatigue strength occurs at a certain force. However, the increase of the fatigue

strength of notched specimens (kt 2) by deep

rolling is much more pronounced than that of

smooth specimens. In the end, the optimal bending fatigue strength of notched specimens

which is a nominal stress amplitudeis higher

than the bending fatigue strength of smooth

specimens.

High-Strength Steels. S-N curves for alternating bending of notched specimens of

quenched SAE 1045 steel in differently ground

conditions are compared in Fig. 42 (Ref 4749).

The grinding parameters (final feed, cutting

speed) and the resulting depth distributions of

residual stresses are given in Fig. 43. Similar to

ground quenched-and-tempered specimens, the

fatigue behavior of specimens with very small or

compressive residual stresses (which have a very

small penetration depth) hardly differ. Tensile

residual stresses, however, not only cause a

strong reduction of bending fatigue strength, but

also of finite fatigue life.

In Fig. 44, the bending fatigue strength that

was evaluated from Fig. 42 and corresponding

data from tests on smooth specimens are plotted

as a function of the surface residual stresses. The

negative influence of tensile residual stresses on

the bending fatigue strength of smooth specimens is even more pronounced compared to

notched specimens. The influence of compressive residual stresses generated by grinding is

800

Smooth

= 0.4

600

400

200

0

1000

Notched

kt = 1.7

=2

500

500

orders of magnitude at high stress amplitudes.

The S-N curves of notched specimens of the

same steel state after grinding and after additional shot peening with shot of different hardness are shown in Fig. 45(b). Compared to

smooth specimens, shot peening produces a

much stronger increase of the bending fatigue

strength. Again, there is also a remarkable in-

stresses.

In Fig. 45(a), the S-N curves of smooth specimens in the ground state and after additional

shot peening are compared (Ref 4649). Similar

to quenched-and-tempered specimens (see Fig.

37), shot peening produces a significant increase

of the bending fatigue strength. Contrary to the

results of the medium-strength steel, however,

there is also a very pronounced increase of finite

1000

800

Smooth

= 0.4

600

400

Ground

Shot peened

Milled

200

Notched

kt = 1.7

=2

0

1000

500

500

1000

rs

1500

rs

smooth and notched specimens of quenched

plain carbon SAE 1045 steel versus surface residual stress.

Source: Ref 4749

and notched specimens made from quenched

plain carbon SAE 1045 steel with different surface conditions versus surface residual stress

Fig. 46

900

Shot peened

800

700

Ground

600

10

4

1

500

400

104

200

400

600

800

Surface hardness, HV 5

105

106

107

Number of cycles to failure, Nf

(a)

DRf relative to changes of surface roughness

height DR h versus surface hardness HV 5 for different heat

treated SAE 1045 and AISI 5115 steels. After Ref 58, 59

Fig. 47

1000

900

0

5458 HRC

800

700

600

500

400

104

Milled

Ground

105

106

107

Number of cycles to failure, Nf

(b)

Alternating bending S-N curves of specimens

made from quenched plain carbon SAE 1045

steel. (a) Smooth specimens after grinding and after additional shot peening. (b) Notched specimens after grinding,

milling, and grinding with additional shot peening with

shot of the indicated hardness. Source: Ref 4649

Fig. 45

rs

SAE 1045

AISI 5115

40

100

1000

crease of finite fatigue life. Additionally, the SN curve of milled specimen is included. Finite

fatigue life and bending fatigue strength of these

specimens are lower compared to shot-peened

ones, even though they contain very high surface

compressive residual stresses of 1200 MPa.

In Fig. 46, the bending fatigue strength data

already plotted in Fig. 44 are complemented by

data evaluated from Fig. 45. The arrows mark

the shift in bending fatigue strengths and surface

residual stresses produced by shot peening.

Similar to the discussion of Fig. 39 and 40, it

becomes obvious that the magnitude of surface

residual stress is not a suitable parameter for the

assessment of the influence of shot-peening-induced residual stresses on the fatigue strength.

200

325

400

280

200

600

185

0

10

Number of cycles, N

upcut milled specimens made from normalized plain carbon SAE 1045 steel during alternating bending fatigue at different nominal loading amplitudes versus

number of cycles. Source: Ref 49

Fig. 48

33, different machining processes or process parameters were applied to the specimens investigated producing different surface topographies.

In Fig. 47, the ratio DRf /DRh, which gives the

change of the alternating bending fatigue

strength of differently treated SAE 1045 and

AISI 5115 steels per lm increase of roughness

height is plotted as a function of the hardness at

the surface (Ref 58, 59). As is well known from

many other investigations, the susceptibility of

the bending fatigue strength to roughness increases with increasing hardness. For the steel

states investigated, a value of 2 MPa/lm can

serve as a guide. Since the roughness height of

the specimens did not vary by more than 2 lm,

the results plotted in Fig. 31 and 32 are hardly

influenced by surface topography. It should be

kept in mind, however, that a large increase of

surface roughness will significantly decrease the

fatigue strength even in low-strength material

states.

In low-strength steels, cyclic plasticity appears if the cyclic loading approaches the fatigue

strength. Extensive cyclic plasticity occurs in the

range of finite fatigue life, being combined with

cyclic softening and/or hardening processes (see

the section Some Aspects of Fatigue of Steels

in this article). Cyclic plasticity results in macro

residual stress relaxation, which is more rapid

the higher the cyclic loading is (see the article

Stability of Residual Stresses in this Handbook). Figure 48 (Ref 49) shows the relaxation

of the surface residual stresses during alternating

bending fatigue in notched upcut milled specimen made from normalized SAE 1045 steel

starting at an initial value of 590 MPa. The

bending fatigue strength amounts to 190 MPa

(see Fig. 32). During loading with an amplitude

of 185 MPa just below the fatigue limit, the

change of the macro residual stresses is almost

negligible. During all other fatigue loadings, extensive residual stress relaxation occurs. Therefore, almost no influence of macro residual

stresses is found in Fig. 31 and 32 (Ref 4752).

The amount of cyclic plasticity, which occurs

during cyclic loading resulting in a given finite

fatigue life or which corresponds to the fatigue

limit, decreases with increasing strength of the

steel. This results in decreasing residual stress

relaxation rates and finally in stable macro residual stresses. On the other hand, in a low-strength

steel, local work hardening may occur during

manufacturing processes such as milling, turning, or grinding. In other words, the micro residual stress state is changed, mainly by the increase

of the dislocation density. As discussed in the

article Stability of Residual Stresses in this

Handbook, the resistance of the micro residual

stress state against cyclic relaxation is much

higher compared to that of the macro residual

stress state. So one may assume that macro residual stresses that exist in a part of a component

that has undergone work hardening by manufacturing processes are more stable than expected

from the basic strength of the material. Looking

at Fig. 32, this relationship may be responsible

for the small reduction of the bending fatigue

strength by tensile residual stresses that occurs

in some of the specimen series investigated.

A much more important consequence of the

change of the micro residual stress state by local

work hardening is the direct alteration of the local fatigue behavior. Increasing strength reduces

the local cyclic plastic deformation at a given

external loading, thus increasing the number of

propagation rate of short cracks (Ref 60, 61).

Therefore, as is evident from Fig. 31, a residual

stress relief heat treatment that eliminates the

machining-induced work hardening reduces the

finite fatigue life and the fatigue limit of normalized SAE 1045 steel. At rather high loading

amplitudes, this influence vanishes because then

the cyclic deformation behavior is almost entirely governed by cyclic softening and hardening processes during the fatigue loading (see the

section Cyclic Deformation Behavior in this

article) irrespective of initial variations of the local work-hardening state.

A smaller influence of the micro residual

stress state appears in Fig. 33(a) and a somewhat

stronger one in Fig. 33(b). Here, it becomes clear

that even in a rather soft material state such as

normalized SAE 1015 steel, an improvement of

the fatigue behavior can be obtained from local

work hardening. In this case, the depth of the

surface layer of the smooth specimens influenced

by deep rolling is relatively large compared to

the milled specimens, which are the basis of Fig.

31 and 32. Nevertheless, push-pull loading (Fig.

33a) without stress gradient reveals a small influence of deep rolling, because of two major

effects. First, the crack initiation site may be

shifted below the work-hardened zone. Then, the

small increase of the fatigue strength mirrors the

benefit of omitting surface effects (such as spec-

600

Smooth

= 0.4

n,a = 200 MPa

rs

Ratio Rf/HV 5

Rh = 3 m

1

Rh = 15 m

0

0

200

400

600

800

1000

Surface hardness, HV 5

400

Notched

kt = 1.7, = 2

n,a = 150 MPa

200

0

0

DRf relative to changes of surface hardness

DHV 5 versus surface hardness HV 5 for different surface

roughnesses heights Rh. After Ref 58, 59

10

Fig. 49

600

=2

kt = 2.5

=5

1.5

kt = 1.7

400

200

200

0

0

200

400

106 107

10

102

103

104

105

106

Number of cycles, N

600

rs

Surface residual stress (l ), MPa

quenched-and-tempered (600 C/2 h) plain

carbon SAE 1045 steel versus surface residual stress.

Source: Ref 4749

Fig. 50

105

Notched

n,a = 350 MPa

kt = 1.7

=2

kt n,a = 595 MPa

=2

1.0

400

104

Smooth

= 0.4

n,a = 400 MPa

rs

2.0

103

(a)

Surface residual stress (l ), MPa

2.5

102

Number of cycles, N

(b)

Surface residual stress in smooth and notched

specimens made from quenched-and-tempered (600 C/2 h) plain carbon SAE 1045 steel versus

number of cycles (a) at a lifetime of approximately 2 106

cycles and (b) at a lifetime of approximately 105 cycles.

Source: Ref 4749

Fig. 51

crack initiation. Secondly, if crack initiation

occurs at the surface, the influence of the workhardened zone may also be rather small, becauseat a given loadinga stress redistribution may occur from the softer core region of the

specimen to the work-hardened zone. This effect

is combined with cyclic plasticity in the core region and, hence, with rather high loading amplitudes. The first effect predominates in the range

of the fatigue limit, where cyclic plasticity is

small, and the second one in the range of finite

fatigue life. On the other hand, during rotating

bending (Fig. 33b) with a distinct stress gradient,

crack initiation always occurs at the surface. If

the work-hardened zone is thick enough, there is

no stress redistribution from the core region

which is then purely elastically loadedto the

work-hardened zone. Again, at very high loading

amplitudes, the cyclic softening/hardening processes produced by the cyclic loading itself push

the influence of the initial work-hardening state

into background.

In Fig. 49, the ratio DRf /DHV, which gives

the increase of alternating bending fatigue

strength per unit hardness, is plotted as a function of the hardness itself (Ref 58, 59). The diagram shows that the sensitivity of the fatigue

strength to the hardness decreases with increasing hardness. Taking into account that the resistance of the macro residual stress state against

cyclic relaxation is low in low-strength steels, it

becomes clear from Fig. 49 thatbesides the

change of the surface topographythe alteration of the micro residual stress state is the governing parameter for the assessment of the influence of machining on the fatigue behavior of

components made from low-strength steel.

Medium-Strength Steel. A detrimental effect

of tensile grinding residual stresses appears in

Fig. 34 and 36 on alternating bending fatigue

strength, being more pronounced for smooth

than for notched specimens. Therefore, the fatigue notch factor decreases with increasing residual stress, as shown in Fig. 50. There is relaxation of surface residual stresses during

bending fatigue loadings of smooth and notched

specimens, which results in lifetimes of approximately 2 million cycles, as shown by Fig. 51(a).

The stress relaxation in the notched specimens

is significantly more pronounced than in smooth

ones. An elastic estimation of the stress in the

notch root r* kt rn,a yields a much higher

value than the stress amplitude in the smooth

specimens. Even though r* is only an upper

bound for the true stress at the notch root, it becomes clear that it is the higher local cyclic loading in notched specimens that causes the stronger

cyclic stress relaxation and, hence, the smaller

residual stress sensitivity of the bending fatigue

strength compared to smooth specimens. Regarding the relatively high loadings indicated in

Fig. 51(b), stress relaxation in notched specimen

is almost complete at the end of fatigue life, and

therefore only a small influence of initial macro

residual stresses on fatigue life at high loadings

is expected. This is shown in Fig. 34.

According to Fig. 34 and 36, the beneficial

effect of compressive grinding residual stresses

on bending fatigue strength is small regarding

smooth and almost negligible regarding notched

specimens. Again, the smaller effect in notched

specimens is due to a stronger residual stress relaxation. However, during cyclic loading corresponding to the fatigue limit, a considerable portion of the initial residual stresses is retained to

the ultimate number of cycles even in notched

specimens. Therefore, and in view of the small

penetration depth of the compressive residual

stresses (

20 lm, see Fig. 35), one would assume that the crack initiation site is shifted below the surface. However, this could not be

proved unambiguously in either smooth or

notched specimens (Ref 49). Therefore, it is

thought that very thin surface zones bearing

compressive residual stresses do not retard the

initiation and early growth of microcracks.

Regarding smooth specimens in Fig. 38, a

smaller effect of compressive residual stresses

produced by grinding or shot peening appears

compared to tensile grinding residual stresses,

even though crack initiation occurred always at

the surface. The small influence of compressive

rs

500

N 107

Before loading

1000

1500

0

0.1

0.2

0.3

0.4

0.5

Residual stress of smooth ground and additional shot peened specimens made from

quenched-and-tempered (600 C/2 h) plain carbon SAE

1045 steel before loading and after 107 cycles versus distance from surface. Source: Ref 4649

Fig. 52

rs

0

n,a = 460 MPa

N 103

n,a = 300 MPa

500

N 107

Before loading

1000

1500

0

0.1

0.2

0.3

0.4

0.5

Residual stress of notched ground and additional shot peened specimens made from

quenched-and-tempered (600 C/2 h) plain carbon SAE

1045 steel before and after loading at different nominal

stress amplitudes versus distance from surface. Source: Ref

4649

Fig. 53

to their small penetration depth. Contrarily, as

evident from Fig. 40, compressive residual

stresses generated by shot peening have a rather

large penetration depth. However, since the

bending fatigue strength of shot peened specimens (414 MPa) is much higher than that of

specimens with tensile grinding residual stresses

(195 MPa), also the residual stress relaxation is

more pronounced during corresponding cyclic

loadings. As shown in Fig. 52, residual stresses

relax significantly during fatigue loading, which

does not cause failure within 107 cycles. Furthermore, the surface roughness of the shot

peened specimens (Rh 35.6 lm) was much

higher than that of ground specimens (Rh 6.5

lm) (Ref 4649). If the bending fatigue strength

of shot peened specimens is corrected for the

high surface roughness by means of Fig. 47 and

is plotted in Fig. 38 at the surface residual stress

remaining after 107 cycles (see Fig. 52), one gets

a data point that lies significantly above the extrapolated relationship between bending fatigue

strength and tensile residual stresses.

The relationship between bending fatigue

strength and surface residual stresses of notched

specimens (Fig. 38) is described by a straight

line except for specimens with compressive

grinding residual stresses as already discussed.

In all cases, crack initiation occurred at the surface of the specimens that failed. Since the extrapolated straight lines for smooth and notched

specimens cross at a tensile surface residual

stress of approximately 750 MPa the fatigue

notch factor approaches a unity with increasing

tensile residual stress (see the paragraphs on medium-strength steel in the section Haigh Diagram in this article). Again, different residual

stress relaxation plays an important role. As

shown in Fig. 51(a) (Ref 49), and discussed previously, the residual stress relaxation in a smooth

specimen loaded just above the fatigue limit,

which failed after 2 million cycles, is significantly smaller than in a notched specimen that

failed after approximately the same number of

cycles. Figure 53 proves that residual stress relaxation is even more significant up to a distance

of 100 lm from the surface during fatigue loading of shot peened notched specimens, which did

not fail up to 107 cycles. During fatigue loading

resulting in a fatigue life of approximately 104

cycles, residual stress relaxation is almost complete. Therefore, no influence of macro residual

stresses appears in the low-cycle fatigue range

(see Fig. 37b).

From Fig. 51(a) and 53, it follows that at the

fatigue limit, the residual stress relaxation in shot

peened notched specimens with compressive

tensile residual stresses is more pronounced than

in ground notched specimens with tensile residual stresses. For this comparison, in Fig 51(a)

only the stress relaxation up to the appearance

of a crack (arrow at 5 105 cycles) may be

regarded, because the final relaxation is influenced by crack propagation and does not exist

in Fig. 53. Furthermore, the surface roughness

of shot peened notched specimens (Rh 18.0

(Rh 5.5 lm). Therefore, the question arises

why the data points of shot peened specimens do

not fall below the solid line drawn in Fig. 38.

Careful inspection of microhardness distribution

and the width of x-ray interference lines as a

function of the distance from the surface show

thatcompared to grindingshot peening produces a more intense work hardening which, in

addition, penetrates into a larger depth. The same

holds for smooth specimens. The work hardening reduces cyclic plasticity and retards crack

initiation. Hence, it becomes clear that the bending fatigue strength of medium strength steel is

strongly influenced by both the micro and the

macro residual stress states and their stability

against cyclic loading. However, in the low-cycle fatigue (LCF) range, the macro residual stress

state is unstable and the influence of the (more

stable) micro residual stress state vanishes in

view of the strong cyclic-softening processes occurring during this loading condition (see the

section Cyclic Deformation Behavior in this

article). Therefore, neither an influence of the

macro residual stress state nor an influence of

the micro residual stress state appears in Fig. 37

at high loading amplitudes.

Figure 40 shows that rather different depth

distributions of compressive residual stresses

can be produced in smooth specimens (g 1

mm1) made from a medium-strength steel by

shot peening with different peening parameters

(see also the article Stability of Residual

Stresses in this Handbook). To understand the

influence of the surface roughness, the work

hardening and the macro residual stress state

produced by shot peening on the resulting bending fatigue behavior it is necessary to discuss

crack initiation and crack propagation separately, and this is done in the section Propagating and Nonpropagating Cracks in this article.

Some relationships, however, should be pointed

out here. The highest bending fatigue strength

(curve 6 in Fig. 39) was determined for specimens shot peened according to curve 4 in Fig.

40. Then, a 100 lm thick surface zone was electrolytically removed, shifting the maximum

compressive residual stresses to the surface and

reducing the surface roughness. The smooth surface (Rh 3 lm) and the work hardening of the

surface zone oppose crack initiation. The relatively high compressive residual stresses at the

surface hinder microcrack propagation. This influence outweighs the larger maximum and the

greater penetration depth of the compressive residual stresses according to curve 5. In this shot

peening state, a higher surface roughness (Rh

43.8 lm) and a lower work hardening at the surface result in crack initiation at a lower stress

amplitude, and the lower surface compressive residual stresses offer less resistance to early microcrack propagation. On the other hand, in the

range of finite fatigue life, there is almost no advantage of the electropolished state 6 compared

to the initial shot peened state 4 and a clear disadvantage compared to state 5. Here, the higher

number of cycles to crack initiation (state 6) is

almost compensated (state 4) or overbalanced

(state 5) by a slower rate of crack propagation

due to the larger penetration depth andin the

case of state 5the higher amount of the compressive residual stresses. Since the surface

roughness, the surface work-hardening state and

the surface residual stress state (see Fig. 40) of

the material states 2 to 5 do not differ much, they

have similar resistance to crack initiation at the

surface. Hence, the different fatigue strength appearing in Fig. 39 is determined by the appertaining resistance to microcrack propagation,

which, in turn, depends on the depth distribution

of the compressive residual stresses. This is discussed in more detail in the section Propagating

and Nonpropagating Cracks in this article.

A very interesting example of the interaction

of changing surface roughness as well as micro

and macro residual stress state in a mediumstrength steel are given in Fig. 41. The strong

increase of the rotating bending fatigue strength

of notched specimens by deep rolling is clearly

attributed to the generation of micro- and macrostresses with a rather large penetration depth. It

is interesting to note that the resistance to microcrack initiation is hardly influenced by the deeprolling treatment (Ref 57), and the increasing

bending fatigue strength is almost entirely combined with an increasing resistance to micro-

2.5

2.0

kt = 1.7

=2

1.5

1.0

500

500

1000

rs

Fatigue notch factor kf of specimens made from

quenched plain carbon SAE 1045 steel versus

surface residual stress. Source: Ref 4749

Fig. 54

rs

0

Before loading

n,a = 750 MPa

500

N 107

1000

1500

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

Residual stress of smooth ground and additional shot peened specimens made from

quenched plain carbon SAE 1045 steel before loading and

after 107 cycles versus distance from surface. Source: Ref

4649

Fig. 55

stress amplitudes increasing with the deep-rolling force. At relatively high forces, however, the

bending fatigue strength decreases, because then

the impairment of the surface zone by the strong

increase of the surface roughness and microcracking as well as an unfavorable residual stress

state at the surface enhances the formation of

rather long initial cracks that only arrest at reduced stress amplitudes. The bending fatigue

strength of smooth specimens increases only

weakly with increasing rolling force. Regarding

the maximum value, an increase of only 15% is

achieved, compared to 120% in the case of

notched specimens. Since the stress gradient is

rather small in smooth specimens, there is some

speculation (Ref 56, 57) that the crack initiation

site is shifted below the surface by deep rolling.

Then, the small increase of the fatigue strength

mirrors the benefit of omitting surface effects

(such as specimen roughness, corrosion, oxidation) and the effect of a decreasing local stress

amplitude on the crack initiation. This mechanism is discussed in detail in the section Haigh

Diagram in this article. At high rolling forces,

however, the crack initiation occurs at the surface again, where it is enhanced by a high roughness, an increased roughness sensitivity due to

work hardening, and an unfavorable residual

stress state. Therefore, the bending fatigue

strength decreases, and there is no improvement

by deep rolling at all. In the end, the maximum

bending fatigue strength of the notched specimens (remember, evaluated as nominal stress

amplitude) is higher than the maximum bending

fatigue strength of the smooth specimens. These

results show that the gradients of the micro and

the macro residual stresses in comparison to the

gradient of the loading stresses are of great importance, and this will become even more evident in the next section.

High-Strength Steels. The S-N curves of

quenched SAE 1045 steel in Fig. 42 prove that

there is a detrimental effect of tensile grinding

residual stresses on bending fatigue strength and,

contrary to the quenched-and-tempered state (see

Fig. 34) also on finite fatigue life. This is because

there is only a weak relaxation of the initial residual stresses even at rather high stress amplitudes. Similar to quenched-and-tempered specimens, compressive grinding residual stresses

with very small penetration depth (

20 lm, see

Fig. 43) have almost no effect on bending fatigue

strength and finite fatigue life. Therefore and

since in the quenched state the compressive residual stresses are retained to the largest part during fatigue loading in the range of the fatigue

limit, one may expect that crack initiation occurs

below the surface. Analogous to the quenchedand-tempered state, this could not be proved definitely (Ref 49). Since in the quenched state a

grinding-induced work softening occurs close to

the surface and an influence of the surface roughness always exists, it is very probable that crack

initiation occurs at the surface. Then, one must

conclude that compressive residual stresses with

a penetration depth of only 20 lm do not retard

after crack initiation.

The extrapolated bending fatigue strength versus surface residual stresses curves in Fig. 44

cross in the range of tensile residual stresses at

1125 MPa. This means that the fatigue notch

factor becomes unity. On the other hand, at negligible or compressive residual stresses, the fatigue notch factor approaches the stress-concentration factor, as shown in Fig. 54. Qualitatively

similar to the quenched-and-tempered state, the

influence of tensile residual stresses is stronger

than that of compressive residual stresses and the

residual stress sensitivity of notched specimens

is smaller than that of smooth ones. However,

this cannot be explained by different residual

stress relaxation so far. Careful x-ray analyses

prove that in notched specimens loaded with a

stress amplitude of 450 MPa and in smooth specimens loaded with 750 MPa (corresponding to

the fatigue limit of each) the relaxation of compressive residual stress is small and of equal

amount. This is not astonishing, since an elastic

estimation of the stress amplitude in the notch

root (r* kt r 765 MPa) yields almost the

same value as the stress amplitude of the smooth

specimen. No corresponding measurements for

smooth and notched specimens with tensile residual stresses were performed. However, since

the fatigue limit is reduced by tensile residual

stresses, it is clear that their relaxation is negligible.

A comparison of Fig. 45(a) with Fig. 37(a)

shows that the shot peening treatments performed increases the bending fatigue strength of

smooth specimens in the quenched state less

than in the quenched-and-tempered state. In

view of the different residual stress states and

their different stabilities in both heat treating

states, this finding is rather astonishing. In Fig.

55, the depth distribution of residual stresses in

shot peened quenched specimens before fatigue

loading and after 107 cycles at the fatigue limit

is shown. In the range of specimen scatter, the

results are identical. Contrary to quenched-andtempered specimen (see Fig. 52), the shot peening residual stresses are much higher and much

more stable in quenched specimens. However,

in the former, crack initiation occurs at the surface. Hence, the shot-peening-induced changes

of the micro and the macro residual stress state

have a strong influence on the measured bending

fatigue strength. In smooth quenched specimens,

cyclic loading at the fatigue limit initiates cracks

below the surface. Therefore, the influence of the

shot-peening-induced changes of the residual

stress state have a rather small influence on the

measured bending fatigue strength. This is

treated more detailed in the section Modeling

the Influence of Residual Stresses on Fatigue Behavior. Contrarily, in the range of finite fatigue

life, crack initiation in both states occurs at the

surface, and there is a strong influence of shot

peening regarding quenched specimens and a

very small one regarding quenched-and-tempered specimens. This finding is directly correlated with very different residual stress relaxa-

definitely complete in quenched-and-tempered

specimens after 50,000 cycles at a stress amplitude of 570 MPa, there is hardly any change of

the surface residual stresses in the quenched

specimens after the same number of cycles even

at 1200 MPa (Ref 49), although there may be

some residual stress relaxation below the surface.

Regarding notched specimens, a different picture appears (compare Fig. 45b with Fig. 37b).

Now, shot peening of quenched specimens pro-

800

Smooth

= 0.4

600

400

Ground

Shot peened

Milled

200

Notched

kt = 1.7

=2

0

1500 1000 500

0

500

1000

rs

Maximum residual stress (max ), MPa

and notched specimens made from quenched

plain carbon SAE 1045 steel with different surface conditions versus surface residual stress

Fig. 56

Before loading

500

rs

N 107

1000

1500

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

(a)

after

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

stresses occur.

Stresses on Fatigue Behavior

Haigh Diagram

Basic Relationships. One obvious way to account for the influence of (macro) residual

stresses on the fatigue behavior is to treat them

as local mean stresses. In doing so, one has to

realize that there are several important differences between (loading) mean stresses and residual stresses, as discussed in the section Comparison between Loading Stresses and Residual

Stresses in this article. Figure 58 shows a Haigh

diagram for smooth and notched specimens

made from a medium-strength steel (Ref 1, 62).

The Goodman approximation (see Eq 9) is used

to account for the influence of residual stresses

on the fatigue strength. If the amount of the minimum stress or the maximum stress in smooth

specimens does not exceed the critical stress amplitude ra,crit, which is a function of the cyclic

yield strength (see the article Stability of Residual Stresses in this Handbook), the residual

stresses do not relax, and the line AB gives the

influence of the residual stress on the fatigue

strength. Then, all combinations of residual

stress and stress amplitude inside the shaded area

result in neither residual stress relaxation nor fatigue failure. However, if the amount of the minimum stress or the maximum stress exceeds the

critical stress amplitude, it is assumed that the

residual stresses relax to the value given by the

points A and B, respectively, and the fatigue

strength remains constant at the value given by

these points. In the case of notched specimens,

the cyclic yield strength and the notch fatigue

strength (both in terms of nominal stress amplitudes) are less than the respective values of

smooth specimens. However, the ultimate tensile

strength of notched specimens is larger than that

of smooth ones in such a material state because

Before loading

500

rs

finite fatigue life. Milling, which may generate

rather high compressive residual stresses, is less

effective. In all cases, crack initiation was observed at the surface. However, as evident from

Fig. 46, the surface residual stress state is not an

appropriate parameter to assess bending fatigue

life in the presence of compressive residual

stresses. In Fig. 56, the bending fatigue strength

is plotted as a function of the maximum residual

stresses. Again, the arrows mark the shift of the

residual stresses and bending fatigue strength

obtained from shot peening. Now, the data of

ground notched specimens with tensile or negligible residual stresses and of notched specimens peened with shot of different hardness are

described satisfactorily by one common line

covering a residual stress range of 2350 MPa.

The data points of ground and those of milled

specimens bearing compressive residual stresses

fall below this line, which has the slope 0.214.

In Fig. 56, it becomes even more clear than in

Fig. 46 that the improvement of the bending fatigue strength of smooth specimens by shot

peening is very limited. As a consequence, the

notch effect on bending fatigue strength is completely eliminated at high compressive residual

stresses.

It is interesting to look at the stability of very

high compressive residual stresses. As already

shown in Fig. 55, fatigue loading of smooth

specimens in the range of the fatigue limit does

not cause any significant change of the residual

stress state. As shown in Fig. 57(a), the same

holds for notched specimens peened with shot of

relatively low hardness, which have a very similar depth distribution of the residual stresses.

Elastic calculation of the minimum stress 70 lm

below the notch root (at the maximum of the

compressive stresses) yields the value 2035

MPa. This stress value may occur more than 107

times without effecting any change of the residual stress state. As shown in Fig. 57(b), notched

specimens peened with shot of higher hardness

bear even higher compressive residual stresses.

In this case, elastic calculation of the minimum

stress in a depth of 100 lm yields the initial

value 2475 MPa and the value 2130 MPa

1000

N 107

1500

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

(b)

Residual stress of notched ground and additional shot peened specimens made from

quenched plain carbon SAE 1045 steel before loading and

after 107 cycles versus distance from surface. (a) Shot

peened with shot of a hardness of 46 to 50 HRC. (b) Shot

peened with shot of a hardness of 54 to 58 HRC. Source:

Ref 4649

Fig. 57

Fig. 58

Haigh diagram. Bending fatigue strength Rf of smooth and notched specimens made from a medium-strength

steel versus residual stress. After Ref 1, 62

of the triaxial stress state in the interior of the

notched specimens. Now, the Goodman relationship holds between points C and D, and residual

stress relaxation occurs outside the lightly

shaded area. From these relationships, it is expected that the residual stress sensitivity of

notched specimens is less than that of smooth

specimens.

Application to a High-Strength Steel. Figure 59 shows Haigh diagrams for smooth and

notched specimens made from a high-strength

steel, which are adopted to Fig. 44, 46, and 56.

In Fig. 59(a), the data points give the correlation

(a)

(b)

Haigh diagrams. Bending fatigue strength Rf of

smooth and notched specimens made from

quenched plain carbon SAE 1045 steel with different surface conditions. (a) Versus surface residual stress. (b) Versus

maximum residual stress

Fig. 59

stress amplitude at the surface. In Fig. 59(b), the

data points give the correlation between the maximum residual stress and the nominal stress amplitude at the locus of the maximum residual

stress. In the range of tensile residual stresses,

this is always the surface. The arrows connecting

some data points illustrate the residual stress relaxation that occurs during bending fatigue loading. The ultimate tensile strength Rm 1910

MPa of smooth specimens was taken from Ref

49. The ultimate tensile strength Rm,no of notched

specimens is unknown, but the value 2000 MPa,

which results from an extrapolation of the data

given in Fig. 44, is reasonable for this highstrength material state and the stress-concentration factor 1.7 (Ref 63, 64). The cyclic yield

stresses of smooth and notched specimens are

also unknown. However, using the results shown

in Fig. 55 (smooth specimens) and Fig. 57(a)

(notched specimens), a border line for residual

stress relaxation can be obtained. As obvious

from the figures and discussed in the previous

section, the cyclic loading indicated there causes

no residual stress relaxation. The corresponding

data points are given in Fig. 59(b) as open square

(smooth specimens) and open circle without arrow (notched specimens). Assuming that residual stress relaxation starts at a limiting amount

of the minimum or maximum stress irrespective

of the fractions of the mean stress and the stress

amplitude, one gets the shaded areas in which

the residual stresses are stable. These areas are

transferred to Fig. 59(a), too. The two open circles connected with an arrow illustrate the

change of the amount of the surface residual

stress (Fig. 59a) and of the minimum stress (Fig.

59b) during the cyclic loading of notched specimens indicated in Fig. 57(b). In milled specimens, there is also some residual stress relaxation as indicated by the triangles connected with

an arrow. Obviously, the amounts of the minimum stresses are reduced to a value corresponding to the border line, confirming this estimation.

Extrapolation of this line to zero residual stress

yields the values 1785 MPa (smooth specimens)

and 1660 MPa (notched specimens), respectively, which are higher than the corresponding

yield strengths (Ref 49).

Even though there are some uncertainties in

establishing the relationships in Fig. 59, the following conclusions can be drawn for a highstrength steel with tensile residual stresses:

Residual stress relaxation has no influence re-

smooth and notched specimens made from

quenched-and-tempered (600 C/2 h) plain carbon SAE

1045 steel with different surface conditions versus surface

residual stress

Fig. 60

on the fatigue strength.

The influence of tensile residual stresses on

the bending fatigue strength corresponds satisfactorily to the Goodman approximation.

This means that the residual stress sensitivity

and the mean stress sensitivity of the bending

fatigue strength is almost the same.

In this formalism, the different residual stress

sensitivity of smooth and notched specimens,

which causes a reduction of the fatigue notch

factor with increasing residual stress (see Fig.

54), is related to the different stress states in

smooth and notched specimens via the different ultimate tensile strengths.

In the range of compressive residual stresses,

the relationships are much more difficult. The

fatigue strength of smooth specimens is much

less than expected by the Haigh diagrams, regardless of whether the existing surface or the

maximum residual stresses are concerned. In the

case of shot peened specimens, this finding is

clearly related to subsurface crack initiation.

Therefore, neither the stress state at the surface

(Fig. 59a) nor the stress state in the depth of the

maximum compressive stresses (Fig. 59b) is directly relevant for the fatigue strength. Regarding notched specimens, the fatigue strength of

shot peened states is much higher than expected

by the Haigh diagram, if it is correlated with the

surface residual stresses (Fig. 59a). This simply

means that crack initiation must occur at the surface, if the specimens are loaded with amplitudes

in the range of the fatigue limit. On the other

hand, the stress amplitude in the depth of the

maximum of the compressive stresses of shot

peened specimens is somewhat lower than expected from the Haigh diagram (Fig. 59b). The

same holds for the data points of milled specimens, which have maximum compressive residual stresses at the surface. These findings can

only be understood if crack initiation and crack

propagation or crack arrest are treated separately.

This is done in the sections Concept of Local

Fatigue Strength and Propagating and Nonpropagating Cracks.

Application to a Medium-Strength Steel. In

medium-strength steels, residual stress relaxation definitively influences the relationships in

the Haigh diagram. Since the manufacturing process, which generates the macro residual

stresses, almost always changes the micro residual stresses and, hence, the local cyclic yield

strength, there are no unique borderlines for the

onset of residual stress relaxation as assumed in

Fig. 58. Nevertheless, Fig. 60 tries to bring the

results shown in Fig. 36 and 38 into a Haigh

diagram.

Regarding smooth specimens, the ultimate

tensile strength was taken from Ref 49. In

ground specimens with tensile residual stresses,

stress relaxation (marked by an arrow) brings the

corresponding data point rather close to the

Goodman line. The border lines for the onset of

residual stress relaxation are constructed using

this data point. However, these border lines are

not valid for shot peened specimens, in which

the residual stresses relax to a much smaller extent than expected as shown by the lower arrow.

Obviously, the resistance to residual stress relaxation is enhanced by shot peening, which results

in a stronger work hardening and, hence,

stronger increase of micro residual stresses than

grinding. Work hardening, however, should also

increase the fatigue strength at zero residual

stress and, hence, should produce deviations

from the solid line established with the aid of the

data points of ground specimens. However, as

already discussed in the paragraphs on medium-

strength steel in the section Evaluation of Experimental Results, there is a detrimental effect

of a very high surface roughness. The second

arrow illustrates the correction, which is based

on Fig. 47, for this effect. Compressive residual

stresses produced by grinding show some relaxation, although their initial value is inside of the

border lines. This finding is consistent with the

consideration given so far, since work hardening

is almost negligible in this specimen. Moreover,

the data point corresponding to the relaxed compressive grinding residual stresses still falls below the Goodman line. This finding was already

attributed to the low penetration depth of the

grinding residual stresses.

The residual stress sensitivity of notched specimens with kt 1.7 is significantly smaller than

that of smooth specimens, even though the consideration is based on the relaxed residual

stresses. Therefore, the different residual stress

sensitivity of the fatigue strength of smooth and

notched specimens, respectively, is not solely

based on different residual stress relaxation, and

the third conclusion given previously for a highstrength steel is also valid for a medium-strength

steel to some extent. From an extrapolation of

the Goodman line plotted, an ultimate tensile

strength of 1133 MPa is obtained, which is significantly higher than the ultimate tensile

strength of smooth specimens. Again, this finding results from the triaxial stress state in the

interior of the notched specimens and is consistent with data measured on other mediumstrength steels (Ref 63, 64). Regarding material

states with compressive residual stresses, the

data point of milled specimens comes rather

close to the Goodman line. No measurements

were done concerning the stability of this residual stress state. Since the data point falls inside

of the border lines, the residual stress state

should be stable. Contrarily, there is some relaxation of the compressive residual stresses in

ground specimens even though their initial value

falls inside of the border lines, too. Again, this

is a consequence of different micro residual

stress states. Similar to smooth specimens, the

data points fall below the Goodman line as a

result of the small penetration depth of the com-

1000

800

SAE 1045

600

400

AISI 304

300

200

100

103

Untreated

Shot peened

Deep rolled

105

106

104

Number of cycles to failure, Nf

107

from normalized SAE 1045 and AISI 304 steel in different

surface conditions. After Ref 28

Fig. 61

holds for notched specimens with kt 2.5, g

5. The data points of shot peened specimens fall

above the expected relationship as a consequence of pronounced work hardening, which

increases both fatigue strength and resistance of

the compressive residual stresses against relaxation.

From the relationships shown in Fig. 60, the

following conclusions can be drawn for a medium-strength steel:

Residual stress relaxation plays a significant

fatigue behavior.

Residual stress relaxation depends on the

work hardening produced by the various

manufacturing processes. Not only the

amount, but also the penetration depth of

work hardening is important. As a consequence, even for one initial material state

there are no unique borderlines for the onset

of residual stress relaxation.

If comparable work-hardening states (or micro residual stress states) are concerned, the

influence of relaxed residual stresses on the

bending fatigue strength corresponds satisfactorily to the Goodman approximation. This

means that the relaxed residual stress sensitivity and the mean stress sensitivity of the

bending fatigue strength is almost the same.

In this formalism, the different initial residual

stress sensitivity of smooth and notched specimens, which causes a reduction of the fatigue

notch factor with increasing residual stress, is

related to both the different residual stress relaxation and the different stress states in

smooth and notched specimens via the different ultimate tensile strengths.

Damage Parameters

As discussed in the section Lifetime Behavior, there are some so-called damage parameters

that correlate fatigue life with loading parameters. The well-known Manson-Coffin relationship (Eq 5) is not applicable to material states

with residual stresses, because it does not account for mean stresses. On the other hand, the

Ostergren parameter (Eq 12) and the Smith-Watson-Topper parameter (Eq 11) include the mean

stress via the maximum stress. With the limitations already outlined in the previous section,

one could try to use these parameters to assess

the effect of residual stresses on fatigue life.

However, the application of the Ostergren parameter requires the knowledge of the local

plastic strain amplitude, which in turn needs sophisticated modeling of the local cyclic elasticplastic deformation behavior. This seems not to

be a useful approach. However, the Smith-Watson-Topper parameter (SWT parameter) is easier

to evaluate and was used in Ref 28 to account

for surface treated steels (normalized SAE 1045

and AISI 304). Figure 61 shows characteristic

plots of the SWT parameter versus the number

of cycles of stress-controlled tests on both steels

in comparison with the untreated states. For each

steel, a unique curve is obtained for the three

material conditions. Hence, concerning these

steels the quantities that are included in the SWT

parameterthe maximum stress and the total

strain amplitude (see Eq 11)form a reliable

basis for the assessment of the influence of residual stresses on the fatigue life.

Basic Relationships. As shown in the previous section, in the paragraphs on applying the

Haigh diagram to high-strength and mediumstrength steels, the use of the Haigh diagram for

the assessment of the effect of residual stresses

on the fatigue behavior has serious limitations.

Regarding, for example, smooth specimens

made from a high-strength steel, shot peening

produces much less improvement in fatigue

strength than expected from the residual stress

state. This finding is combined with subsurface

crack initiation. Regarding medium-strength

steel, difficulties arise because not only the

(macro) residual stress state is changed by manufacturing processes, but also the micro residual

stress state. The latter influences the fatigue

strength at zero mean stress and the resistance of

the residual stress state against relaxation. Therefore, it is appropriate to look for the local fatigue

strength, which depends on the local micro residual stress state, the local macro residual stress

state, and (as far as crack initiation at the surface

is concerned) on the topography of the surface

of the component. Strictly speaking, the local resistance against fatigue crack initiation is concerned, as shown later.

The concept of the locally effective fatigue

strength, which has its origin in Ref 38 and 39,

enables quantitative predictions of the effect of

the depth distributions of residual stresses on the

locus of crack initiation as well as on the fatigue

strength. The basic assumption of the concept is

that a crack can only be initiated at or below the

surface if the local loading stress exceeds the local fatigue strength. Especially in the case of

relatively hard materials, for example, hardened

steels, this concept yields to a good estimation

of the corresponding properties. For that purpose, it is necessary to have a good knowledge

of the depth distributions of the fatigue strength

in the residual-stress-free condition R0f (z) as well

as of the macro residual stress rrs(z) and the residual stress sensitivity m(z). The locally effective fatigue strength Rf(z) as a function of the

distance x from the surface is calculated by the

relationship:

Rf(z) R0f (z) m(z) rrs(z)

(Eq 18)

where the residual stress sensitivity m(z) approaches the mean stress sensitivity M of the

Goodman relationship (e.g., Eq 9) if the residual

stresses are stable. Then, the residual stress sensitivity is determined approximately by m(z)

R0f (z)/Rm(z) (Ref 39). However, if residual stress

relaxation occurs, the residual stress sensitivity

m is smaller than the mean stress sensitivity M,

if the initial residual stress distribution is used

(Ref 39, 62). However, if the relaxed residual

stress distribution is used, the residual stress sen-

sitivity m again approaches the mean stress sensitivity M as shown in the paragraphs applying

the Haigh diagram to medium-strength steels in

the previous section. The depth distributions of

the tensile strength Rm(z) and of R0f (z) can be

estimated from appropriate correlations with

Notched

Residual stress, rs

Residual stress, rs

Smooth

0

(a)

(b)

N f = 104

a and Rf,loc

a and Rf,loc

N f = 104

Nf =

Loading stress

0

Nf =

0

(c)

(d)

Distribution of residual stress (a) and (b) and distribution of loading stress and local fatigue strength for loadings

in the range of finite and infinite lifetime (c) and (d) for smooth (a) and (c) and notched (b) and (d) specimens

made from a high-strength steel

Fig. 62

Propagating crack

Not propagating

crack

d = 0.6 mm

1000

Rf

800

Loading

stresses

1200

d = 0.3 mm

1200

1000

Rf

800

600

600

0

0.1

0.2

0.3

Distance from surface, mm

0.4

400

rs

400

800

1200

0.4

0.1

0.2

0.3

Distance from surface, mm

0.4

400

800

1200

0

(c)

0.1

0.2

0.3

Distance from surface, mm

(b)

400

l, MPa

rs

l, MPa

(a)

0.1

0.2

0.3

Distance from surface, mm

0.4

(d)

Distribution of the local fatigue strength and the loading stress for different stress amplitudes with marking of

the locus of crack initiation (a) and (b) and distribution of residual stress (c) and (d) for specimens made from

a quenched SAE 1045 steel after shot peening with shot with a mean diameter of 0.6 mm (a) and (c) and 0.3 mm (b) and

(d), respectively. After Ref 38, 39

Fig. 63

(see e.g., Ref 65).

Application to High-Strength Steels. In a

high-strength steel, manufacturing processes

may generate very large (macro) residual

stresses, as shown by several examples in the

paragraphs on high-strength steel in the section

Characteristic Examples. Contrarily, the

changes of the micro residual stress state are

rather small. Therefore, the local fatigue strength

is almost entirely determined by the initial material state and the depth of the (macro) residual

stresses. Figure 62(a) and (b) show depth distributions of compressive residual stresses in

smooth and notched bending specimens produced by shot peening. In Fig. 62(c) and (d), the

resulting depth distributions of the fatigue

strength for infinite life and for one arbitrarily

chosen finite life are plotted together with the

corresponding depth distributions of the loading

stress. It becomes clear from this figure that the

improvement of the fatigue strength for infinite

life of smooth specimens (low stress gradient)

by shot peening is small because crack initiation

occurs below the surface in a depth with vanishing residual stresses. Contrarily, with increasing

stress amplitude, the crack initiation site is

shifted toward the surface, and the shot-peeninginduced stresses influence crack initiation and

early fatigue crack growth. Hence, there is a

large influence of shot peening on the low-cycle

fatigue strength and a small influence on the

high-cycle fatigue strength. This is shown in Fig.

45(a). Contrarily, crack initiation in notched

specimens (high stress gradient) occurs always

at the surface and both low-cycle and high-cycle

fatigue strength are improved, as shown in Fig.

45(b).

The assumption of the local fatigue concept

that fatigue life determining cracks initiate only

in places where the loading stress exceeds the

local fatigue strength is verified by several experimental results (e.g., Ref 38, 39, 62). Figure

63 presents for two series of specimens of the

aforementioned investigations from Ref 38 and

39 on hardened and shot peened SAE 1045 steel

graphs of the depth distribution of the local fatigue strength (see Fig. 63a and b). Rf,loc can be

calculated from the compressive residual stress

distributions (Fig. 63c and d) under the assumption of uniform hardness. The closed circles

mark the initiation sites of damage-relevant

cracks for each load stress distribution, which

could be taken from SEM of the fracture surface

(see Fig. 24). As can be seen, the actual crack

initiation positions agree with the expectation after the concept of the local fatigue strength. For

high stress amplitudes (1150 MPa), the cracks

that determine fatigue life can initiate at the surface. For lower stress amplitudes (1150 MPa),

the damage behavior is changed. Of course,

stress amplitudes between 1000 and 1100 MPa

are able to initiate cracks at the surface (open

circles in Fig. 58a and b). However, these stress

amplitudes are too small to propagate the cracks

into greater depth because of the high compres-

sive residual stresses existing there. This aspect

is treated in more detail in the paragraphs on

high-strength steel in the section Propagating

and Nonpropagating Cracks. At stress amplitudes below 1000 MPa, cracks are generated

only beneath the surface as expected by the plotted straight lines of the loading stress in comparison with the courses of the local fatigue

strength (Fig. 62a and b). The fatigue strength Rf

is determined by the maximum loading stress

amplitude, which in any cross section of the

specimen exceeds the local fatigue strength and

initiates a crack that is able to propagate. The

corresponding distributions are plotted in Fig.

63(a) and (b) as dashed straight lines.

Figure 64 applies the concept of local fatigue

strength to shot peened smooth and notched

specimens according to Fig. 55 to 57. In notched

specimens peened with shot of low (Fig. 64a)

and high (Fig. 64b) hardness and loaded in the

range of the fatigue strength, crack initiation is

predicted at the notch root, as actually observed.

However, the surface loading stress amplitude is

much higher than the surface fatigue strength.

The same result is obtained from the Haigh diagram in Fig. 59(a). Therefore, one has to assume that loading stress amplitudes below the

fatigue limit can initiate a crack at the surface

by the maximum loading stress amplitude that

does not initiate a crack. This shows the limitations of the concept of the local fatigue strength

and is treated in more detail in the paragraphs on

high-strength steel in the section Propagating

and Nonpropagating Cracks.

Figure 64(c) shows that crack initiation in

smooth specimens is expected below the surface

where the local fatigue strength is hardly increased by compressive residual stresses. This is

verified by metallographical investigation (Ref

46, 49). This is the reason for the small improvement of the bending fatigue strength by shot

peening shown in Fig. 45, 46, 56, and 59. However, from Fig. 45(a) it becomes clear that shot

peening significantly improves finite fatigue life.

This finding cannot be understood on the basis

of the local fatigue strength alone and is treated

in the paragraphs on high-strength steel in the

section Propagating and Nonpropagating

Cracks.

In the case of notched specimens finished by

milling (see Fig. 45b, 46, and 56), the concept

of local fatigue strength predicts crack initiation

below the surface, because the gradient of the

residual stresses is steeper than the gradient of

the loading stresses. Hence, the increase of the

1500

Local fatigue

strength

1000

Cracks

Loading stress

500

0

0

0.2

0.4

(a)

0.2

0.4

0.2

0.4

(b)

(c)

Distribution of the local fatigue strength and the loading stress with marking of the locus of crack initiation

for notched (a) and (b) and smooth (c) specimens made from a quenched SAE 1045 steel after shot peening

with a shot of a hardness of 46 to 50 HRC (a) and 54 to 58 HRC (b) and (c), respectively. Source: Ref 49

Fig. 64

1500

1000

Local fatigue

strength

500

Loading stress

0

0

0.2

0.4

(a)

0.2

0.4

(b)

Distribution of the local fatigue strength and the loading stress with marking of the locus of crack initiation

for notched (a) and smooth (b) specimens made from a quenched-and-tempered (600 C/2 h) SAE 1045 steel

after shot peening with a shot of a hardness of 46 to 50 HRC

Fig. 65

in view of the high surface compressive residual

stresses (see the Haigh diagram in Fig. 64a), but

larger than in shot peened smooth specimens

with subsurface crack initiation. This is a consequence of the higher loading stress gradient in

notched specimen compared to smooth specimens. However, crack initiation below the surface of these specimens was not proved (Ref 49).

Application to Medium-Strength Steels.

The application of the concept of the local fatigue strength to medium-strength steels is somewhat more difficult compared to high-strength

steels because residual stress relaxation becomes

important. Therefore, a knowledge of the initial

depth distribution of the residual stresses and of

the residual stress sensitivity m or of the relaxed

residual stress distribution is necessary. In Fig.

65, the concept is applied to shot peened smooth

and notched specimens made from quenchedand-tempered SAE 1045 steel. The initial and

relaxed residual stress distributions are given in

Fig. 52 and 53. Again, in notched specimens

(Fig. 65a) crack initiation at the surface is predicted as actually observed. Regarding smooth

specimens, crack initiation below the surface is

expected. Contrarily, crack initiation at the surface is found in metallographical investigation.

In both cases, the difference between local loading stress and local fatigue strength is small

throughout the depth range under consideration.

Therefore, factors such as the high roughness of

the smooth specimens may well account for the

deviation of the expected locus of crack initiation.

notched specimens or components that are

loaded in the range of the fatigue strength, cracks

may initiate in the root of sharp notches and may

arrest in a certain depth, where the driving force

for crack propagation falls below its threshold

value because of the steep drop of the loading

stress. In most cases, such sharp notches are not

relevant for components (Ref 57, 66). However,

in the presence of high residual stresses, the interaction of loading stresses and residual stresses

may produce strong gradients of the driving

force for crack propagation making crack arrest

possible even in notches with low stress-concentration factor. On the other hand, in smooth specimens with large compressive residual stresses

at and below the surface, crack initiation may

occur below the surface because the gradient of

loading stresses is small. Then, the question

arises whether or not the crack can propagate

toward the surface where its propagation is hindered by the residual stress field. Both situations

are illustrated by the Haigh diagrams in Fig. 59.

In the case of shot peened smooth specimens,

the fatigue strength falls below the values expected by the Haigh diagrams, because crack initiation occurs below the surface, where no ben-

eficial effect of the compressive residual stresses

is effective. Contrarily, the combination of surface loading stress and surface residual stress occurring at the notch root of specimens with kt

1.7 (Fig. 59a) falls significantly above the corresponding Goodman line. As already men-

Keff, MPa m

(Eq 19)

Without residual stresses

4

Range of Kth,eff

0

0

0.05

0.1

0.15

0.2

0.25

0.3

Range of effective stress intensity factor DKeff

with and without consideration of the residual

stresses versus distance from surface for ground notched

specimens made from a quenched SAE 1045 steel at cyclic

bending loading (kt 1.7, g 2)

Fig. 66

(Eq 20)

for smooth (kt 1) and notched (kt 1) specimens without residual stresses. rn,a is the nominal stress amplitude. If residual stresses are

present, the maximum stress is increased or decreased, depending on the sign and the amount

of the residual stresses, leading to:

DKeff (kt rn,a

rrs) p aY Kmax

(Eq 21)

if Kmin 0 or:

DKeff kt Drn p aY

60

Without residual stresses

Keff, MPa m

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

12

root, but are arrested below the surface.

A crack is arrested when the driving force for

its propagation falls below the threshold value:

Initial residual

stresses

40

Relaxed

residual

stresses

20

Range of Kth,eff

0

0

0.1

0.2

0.3

0.4

0.5

0.6

Range of effective stress intensity factor DKeff

without and with consideration of the initial

and the relaxed, respectively, residual stresses versus distance from surface for shot peened (shot of a hardness of

54 to 58 HRC) notched specimens made from a quenched

SAE 1045 steel at cyclic bending loading (kt 1.7, g

2)

Fig. 67

(Eq 22)

if Kmin 0.

Application to a High-Strength Steel. In a

high-strength SAE 1045 steel, literature data for

the threshold value Kth,eff fall in the range from

2.2 to 3.2 MPa m (Ref 67). As a first example,

ground notched specimens (kt 1.7, g 2)

with tensile residual stresses (see Fig. 43, curve

with open circles) are treated. Concerning the

half-elliptical cracks that initiate at the notch

root, a geometry factor Y 1 is assumed. The

interaction of this residual stress state with the

loading stress amplitude ra 230 MPawhich

corresponds to the fatigue strengthresults in

the depth distribution of DKeff shown in Fig. 66.

In a depth of 5 lm only, DKeff exceeds the

threshold value. Therefore, a crack that initiates

at the notch root will always propagate into the

interior of the specimen, and the fatigue limit

40

30

Relaxed residual

stresses

Keff, MPa m

Keff, MPa m

15

Initial

residual

stresses

Range of Kth,eff

20

10

Initial residual

stresses

Relaxed

residual

stresses

10

Range of Kth,eff

Crack initiation

0

0

0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.2

0.4

0.6

0.8

1.0

1.2

Range of effective stress intensity factor DKeff

without and with consideration of the initial

and the relaxed, respectively, residual stresses versus distance from surface for shot peened (shot of a hardness of

4650 HRC) notched specimens made from a quenched

SAE 1045 steel at cyclic bending loading (kt 1.7, g 2)

Fig. 68

DKeff for subsurface crack initiation without and

with consideration of the initial and the relaxed, respectively, residual stresses versus distance from surface for

ground and additional shot peened smooth specimens

made from a quenched SAE 1045 steel at cyclic bending

loading (g 0.4)

Fig. 69

that does not initiate a crack. This is in full agreement with the Haigh diagram in Fig. 59(a). It is

interesting to note that DKeff has a local maximum in a depth of 80 lm. Up to a depth of 150

lm, the residual stress field increases significantly DKeff as compared with the loading of a

specimen without residual stresses (dotted line).

A completely different situation prevails regarding shot peened notched specimens (kt

1.7, g 2) with the initial and the relaxed residual stress distributions shown in Fig. 57(b).

As shown in paragraphs on high-strength steel

in the sections Haigh Diagram and Concept

of Local Fatigue Strength, a crack will initiate

at the notch root if a specimen is loaded in the

range of the fatigue strength. Hence, if the crack

has the length a, the loading stress and the residual stress in the distance a from the surface determines if the crack will continue propagation

or not. In Fig. 67, DKeff is plotted versus the

distance from the surface. If the initial residual

stress distribution is concerned, DKeff becomes

zero at a depth of 40 lm, and higher stress amplitudes than the measured bending fatigue

strength would be necessary to propagate a crack

into the interior of the specimen. However, if the

relaxed residual stress distribution is concerned,

DKeff approaches the threshold range between 50

and 200 lm below the surface, and cracks may

or may not propagate. Hence, the fatigue limit

corresponds to the boundary between propagating or nonpropagating cracks that initiate at the

notch root. It is interesting to note that at depths

greater than 0.3 mm, DKeff becomes much larger

than in Fig. 66 as a consequence of the high loading amplitude. This means that the strong increase of finite fatigue life by shot peening is

entirely based on very small crack propagation

rates at small distances from surface.

Regarding shot peened notched specimens

with the residual stress distributions shown in

Fig. 57(a), almost no residual stress relaxation

occurs during loading in the range of the fatigue

limit. Therefore, the depth distributions of DKeff

corresponding to the initial and the relaxed residual stress distributions shown in Fig. 68 are

almost the same. At a depth of 40 lm, DKeff

becomes zero and exceeds the threshold value

only at a depth of 170 lm. Therefore, one would

expect that cracks could still be arrested at stress

amplitudes higher than the measured bending fatigue strength. This shows the limitations of the

simple estimation given here.

Figure 64(c) shows that crack initiation in shot

peened smooth specimens with the residual

stress distribution plotted in Fig. 55 occurs below the surface at a depth of 500 lm. For the

following rough estimation, it is assumed that

the crack has a circular shape (see Fig. 24), that

it propagates with the same rate toward the surface as well as into the interior of the specimen,

and that the residual stresses in depths greater

than 500 lm are zero. If the loading stress in the

depth of the crack front and a geometry factor Y

0.63 (Ref 68) are used for the calculation of

DKeff, one finds the relationship shown in Fig.

8

With residual

stresses

Keff, MPa m

2

Range of Kth,eff

0

0

0.05

0.1

0.15

0.2

0.25

0.3

Range of the effective stress intensity factor

DKeff with and without consideration of the residual stresses versus distance from surface for ground

notched specimens made from a quenched-and-tempered

(600 C/2 h) SAE 1045 steel at cyclic bending loading (kt

1.7, g 2)

Fig. 70

Keff, MPa m

20

Without residual

stresses

15

Initial residual

stresses

Relaxed

residual

stresses

10

Range of Kth,eff

5

0

0

0.1

0.2

0.3

0.4

Range of the effective stress intensity factor

DKeff without and with consideration of the initial and the relaxed, respectively, residual stresses versus

distance from surface for ground and additional shot

peened notched specimens made from a quenched-andtempered (600 C/2 h) SAE 1045 steel at cyclic bending

loading (kt 1.7, g 2)

Fig. 71

Relaxed residual

stresses

Without residual

stresses

Keff, MPa m

16

12

8

Initial residual

stresses

4

Range of Kth,eff

0

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

Fig. 72 Rangewithout

and with consideration of the iniDKeff

tial and the relaxed, respectively, residual stresses versus

distance from surface for ground and additional shot

peened smooth specimens made from a quenched-andtempered (600 C/2 h) SAE 1045 steel at cyclic bending

loading (g 0.4)

finally crack arrest. However, in this stage of

crack propagation the crack extension will be

large compared to the depth of the compressive

residual stresses. Therefore, loading stress redistribution will occur, as will relaxation of the

compressive residual stress field in the process

zone ahead of the crack front. The crack will

continue to propagate toward the surface causing

final failure. Then, the fatigue limit is correlated

with the maximum stress amplitude that does not

cause crack initiation below the surface. Therefore, the improvement of the fatigue strength of

smooth specimens by shot peening is limited, as

shown in Fig. 45(a). However, Fig. 69 illustrates

that the overall value of DKeff and, hence, the

driving force for crack propagation is rather

small (compare, for example, Fig. 69 with Fig.

66 in view of the very different loading stress

amplitudes). Therefore, already in the transition

from infinite to finite fatigue lifewhere crack

initiation still occurs below the surface (see Fig.

63)there is a strong influence of shot peening

on fatigue life, as shown also in Fig. 45(a). The

influence of shot peening on finite fatigue life

increases with increasing stress amplitude, because the crack initiation site is shifted toward

the surface, as discussed in the paragraphs on

high-strength steel in the section Concept of

Local Fatigue Strength.

Application to a Medium-Strength Steel. In

a medium-strength SAE 1045 steel, the threshold

value Kth,eff ranges from 2.3 to 3.3 MPa m (Ref

67). In the following, ground notched specimens

(kt 1.7, g 2) with tensile residual stresses

(see Fig. 35, curve with open circles), are treated.

The interaction of this initial residual stress state

with the loading stress amplitude ra 143

MPawhich corresponds to the fatigue

strengthresults in the depth distribution of

DKeff shown in Fig. 70. At a depth of 15 lm,

DKeff exceeds the range of threshold values.

Therefore, a crack that initiates at the notch root

will always propagate into the interior of the

specimen, and the fatigue limit corresponds to

the maximum stress amplitude that does not ini-

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

1200

R m = 1650 MPa

Final failure

1000

Transition range

1400 MPa

No final

failure

1150 MPa

800

600

400

0

0.2

0.4

0.6

0.8

1.0

1.2

Influence of the nominal bending stress amplitude on the length of arrested cracks in deeprolled notched specimens (kt 2) made from SAE 5135

steel with the ultimate tensile strengths indicated. After Ref

66

Fig. 73

(see Fig. 66), DKeff has a local maximum in a

depth of 70 lm, and up to a depth of 150 lm the

residual stress field increases significantly DKeff

as compared with the loading of a specimen

without residual stresses (dotted line).

The corresponding relationships for shot

peened notched specimens with the residual

stress distributions shown in Fig. 53 are given in

Fig. 71. The loading stress amplitude ra 310

MPa which corresponds to the bending fatigue

strength, is considered. From Fig. 60 it becomes

clear that crack initiation at the notch root occurs

under such loading conditions because the corresponding data point (open square) falls significantly above the Goodman relationship. On the

basis of the initial residual stress distribution,

there is no driving force for the propagation of

such a crack up to a distance of 170 lm from

the notch root. However, if the distribution of

the relaxed residual stresses is considered, DKeff

comes close to the threshold in depths ranging

from 30 to 200 lm. Therefore, the fatigue limit

corresponds to a loading condition where cracks

that initiate at the notch root may propagate or

not.

In the case of smooth shot peened specimens

with the residual stress distributions shown in

Fig. 52, it is difficult to determine on the basis

of the concept of the local fatigue strength

whether crack initiation occurs below or at the

surface, if the loading approaches the fatigue

limit (see paragraph on medium-strength steel in

the section Concept of Local Fatigue Strength

and Fig. 65b). Actually, crack initiation at the

surface is observed that is favored by the rather

high roughness. Figure 72 gives the depth distribution of DKeff for the loading stress amplitude

ra 414 MPa, which corresponds to the bending fatigue limit. Regarding the initial residual

stress state, the longest nonpropagating crack

would have a length of 250 lm. Due to stress

relaxation, this value is reduced by 50%. Still,

rather long cracks have to be formed during the

initiation stage to be able to propagate into the

interior of the specimen.

The difference between the minimum loading

stress amplitude for crack initiation at the root

of notched specimen and the maximum loading

stress amplitude at which crack arrest is observed may be very large in deep-rolled specimens, which have rather small or even tensile

residual stresses at the notch root, but large compressive residual stresses down to a relatively

great depth below the surface. Then, the length

of an arrested crack will strongly depend not

only on the depth distribution of the residual

stresses, but also on the loading stress amplitude.

This is proved by Fig. 73 concerning notched

specimens (kt 2) made from differently heat

treated SAE 5135 steel with the ultimate tensile

strength indicated (Ref 66). The diagram gives

the length of arrested cracks as a function of the

nominal stress amplitude. The higher the

strength of the material state, the less the crack

length increases with increasing stress amplitude

because the amount and the stability of the com-

pressive residual stresses produced by deep rolling increase with increasing hardness. From this

figure it becomes clear that the large increase of

the notch fatigue strength visible in Fig. 41 is

almost entirely based on an increase of the resistance against crack propagation produced by

deep rolling.

In view of the relationships presented so far,

it is interesting to look again at Fig. 39 and 40.

The lowest bending fatigue limit was determined

with ground specimens having low tensile residual stresses at the surface (curve 1 in both figures). An increase of 324 MPa (curve 6 in Fig.

39) was achieved with specimens that were shot

peened according to curve 4 in Fig. 40 and from

which a surface layer of 100 lm thickness was

the maximum compressive residual stresses into

the surface. Taking the slope of the straight line

for smooth specimens, g 0.4, in the range of

(initial) tensile residual stresses in Fig. 36 as a

rough approximation, one would expect a difference of both specimen series of 316 MPa

caused by the different residual stresses. On the

other hand, on the basis of the Goodman line in

Fig. 60, one would expect a much higher difference of 429 MPa. This shows that stress relaxation occurs in this medium-strength steel state,

as expected. Anyway, the bending fatigue limit

of both specimen series is determined by crack

initiation at the surface. The specimens corresponding to curve 5 in both figures have a lower

30

Without residual stresses

n,a = 640 MPa

n,a = 600 MPa

Keff, MPa m

Initial residual

stresses

20

Condition 6

n,a = 640 MPa

Condition 5

n,a = 600 MPa

Assumed

relaxed

residual

stresses

10

Range of Kth,eff

0

0

0.1

0.2

0.3

0.4

Distance from surface, mm

0.5

0.6

Range of the effective stress intensity factor DKeff without and with consideration of the initial and the relaxed,

respectively, residual stresses versus distance from surface for (corresponding to Fig. 39) shot peened smooth

specimens made from a blank-hardened AISI 5115 steel at cyclic bending loading

Fig. 74

rs

ls

Smooth

Fatigue strength, Rf

kt

Rh

the specimens corresponding to curve 6. Applying the same procedure with the aid of Fig. 36

as above to the surface residual stress state, an

increase of the bending fatigue limit of 220 MPa

compared to ground specimens is expected. Actually, a difference of 279 MPa is determined.

This finding means that the bending fatigue limit

is not determined by crack initiation at the surface, but by the largest amplitude at which crack

arrest below the surface occurs. These relationships become more clear in Fig. 74. There, DKeff

for surface cracks developing in conditions 5 and

6 at nominal stress amplitudes corresponding to

the bending fatigue strength are plotted as a

function of the distance from the surface. Regarding the initial residual stress distribution,

cracks smaller than 320 lm (condition 5) or 100

lm (condition 6) would not be able to propagate.

Actually, stress relaxation occurs, the extent of

which, however, is not given in Ref 54 to 56. If

it is assumed that the residual stresses relax by

50%, one gets the curves for the relaxed residual

stress state in the figure. Now, one can see that

a crack once initiated in condition 6 will continue propagation. Hence, the bending fatigue

strength of this condition is determined by the

resistance against crack initiation, as stated

above. Contrarily, in condition 5 the bending fatigue strength corresponds to the maximum

nominal stress amplitude at which crack arrest

below the surface occurs.

From these considerations, it becomes clear

that the depth distribution of relaxed residual

stresses must be taken into account to gain a reliable assessment of the influence of residual

stresses on the fatigue behavior of mediumstrength steel. Frequently, these data are not

known. Then, the use of initial tensile residual

stresses for the dimensioning of components is

a conservative procedure, because their detrimental effect on the fatigue behavior of a component will be reduced by residual stress relaxation. However, the beneficial effect of

compressive residual stresses will be overestimated by using the initial values, and this results

in a nonconservative dimensioning. Therefore, it

is important to estimate the remaining residual

stresses conservatively, for example with the article Stability of Residual Stresses in this

Handbook.

rs

micro

rs

ls

kt

Rh

Notched

Smooth

rs

micro

ls

Notched

Smooth

kt

Rh

Notched

drawn from this article are illustrated by Fig. 75,

in which the fatigue strength Rf is plotted as a

function of the macro residual stress. The following parameters are regarded:

The material strength; a low-strength steel

Influence of the macro residual stress and some other parameters on the fatigue strength of smooth and notched

specimens made from a low-strength steel (lower band), a medium-strength steel (middle band), and a highstrength steel (upper band)

Fig. 75

in the figure), a medium-strength steel (for

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

figure), and a high-strength steel (a quenched

or quenched-and-tempered (at a low temperature) steel, upper band in the figure) are concerned

The depth distribution of the macro residual

stress rrs, characterized by its sign, magnitude, and gradient grs

The depth distribution of the micro residual

stress rrs

micro

The notch factor kt (smooth and mildly

notched specimens are regarded in the figure)

The gradient of the loading stress gls

The surface topography, characterized by the

roughness height Rh

Low-Strength Steel

In a low-strength steel, there will be no or very

little influence of the macro residual stress, because it is relaxed more or less completely if the

cyclic loading approaches the fatigue strength

(see Fig. 32 and 48). A change of the micro residual stress state by work hardening may significantly increase Rf since the resistance against

cyclic plastic deformation and hence, crack initiation increases (see Fig. 31). Then, also the resistance against macro residual stress relaxation

is raised resulting in a certain sensitivity of the

work-hardened zone to macro residual stress.

This may be detrimental or beneficial (see Fig.

33b) for Rf depending on the sign and the magnitude of the macro residual stress. The influence

of the surface topography is rather small in a

low-strength steel (see Fig. 47). The fatigue

notch factor kf is significantly smaller than the

notch factor kt because cyclic plastic deformation and stress redistribution occur in the notch

root (see Fig. 32). With increasing loading stress

gradient at a given kt, the fatigue strength increases due to the decrease of the highly stressed

volume of the component or specimen (compare

kt 2.5, g 2 with kt 2.5, g 5 in Fig.

32).

Medium-Strength Steel

In a medium-strength steel, there is a significant influence of the macro residual stress on Rf

since only a small part of rrs relaxes during cyclic loading in the range of the fatigue limit (see

Fig. 36, 38, 51a, and 52). However, in the lowcycle fatigue range, relaxation becomes more

complete with increasing amplitude, and the influence of the macro residual stress vanishes (see

Fig. 34, 37, 39, 51b). Tensile residual stresses

are always detrimental to Rf. Therefore, in the

presence of large tensile residual stresses a medium-strength steel may have equal or even

lower fatigue strength than a low-strength steel

(compare Fig. 32 with 36). If relaxed tensile residual stresses are concerned, the residual stress

sensitivity m of the fatigue strength approaches

the mean stress sensitivity M (see Fig. 60). In

material states of similar hardness loaded in the

range of the fatigue limit, compressive residual

stresses relax stronger than tensile ones simply

because of the different corresponding stress am-

may be changed significantly by the processes

that generate the macro residual stress. Hence,

in a medium-strength steel the local resistance to

residual stress relaxation may be quite different

even if the same initial material state is concerned, and unique borderlines for the onset of

residual stress relaxation do not exist (see Fig.

60). A significant benefit from compressive residual stresses is only obtained if their penetration depth is sufficiently high and/or their gradient is sufficiently low as compared to the

gradient of the loading stresses. Therefore, compressive residual stresses produced by grinding

do hardly influence the fatigue strength of

smooth and notched specimens (see Fig. 36).

Contrarily, compressive residual stresses generated by deep rolling or shot peening increase the

notch fatigue strength (see Fig. 38, 39, and 41).

The fatigue strength of smooth specimens is less

improved, if subsurface crack initiation occurs.

Then, the fatigue strength of notched specimens

may even be higher than corresponding values

of smooth specimens (see Fig. 41). This means

that the fatigue notch factor kf becomes less than

unity. At vanishing residual stresses, kf comes

rather close to kt if the loading stress gradient is

small, but is significantly less at higher gls values. With increasing tensile residual stresses, the

fatigue notch factor is reduced again, as expected

on basis of the Goodman relationship for smooth

and notched specimens (see Fig. 58 and 60).

There is a considerable influence of the surface

roughness on the fatigue strength, as shown in

Fig. 47.

High-Strength Steel

In high-strength steels, stress relaxation during cyclic loading in the range of the fatigue

strength only occurs in notched specimens bearing very high compressive residual stresses.

Then, the resulting fatigue strength is also high,

and during corresponding cyclic loading very

high magnitudes of the minimum stress occur,

which leads to some residual stress relaxation

(see Fig. 57b). Contrarily, in the range of high

tensile residual stresses and cyclic loadings that

lead to infinite life or to technically relevant lifetimes the occurring maximum stresses are much

lower and no residual stress relaxation is observed even in the range of low-cycle fatigue.

Consequently, the residual stress sensitivity and

the mean stress sensitivity of Rf are identical (see

Fig. 59), and the fatigue strength is strongly reduced with increasing tensile residual stress.

This is also true for the finite fatigue life (see

Fig. 42). In the range of compressive residual

stresses, complex relationships exist. A strong

effect of rrs will only occur if cracks are initiated

at the surface. However, in thick smooth specimens or components cyclically loaded in the

range of the fatigue strength, the loading stress

gradient is usually lower than the gradient of the

local fatigue strength, which depends on the

depth distribution of the residual stresses. Consequently, cracks are initiated below the surface

(see Fig. 62a and c, 63, and 64c) and the improvement of the fatigue strength even by high

compressive residual stresses is limited (see Fig.

45a, 46, and 56). With increasing stress amplitude and hence loading stress gradient, the crack

initiation site is shifted to the surface. Therefore,

the finite fatigue life is much more improved

than the fatigue strength (see Fig. 62a and c, 63,

and 45d). In notched specimens, the loading

stress gradient is larger than in smooth ones. If

high compressive residual stresses with a sufficient depth and/or a small gradient exist, crack

initiation will occur in the notch root resulting

in a strong improvement of the fatigue strength

(see Fig. 64a and b, 45b, 46, and 56). However,

if the penetration depth of the compressive residual stresses is low and therefore the gradient

of the local fatigue strength highfor example,

after grindingthe improvement of the notch

fatigue strength will be small (see Fig. 44 and

59). Frequently, after shot peening or deep rolling, maximum compressive residual stresses occur below the notch root. Then, the fatigue

strength may not be determined by the maximum

cyclic loading, which does not result in crack

initiation, but by the maximum cyclic loading at

which crack arrest below the surface is possible

(see Fig. 67, 68, and 73). As a consequence of

all these relationships, the fatigue notch factor

may vary strongly in the presence of compressive residual stresses regarding one-notch geometry and may take values ranging from less

than unity to the notch factor kt, as sketched in

Fig. 75. In fact, regarding smooth and notched

specimens, Fig. 44 and 54 prove that kf of ground

specimens with compressive residual stresses

produced by grinding approaches kt. From Fig.

56 it can be deduced that kf of shot peened specimens approaches unity. At vanishing residual

stresses, kf comes close to kt as expected in a

high-strength steel. With increasing tensile residual stress, the fatigue notch factor is reduced

and finally approaches unity (see Fig. 44 and 54).

Again, this finding is in correspondence with the

Goodman relationship for smooth and notched

specimens (see Fig. 58 and 59). The surface

roughness has principally a large influence on

the fatigue strength of high-strength steel, as

shown in Fig. 47. On the other hand, in practice

the roughness height of hard steel is rather low

even after mechanical surface treatments such as

shot peening.

Recommendations

From all of these relationships, some recommendations may be deduced. In medium- and

high-strength steels tensile macro residual

stresses must strictly be avoided since they always promote crack initiation and crack propagation and are detrimental to the fatigue strength

andat least in higher strength steelto finite

fatigue life. In a low-strength steel, the influence

of tensile macro residual stresses is usually small

or negligible, and the change of the micro residual stress state by work hardening is much more

important. In most cases, work hardening will be

beneficial for fatigue strength andif the micro

residual stress state is sufficiently stablefor finite fatigue life. However, processes must be

avoided in which strong work hardening is combined with the generation of tensile residual

stresses. Then, residual stress relaxation during

cyclic loading will be incomplete, and the detrimental influence of tensile residual stresses on

the fatigue strength will appear.

The beneficial influence of compressive

macro residual stresses ranges from almost nil to

the full extent expected after the mean stress sensitivity of the material state under consideration

and depends on the wholeness of the parameter

considered in Fig. 75. In a low-strength steel, it

is recommended to use processes in which the

generation of compressive residual stresses is

combined with strong work hardening. Then,

there will be on one hand the beneficial influence

of work hardening on the reduction of cyclic

plasticity and the increase of the materials resistance against crack initiation. On the other hand,

the resistance against residual stress relaxation

will also increase, resulting in some beneficial

influence of the compressive macro residual

stresses on crack initiation and crack propagation. One example is deep rolling (see Fig. 33).

Another example would be shot peening. However, shot peening of a low-strength steel will

produce a rather rough surface, and this will

eventually counterbalance a large part of the

beneficial influence on the fatigue behavior.

In medium-strength steels it is also recommended to utilize processes that produce both

work hardening and compressive macro residual

stresses. Again, deep rolling (see Fig. 41) or shot

peening (see Fig. 37 and 39) can serve as examples. Now, for a significant effect it is necessary to adjust the penetration depth of the

macro residual stresses and the width of the

work-hardened zone to the depth distribution of

the loading stress. Clearly, this is impossible in

push-pull loading of smooth specimens or components (gls 0) and very difficult in any cyclic

loading of thick smooth specimens or components, where gls is rather small. Then, some improvement of the fatigue behavior will be obtained because crack initiation at the surface is

retarded or crack initiation is shifted below the

surface (see Fig. 37a and Fig. 41, smooth specimens). However, in notched or rather thin

smooth specimens or components with higher gls

it is possible and recommended to produce a

depth distribution of the macro and micro residual stresses and, hence, of the local fatigue

strength that ensures that crack initiation can

only occur at the surface. Then, a strong improvement of the fatigue strength and finite fatigue life is possible by one of two mechanisms.

On one hand, the resistance of the steel against

crack initiation and growth of very short cracks

may strongly be increased by a smooth surface,

strong work hardening, and maximum compressive residual stresses at the surface. Examples

are given in Fig. 37(b) (shot peened notched

specimens) and Fig. 39 (condition 6, shot peened

smooth specimens). In these cases, the fatigue

stress amplitude that does not initiate a crack at

the notch root. On the other hand, by high compressive stresses below the surface, the resistance against propagation of cracks with lengths

up to the distance of the residual stress maximum

from the surface may strongly be increased.

Again, strong work hardening, is essential to prevent the compressive residual stresses as much

as possible from relaxation. Examples are given

in Fig. 39 and 40 (condition 5, shot peened

smooth specimens) and Fig. 41 and 73 (deeprolled notched specimens). In these cases, the

fatigue strength is determined by the maximum

loading stress amplitude at which crack arrest

below the surface is obtained.

In high-strength steels, residual stress relaxation is negligible except for extreme conditions

that were discussed above (see Fig. 57b). Apart

from this item, principally the same relationships

as just discussed are valid. Again, it is difficult

to achieve a large improvement of Rf in thick

smooth specimens or components, since crack

initiation will occur below the surface (see Fig.

56, 63, and 64c). However, a significant improvement of finite fatigue life in the low-cycle

fatigue range may be obtained, since crack propagation toward the surface is hindered by the

high compressive residual stresses (see Fig. 69)

and/or the crack initiation site is shifted to the

surface as a result of the increasing loading stress

gradient (see Fig. 63). In thin smooth or in

notched specimens and components it is recommended to produce high compressive residual

stresses with a large penetration depth by processes such as shot peening, deep rolling, and

others described in this Handbook. Then, crack

initiation and early crack growth is forced to occur at the surface, and enormous increases of the

fatigue strength are obtained by one of the two

mechanisms described previously. However, by

application of the aforementioned processes to a

high-strength steel, frequently rather small surface compressive residual stresses are obtained

and maximum values of |rrs| occur below the

surface. Then, the second mechanism (crack arrest below the surface) is solely effective (see

Fig. 67 and 68), and the fatigue strength may be

much higher than expected by the surface residual stress and loading state plotted in a Haigh

diagram (see Fig. 59a).

Regarding surface roughness, numerous investigations exist. The interaction of an increased roughness with tensile residual stresses

will always be detrimental to the fatigue strength

and to finite fatigue life, especially in highstrength steels. In the presence of compressive

residual stresses, different influences exist. If the

fatigue strength is entirely and the finite fatigue

life mainly determined by crack initiation at the

surface, there will be decisive influence of the

surface roughness. A much smaller influence of

surface roughness, however, may exist if the fatigue strength is determined by crack initiation

or crack arrest below the surface or if the finite

fatigue life is mainly influenced by crack propagation below the surface. If compressive resid-

components, it is of utmost importance to account for their stability. The reader is referred to

the article Stability of Residual Stresses in this

Handbook.

REFERENCES

1. B. Scholtes, Residual Stresses Generated by

Mechanical Surface Treatments, DGM Informationsgesellschaft, Oberursel, 1990, in

German

2. V. Hauk, Ed., Structural and Residual

Stress Analysis by Nondestructive Methods:

Evaluation, Application, Assessment, Elsevier Science B.V., Amsterdam, 1997

3. H. Ishigami, M. Matsui, Y. Jin, and K.

Ando, Proc. ICRS-6, IOM Communications

Ltd., London, 2000, p 667673

4. Fatigue and Fracture, Vol 19, ASM Handbook, ASM International, 1996

5. E. Macherauch, Laboratory Course on Material Science and Engineering, ViewegVerlag Braunschweig, Germany, 1992, in

German

6. D. Eifler, Dr.-Ing. Thesis, University Karlsruhe (TH), 1981

7. D. Eifler and E. Macherauch, Defects, Fracture and Fatigue, G.C. Sih and J.W. Prowan, Ed., Martinus Nijhoff, The Hague,

1983, p 171182

8. D. Eifler and E. Macherauch, Int. J. Fatigue,

Vol 12 (No. 3), 1990, p 165174

9. M. Becker, Dr.-Ing. Thesis, University

Karlsruhe (TH), 1987

10. M. Becker, D. Eifler, and E. Macherauch,

Mater.wiss. Werkst.tech., Vol 24, 1993, p

5764

11. P.C. Paris, M.P. Gomez, and W.E. Anderson, The Trend in Engineering, Vol 13,

1961, p 914

12. P.C. Paris and F. Erdogan, J. Basic Eng.

(Trans. ASME ), Vol 85, 1963, p 528534

13. NASA/FLAGRO: Fatigue Crack Growth

Computer Program, National Aeronautics

& Space Administration, p 5

14. R.C. McClung, K.S. Chan, S.J. Hudak, Jr.,

and D.L. Davidson, Behavior of Small Fatigue Cracks, Fatigue and Fracture, Vol 19,

ASM Handbook, ASM International, 1996,

p 153158

15. A. Wohler, Z. Bauwesen, Vol XX, 1870, p

73106

16. O.H. Basquin, Proc. ASTM, Vol 10, 1910,

p 625630

17. L.F. Coffin, Trans. ASME, Vol 76, 1954, p

931950

18. S.S. Manson, NACA Report 1170, Lewis

Flight Propulsion Laboratory, 1954

19. J.H. Smith, J. Iron Steel Inst., Vol 82, 1910,

p 246318

20. B.P. Haigh, No. 85, Report Brit. Association, 1915, p 163170

21. J. Goodman, Mechanics Applied to Engineering, Longmans-Green, London, 1899

22. H. Gerber, Z. Bayr. Architekt. Ing.-Ver., Vol

6, 1874, p 101110

23. K.N. Smith, P. Watson, and T.H. Topper, J.

Mater., Vol 5, 1970, p 767778

24. W.J. Ostergren, J. Test. Eval., Vol 4 (No.

5), 1976, p 327339

25. A. Ebenau, Dr.-Ing. Thesis, University

Karlsruhe (TH), 1989

26. A. Ebenau, D. Eifler, O. Voehringer, and E.

Macherauch, Proc. ICSP 4, K. Iida, Ed., The

Japan Society of Precision Engineering, Tokyo, 1990, p 327336

27. F. Hoffmann, Dr.-Ing. Thesis, University Bremen, 1987; Fortschritts-Berichte VDI Reihe 5

Nr. 123, VDI-Verlag, Dusseldorf, 1987

28. I. Altenberger, Dr.-Ing. Thesis, University

GH Kassel, 2000

29. I. Altenberger, B. Scholtes, U. Martin, and

H. Oettel, Mater. Sci. Eng. A, Vol 264,

1999, p 116

30. G. Kuhn, Dr.-Ing. Thesis, University Karlsruhe (TH), 1991

31. G. Kuhn, J.E. Hoffmann, D. Eifler, B.

Scholtes, and E. Macherauch, ICRS 3, H.

Fujiwara, T. Abe, and K. Tanaka, Ed., Elsevier Applied Science, London, 1991, p

12941301

32. A. Glaser, D. Eifler, and E. Macherauch,

Mater.wiss. Werkst.tech., Vol 22, 1991, p

266274

33. H. Mughrabi, Fatigue Behavior of Metallic

Materials, D. Munz, Ed., DGM-Informationsgesellschaft Verlag, Oberursel, 1985, p

738, in German

34. J.D. Almer, J.B. Cohen, and B. Moran, Mater. Sci. Eng. A, Vol 284, 2000, p 268279

35. H. Berns and L. Weber, Residual Stresses

in Science and Technology, E. Macherauch

and V. Hauk, Ed., DGM-Informationsgesellschaft Verlag, Oberursel, 1987, p 751

758

36. H. Berns and L. Weber, Shot Peening, ICSP

3, H. Wohlfahrt, R. Kopp, and O. Voehringer, Ed., DGM-Informationsgesellschaft

Verlag, Oberursel, 1987, p 647654

37. R. Herzog, Dr.-Ing. Thesis, University

Braunschweig, 1997

38. E. Macherauch and H. Wohlfahrt, Fatigue

Behavior of Metallic Materials, D. Munz,

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

Oberursel, 1985, p 237283, in German

P. Starker, E. Macherauch, and H. Wohlfahrt, Fatigue Eng. Mater. Struct., Vol 1,

1987, p 319327

S. Jaegg, Dr.-Ing. Thesis, University GH

Kassel, 1999

S. Jaegg and B. Scholtes, ICRS 5, Linkoping, Sweden, 1997, p 10781083

E. Welsch, D. Eifler, B. Scholtes, and E.

Macherauch, Proc. Sixth European Conf. on

Fracture (ECF 6), V. Elst and A. Bakker,

Ed., Eng. Mater. Adv. Services Ltd

(EMAS), Amsterdam 1986, p 13031320

A. Stacey and G.A. Webster, Materials Research Society Proc., Vol 22, 1984, p 215

H.P. Lieurade, Advances in Surface Treatments, A. Niku-Lari, Ed., Pergamon Press,

Oxford, 1987, p 455482

K.J. Kang, J.H. Song, and Y.Y. Earmme,

Fatigue Fract. Eng. Mater. Struct., Vol 12,

1989, p 363376

J.E. Hoffmann, D. Loehe, and E. Macherauch, Proc. ICSP3, H. Wohlfahrt, R. Kopp,

and O. Voehringer, Ed., Garmisch-Partenkirchen, DGM Verlag, Oberursel, Germany,

1987, p 631638

J.E. Hoffmann, D. Loehe, and E. Macherauch, Residual Stresses in Science and

Technology, Proc. ICRS 1, E. Macherauch

and V. Hauck, Ed., Vol 1, DGM Verlag, Oberursel, Germany, 1987, p 801808

J.E. Hoffmann, D. Eifler, and E. Macherauch, Residual Stresses: Origination, Measurement, Assessment, E. Macherauch and

V. Hauck, Ed., Vol 2, DGM Verlag, Oberursel, Germany, 1983, p 287300, in German

J.E. Hoffmann, Dr.-Ing. Thesis, University

Karlsruhe (TH), 1984

B. Syren, H. Wohlfahrt, and E. Macherauch,

Arch. Eisenhuttenwes., Vol 46, 1975, p

735739

B. Syren, H. Wohlfahrt, and E. Macherauch,

Proc. Second Int. Conf. Mech. Behavior of

Materials (ICM2) (Boston), 1976, American Society for Metals, p 807811

53. H. Traiser and K.-H. Kloos, Z. Werkstofftechn., Vol 16, 1985, p 135143

54. R. Schreiber, H. Wohlfahrt, and E. Macherauch, Arch. Eisenhuttenwes., Vol 49,

1978, p 207210

55. R. Schreiber, H. Wohlfahrt, and E. Macherauch, Arch. Eisenhuttenwes., Vol 48,

1977, p 653657

56. H. Wohlfahrt, Mechanical Surface Treatments, H. Wohlfahrt and P. Krull, Ed., Wiley-VCH, Weinheim, Germany, 2000, p

5584, in German

57. K.-H. Kloos, B. Fuchsbauer, and J. Adelmann, J. Fatigue, Vol 9, 1987, p 3542

58. P. Starker and E. Macherauch, Z. Werkstofftechn., Vol 14, 1983, p 109115

59. P. Starker, Dr.-Ing. Thesis, University

Karlsruhe (TH), 1981

60. A. Navarro and E.R. de los Rios, Philos.

Mag. A, Vol 57 (No. 1), 1988, p 1536

61. K.J. Miller and E.R. de los Rios, The Behavior of Short Fatigue Cracks, Mechanical

Engineering Publications Limited, London,

UK, 1986

62. B. Scholtes, Structural and Residual Stress

Analysis by Nondestructive Methods: Evaluation, Application, Assessment, V. Hauk,

Ed., Elsevier Science B.V., Amsterdam,

1997

63. W. Backfisch and E. Macherauch, Arch. Eisenhuttenwes., Vol 50, 1979, p 167171

64. K. Wellinger and D. Dietmann, Calculation

of Strength, A. Kroener Verlag, Stuttgart,

1969, in German

65. B. Winderlich, Mater.wiss. Werkst.tech.,

Vol 21, 1990, p 378389

66. K.H. Kloos, Notches and Structural Durability, H. Nowack, Ed., Deutscher Verband

fur Materialforschung und-pruefung e.V.,

1989, p 740, in German

67. J. E. Hoffmann and D. Loehe: to be published

68. T. Fett and D. Munz, Stress Intensity Factors and Weight Functions, Computational

Mechanics Publications, 1997

G. Totten, M. Howes, T. Inoue, editors, p54-69

DOI: 10.1361/hrsd2002p054

All rights reserved.

www.asminternational.org

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

structural components during manufacturing

processes such as forging, machining, heat treating, shot peening, and many others. These

stresses are always a consequence of inhomogeneously distributed dimensional changes due

to inhomogeneous plastic deformations, thermochemical treatments, and/or phase transformations. They can be either beneficial or detrimental to component behavior in service,

depending on the materials state as well as the

sign, magnitude, and stability of the residual

stresses, mechanical loading, and environmental

conditions (Ref 115). Compressive macro residual stresses in the surface region of materials

with medium and high hardnesses increase the

fatigue life and the fatigue limit at cyclic loading

compared to materials states that are free of residual stresses. This improvement is caused by

an increased resistance against crack initiation

and, to a certain extent, against crack propagation if the residual stresses are sufficiently stable

in the areas of highest loading (in most cases, the

component surface areas). Moreover, compressive macro residual stresses can increase the resistance of certain materials against corrosion fatigue and stress corrosion cracking. They can

also improve the wear resistance. Therefore, in

many cases compressive macro residual stresses

are intentionally generated in near-surface regions by controlled heat treatments or by posttreatments such as shot peening or deep rolling.

The stability or relaxation behavior of these

residual stresses at purely thermal or mechanical

loadings as well as superimposed thermal and

mechanical loadings is thus of decisive importance for the service behavior of components and

hence of great interest from a scientific as well

as a practical point of view (Ref 1618). If residual stresses are relaxed by annealing or mechanical treatment, they naturally have little if

any influence on subsequent component failure.

Residual stresses can be reduced or completely relaxed by the application of mechanical

and/or thermal energy. The elastic residual

strains, ee, that are associated with the residual

stresses via Hookes law can be converted into

micro plastic strains, ep, by suitable deformation

processes. For example, this transformation can

be achieved by dislocation slip, dislocation

creep, grain-boundary sliding, and/or diffusion

only a limited extent or not at all, relaxation of

residual stresses is also conceivable by crack formation and propagation. Relaxation of residual

stresses in the real case occurs by complex interaction of a large number of factors. It depends

not only on the residual stress state itself but also

on the material state, loading condition, geometry, and environment of the component under

consideration.

The best known and most important techniques for inducing residual stress relaxation are

annealing (tempering), uniaxial deformation

(drawing, stretching), and cyclic deformation

(Ref 1618). Relaxation can also be caused by

thermal cycling, quenching, neutron bombardment, the effect of alternating magnetic fields (in

the use of ferromagnetics), vibration, and partial

damage. As discussed in other articles in this

Handbook, residual stresses can be differentiated

by their technological origins (i.e., whether they

were produced during forming, machining, heat

treatment, joining, coating, or casting).

Before the effect of applied thermal or mechanical energy on the residual stress state can

be assessed, the latter must be determined quantitatively. In the case of mechanical methods, the

necessarily large number of measurements involve considerable expenditure of both time and

money. Most investigations have thus employed

x-ray methodsfor example, the sin2w method

(Ref 19) with CrK radiation on the {211}-interference plane of the ferrite, as well as

interference-profile analyses (e.g., Ref 20, 21).

In this article, current knowledge about residual stress relaxation of steels is presented in a

condensed and, as far as possible, systematic

form. The abundance of available experimental

data nevertheless necessitates a limited scope.

Thus, the findings of residual stress relaxation

by annealing, tensile loading, compressive loading, and cyclic loading will be taken as examples

to demonstrate universally valid rules. Shot

peening residual stresses are particularly suitable

for such investigations due to their sign and their

relatively large absolute value. Possible formulas

that can be used to quantify residual stress relaxation are discussed, and the underlying microstructural as well as micromechanical mechanisms are considered.

Annealing

If a pure metal is annealed for several hours

at a temperature of

Ta 0.5 Tm [K]

(Eq 1)

and then cooled slowly to room temperature, almost total relaxation of the residual stresses arising from forming, machining, heat treatment, or

joining operations can be achieved. The necessary annealing time depends essentially on the

workpiece dimensions and the material state.

Since an annealing temperature of 0.5 Tm corresponds to the recrystallization temperature of

ferritic steels, complete relaxation of macro residual stresses can be expected. In the case of

pearlitic steels, macro residual stresses are completely eliminated during the pearlite-austenite

transformation above the A1 temperature. Micro

residual stresses are considerably reduced but

not entirely removed, since lattice defects (particularly dislocations) and, in the case of heterogeneous materials, the different expansion coefficients of the various phases are always

responsible for some micro residual stresses. Residual stress relaxation by annealing is brought

about by so-called thermally activated processes

for which the annealing temperature and the annealing time are interchangeable within certain

limits. In order to achieve comparable residual

stress relaxation at a lower annealing temperature, the annealing time must be increased correspondingly (Ref 16, 17).

Thermal residual stress relaxation is fundamentally affected by the residual stress state itself and by the material state. This is convincingly demonstrated by the findings presented in

Fig. l and 2 (Ref 16, 17), which show the effect

of 1 h anneals on surface macro residual stresses

of a variety of origins in a variety of steels investigated using x-ray methods. Figure 1 is a bar

chart showing the macro residual stresses present

in the original state and after the normal industrial stress relieving for hardened components of

1 h at 200 C (390 F). The materials deformed

in tension have undergone negligible residual

stress relaxation. The greater the original compressive residual stresses, the greater the residual

stress relaxation produced by annealing of

surface-machined specimens: 11% for a plain

carbon steel with 0.45 wt% Carbon, SAE 1045

(Fig. 1C); 17% for SAE 1045 (Fig. 1D); and

19% for the rolling bearing steel AISI 52100

(Fig. 1E). The most pronounced relative residual

stress relaxation, however, is observed in the

hardened materials: 25% for plain carbon steel

with 0.22 wt% C, SAE 1023 (Fig. 1F); and 37%

for SAE 1045 (Fig. 1G).

Figure 2 shows the relative residual stresses

rrs(Ta)/rrs(293 K) for the same steels as a function of the homologous annealing temperature,

Ta /Tm. It can be seen that a 1 h anneal at 0.5 Tm

(about 600 C, or 1110 F) results in complete

relaxation of the macro residual stresses in every

case. Characteristic, material-specific stresstemperature curves are obtained. Clearly, residual stresses produced by hardening are relaxed

at lower temperatures, machining residual

stresses at medium temperatures, and deformation residual stresses at higher temperatures. In

order to achieve the same degree of residual

stress relaxation, (e.g., 50%) in 1 h in hardened

and in deformed SAE 1045, a temperature difference of 150 C (270 F) is necessary. Furthermore, according to Fig. 1, larger initial residual stresses result in a shift of the rrs versus

T curve to lower annealing temperatures in the

case of hardened or machined materials.

Micro residual stresses and their relaxation behavior can be determined directly from the halfwidth values (HW) of the x-ray interference

lines, which are a measure of the microstructural

work-hardening state as well as the dislocation

density. The change of the HW values are approximately proportional to the changes of the

micro residual stresses (DHW Drrs

micro).

Again, the micro residual stresses in steels are

more effectively relaxed at a higher temperature

and, like macro residual stresses (see Fig. 2),

lower temperatures, such as those produced by

machining; likewise, machining micro residual

stresses are relaxed at lower temperatures, such

as those produced by deformation in tension. In

comparable steel states, micro residual stresses

are relaxed only after a longer period or at a

higher temperature than macro residual stresses

(Ref 16, 17). For example, in order to achieve

50% micro residual stress relaxation in tensiledeformed SAE 1045 annealed for 1 h, the annealing temperature must be 100 C (180 F)

higher than that needed to achieve the same degree of macro residual stress relaxation.

Data pertaining to the thermal relaxation of

micro residual stresses for hardened steels are

available by analyzing x-ray interference lines of

different steels in (Ref 2224). The mean lattice

distortions e21 / 2, which are proportional to the

micro residual stresses according to

2 1/2

rrs

micro E110 e

(Eq 2)

direction, increase with growing carbon content

mainly due to the increasing dislocation density

and, secondarily, to the increasing number of

solute carbon atoms in octahedral sites of the

tetragonal martensite lattice. Usually, it can be

assumed that with increasing carbon content the

temperature for the onset of residual stress relaxation is lowered and the recovery rate increases.

However, it must be remembered that with the

annealing-induced formation of carbides on dislocations, which leads to their pinning, the relaxation procedure slows.

The relaxation behavior of residual stresses

depends for a similar surface treatment, as for

example by shot peening, as shown in Fig. 3 for

the plain carbon steel SAE 1045 with 0.45 wt%

C, from the heat treatment before shot peening

400

Ck22

F

Ck45

A

Ck22

B

Ck45

C

Ck45

D

Ck45

G

25). The macro residual stresses (left-hand side

of Fig. 3) and the micro residual stresses (righthand side; characterized by the half-width ratio

DHW/DHW0), yield with the hardness increasing

from the normalized to the quenched and tempered and to the hardened condition, greater relaxation rates. Figure 3 also illustrates the delay

of the relaxation of the micro residual stresses,

which was mentioned above in comparison with

the macro residual stresses. Further experimental

investigations of the macro and micro residual

stress relaxation of metallic materialstempering steel AISI 4140 (Ref 2628), titanium alloy

Ti-6Al-4V (Ref 29), and several CuZn alloys and

AlMg alloys (Ref 30, 31)lead to similar results.

Materials with surface residual stresses produced by grinding, milling, deep rolling, or shot

peening frequently show different residual stress

relaxation behavior when comparing the surface

layer with subsurface areas. In most cases, this

behavior is connected with work-hardening gradients, where an increasing work hardening due

to increasing dislocation densities causes growing mean strains and greater half-width values.

Characteristic examples of depth distributions

of the macro residual stresses and the half-width

values in SAE 1045, which were produced after

different heat treatments and electrolytical removal of thin material layers, are shown Fig. 4

and 5 (Ref 25). Normalized SAE 1045 samples

in the ground, milled, or shot-peened condition

result in typical depth distributions that depend

on the heat treatment (Fig. 4). In this case, the

relaxation behavior of the macro residual

stresses (left-hand side), and of the half-widths

(right-hand side), is more pronounced at the surface than in deeper layers. After selected annealing treatments, this behavior holds to completely relaxed shot peening residual stresses in

the near-surface layers in contrast to the interior,

where remaining residual stresses exist (see Fig.

4, bottom row, left, after annealing 30 min at 550

C, or 1020 F).

The relaxation behavior of normalized SAE

1045 in the shot-peened condition is reproduced

100Cr6

E

Ta, C

400

Normalized

+5% deformed

in tension

Normalized

+ ground

800

Untempered

1200

Fig. 1

1.0

rs(Ta)/rs(293 K)

rs, MPa

Hardened

Hardened +

shot peened

Macro residual stress (rrs) of different steels before and after annealing for 1 h at 200 C (390 F). Ck 45 SAE

1045; Ck 22 SAE 1023; 100 Cr 6 AISI 52100

100

200

300

400

C

B

0.6

0.2

0

0.15 0.20

600

ta = 1 h

0.8

0.4

500

E

F

G

Ta/Tm, K

Macro residual stress ratio versus homologous

annealing temperature of different steels (see

Fig. 1) after annealing for 1 h

Fig. 2

again in Fig. 5 (top row) in comparison with

shot-peened conditions of quenched and tempered as well as hardened states of SAE 1045.

The shot peening residual stresses in the hardened state (bottom row) are reduced by annealing of 1 h at 300 C (570 F) almost 50% at the

surface and in the depth of the maximum of compressive residual stresses. The corresponding

HW values in the area close to the surface are

smaller than those measured in deeper regions.

This finding is well known from relatively hard

materials, such as hardened steels (500 HV)

with relatively high dislocation densities, where

the shot peening process causes decreasing HW

values due to work-softening effects (Ref 32).

The higher the hardness, the more effective the

work softening. In these material states, decreases in HW values are caused by rearrangements of the dislocations with high density in

low-energy structures. However, in the

quenched and tempered condition, and in the

normalized condition, enlarged relaxation effects are observed in the surface layer due to a

dislocation-induced higher driving force and the

relaxation rate decreases with increasing distance from the surface. These findings must be

stress relaxation of corresponding components.

can be linearized with the help of a Zener-WertAvrami function (Ref 33) to:

m

rrs/rrs

0 exp [(Ata) ]

Processes

In order to analyze the thermally activated

processes responsible for residual stress relaxation, the influence of annealing time, ta, and of

annealing temperature, Ta, must be known. If the

time and temperature ranges are sufficiently restricted and if rrs(Ta,ta)/rrs(293 K) constant

(abbreviated in the following to rrs/rrs

0 ), the relationship

ta t0 exp (DH/kTa)

(Eq 3)

should hold between annealing time and temperature. In Eq 3, t0 is a time constant, k is the

Boltzmann constant, and DH is the activation enthalpy for the relaxation process in the ta versus

Ta range under consideration. Data in the form

of rrs versus 1og ta plots as in Fig. 3 are not

suited to a direct evaluation using Eq 3.

The effect of time on residual stress relaxation

dominant relaxation mechanism and A is a function dependent on material and temperature according to

A B exp (DH/kTa)

Ta, C

Ta, C

200

250

300

350

400

450

500

550

200

250

300

0.5

350

400

450

500

550

0.5

Normalized

0

1.0

150

150

1.0

250

250

0.5

350

350

450

1.0

0.5

HW/HW0

1.0

450

1.0

150

150

250

250

0.5

0.5

350

350

450

0

0

10

100

1000

450

Hardened

0

0

10

100

1000

Macro residual stress ratio versus annealing time, and ratio of the changes of the half-width values DHW/DHW0

versus annealing time, at different annealing temperatures at the surface of shot-peened SAE 1045 samples in

normalized, quenched and tempered, and hardened states

Fig. 3

(Eq 5)

rs

log ln (rrs

0 /r ) m log ta m log A

(Eq 6)

with the result that for a constant annealing temperature Ta, a plot of log ln(r0rs/rrs) versus log ta

gives a straight line (see Fig. 6).

Corresponding annealing temperatures and

times for rrs/r0rs constant can be derived from

such plots, as schematically illustrated in Fig. 6

in the log ta versus l/kTa plot. The values for

rrs/rrs0 constant lie on a straight line. Based

on Eq 5 and 6, this line results in

log ta constant

1.0

(Eq 4)

DH

ln 10 kTa

(Eq 7)

of the log ta versus l/kTa plot in Fig. 6 yields to

the activation enthalpy for residual stress relaxation.

An experimental verification of Eq 6 is presented in Fig. 7, with results for shot-peened turbine blades made from a 12% Cr steel that was

annealed for residual stress relaxation in the temperature range from 150 to 600 C (300 to 1110

F) up to 104 h (Ref 34, 35). The measuring data

are fitted very well by straight lines with a slope

of approximately m 0.10. The evaluation of

these lines by Eq 7 leads to the activation enthalpy DH 3.5 eV. This finding will be discussed together with the following results.

A complete quantitative description of the

thermal relaxation of residual stresses is of practical importance. This was achieved by an extensive study on the steel AISI 4140, which was

quenched, tempered 2 h at 450 C (840 F), and

then shot-peened. The shot peening was carried

out with cast steel shot S170 (44 to 48 HRC), at

a pressure of 1.6 bar and a coverage of 98% (Ref

26, 27). The absolute values of the macro residual stresses are plotted in Fig. 8(a) versus the

logarithm of the annealing time for different annealing temperatures. They decrease with increasing time and temperature. Because the conventional way for determining the Avrami

quantities DH, m, and B in Eq 4 and 5 requires

strong extrapolations to very high and low annealing times, a new iterative method was used

that allows determination of these parameters by

a nonlinear minimization of the residual sum of

squares (Ref 26, 27). To achieve this, all measurements in the evaluation are taken into account with the same weight. The algorithm

yields DH 3.3 eV, m 0.122, and B 1.22

1021 min 1 for the surface values of the macro

residual stresses. The curves in Fig. 8(a) were

calculated using these constants in Eq 4 and 5.

They describe the time and temperature dependence of the relaxation process very well.

The alterations of surface half-widths by annealing at temperatures between 250 and 450 C

(480 and 840 F) for different annealing times

are shown in Fig. 8(b) for the same steel. The

reductions of the HW values are similar to the

relaxation of the macro residual stresses presented in Fig. 8(a). Again, the Avrami approach

is used to describe the relaxation of half-widths,

applying the new iterative method to the differences between the half-widths after annealing

and the value HW 1.65 of a normalized specimen related to their starting values instead of

the ratio rrs(Ta,ta)/rrs

0 in Eq 4. The curves in Fig.

8(b) were calculated by means of the material

properties DH 2.48 eV, m 0.116, and B

1.09 1013 min 1 and agree well with the measured values. The dependence of the relaxation

of mean strains e21 / 2 and micro residual

stresses rrs

micro, respectively, on the annealing

time was measured in the temperature range between 250 and 450 C (480 and 840 F). The

data can also be modeled by the Avrami approach using e2(Ta, ta)1/2/e201/2 instead of

used to determine the material parameters DH

2.64 eV, m 0.096, and B 5.32 1012

min1, which allow description of the measured

values as shown in Fig. 8(c).

Comparison of these relaxation data in Ref 26

and 27 shows that the relaxation of macro residual stresses at the surface of the shot-peened state

is a bit faster than the relaxation of half-widths

and mean strains; this corresponds with the fact

that the constant B differs by nearly nine orders

of magnitude (see Table 1). The reason for the

differences in the relaxation rate is that for the

relaxation of macro residual stresses, dislocation

movement is sufficient. For a distinct relaxation

of micro residual stresses, however, additional

dislocation annihilation is necessary. The exponents m show no significant alterations, and the

activation enthalpies approach the values of the

activation enthalpy of self-diffusion of iron: DHS

2.8 eV. Additional data for the same steel in

a normalized and another quenched and tempered state in Table 1 (with exception of the

hardened state) confirm this statement (Ref 28).

Accordingly, volume diffusion-controlled dislocation creep in the residual stress field that is

dominated by climbing of edge dislocations

should be the rate-controlling process for the relaxation of the shot peening residual stresses of

3.0

400

Ground

350 C/22 min

2.5

200

2.0

0

0

2.4

100

2.2

HW, 2

rs, MPa

Milled

2.0

200

0.1

0.2

0.1

0.2

0

Shot peened

3.0

500 C/30 min

200

2.5

400

2.0

0

0.1

0.2

0.3

0.1

0.2

0.3

Fig. 4

Macro residual stress and half-width of the x-ray interference line versus distance from surface before and after

annealing of ground, milled, and shot-peened SAE 1045 samples in a normalized state

and tempered states.

In order to obtain some information about the

relaxation behavior of macro residual stresses in

subsurface layers, the values of the residual

stresses after the annealing processes of AISI

4140 mentioned above were measured for different distances from the surface (Ref 26, 27).

The evaluated parameters of the Avrami approach vary only at the surface itself from those

values measured below the surface. There, the

evaluated data show no significant tendency and

amount to the mean values DH 2.99 eV, m

0.172, and B 6.09 1017 min1. As

shown in Fig. 9, the dependence of the macro

residual stresses after different annealing times

at 450 C (840 F) on the distance from surface

can be described quantitatively using the surface

material properties and the mean values for all

subsurface layers. The agreement between measured values and modeled curves is very good.

During heating to sufficiently high annealing

temperatures, a distinct relaxation of macro residual stresses occurs. An attempt was made to

model this behavior by extending the Avrami approach to nonisothermal stress relaxation (Ref

27). The transient relaxation of the macro residual stresses was calculated for specimens that

were immersed up to 90 s in a salt bath at 450

C (840 F). The temperature at the specimen

surfaces developed according to the T(t) curve

shown in Fig. 10. For calculation of the relaxation of residual stresses, the real T(t) relationship

was partitioned into a staircase curve with small

equidistant steps and isothermal sections. Calculation of the relaxation was determined by a

numerical integration after the so-called stresstransient method (Ref 27) under the application

of the Avrami quantities DH, m, and B (Table 1)

which are valid for the surface state. The magnitudes of the macro residual stresses measured

after interruption of the heating and cooling

down to room temperature are marked with triangles. It becomes evident that the stresstransient method describes the measured values

in an excellent manner. Figure 10 also shows the

isothermally calculated relationship between

heating time and absolute values of residual

stresses. While there are distinct differences during the first stage of heating time of approximately 20 s, the curves calculated isothermally

and by the stress-transient method closely approach at long annealing times.

A similar discussion of thermally activated

processes was carried out for other steels (Ref

16, 17, 28, 30, 31) and for AlMg alloys and

CuZn alloys (Ref 30, 31) with surface residual

stresses originating from a variety of processes.

In the case of these steels, the activation enthalpy

DH depends on the state of the material and lies

in the range of 1.1 to 2.6 eV. DH is lowest for

relaxation of residual stresses due to hardening

and highest for those due to deformation in soft

annealed states. That proves unequivocally that

residual stress relaxation in steels can occur by

several processes.

Characteristic structural changes occur during

the deformation, machining, or hardening of

steels. Typically, an increase in the dislocation

density and a change in the dislocation arrangement are observed. In the case of hardening, the

concentration of solute interstitial atoms differs

from the equilibrium value, and this, together

with the presence of dislocations, has a decisive

influence on the residual stress fields. If a single

heat treatment is carried out in the temperature

range corresponding to recovery (T 0.5 Tm),

the dislocations adopt arrangements of lower energy by elementary processes such as glide and

cross slip by screw dislocations and glide and

climb by edge dislocations. In the case of hardened material, diffusion of carbon atoms dependent on the annealing temperature and accompanied by the formation of characteristic

carbides is superimposed on these processes.

The rate-determining process, with the exception

of the early stages in the annealing of hardened

steels (Ref 36), is clearly the thermally activated

climb of edge dislocations (Ref 17). If diffusion

of matrix atoms occurs along the edge dislocations to the dislocation core, the activation enthalpy should be DH 0.5 DHS, where DHS is

the activation enthalpy of self-diffusion. If volume diffusion predominates, the DH value for

climb is determined by DHS. In the real case,

both processes occur simultaneously but to different degrees. The dislocation density and arrangement are of considerable importance. In the

case of randomly distributed dislocations or tangles of extremely high density, qt for example,

in hardened steels where qt 1012 cm2 (Ref

23, 24)residual stress relaxation is expected

to involve dislocation-core diffusion-controlled

climb by edge dislocations. Predominantly volume diffusion will determine recovery if the dislocation configurations are relatively stable and

consist of plane arrangements, cell walls, or subgrain or low-angle grain boundaries. This recovery process probably occurred in the investigated

shot-peened quenched and tempered steels and

normalized steels as well as nonferrous alloys.

In view of the activation energies, the residual

stress relaxation in the hardened steels can probably be classified between these two extremes

and occurred by two recovery mechanisms in

competition with each other.

Up to now, discussion has centered on residual stress relaxation at temperatures Ta 0.5

Tm that is, those temperatures brought about

by typical recovery processes. In this case, mechanical parameters such as hardness and yield

strength are not significantly altered. During a

recrystallization anneal at Ta 0.5 Tm, the dislocation density rapidly takes very small values

as a result of the growth of new grains. This leads

to complete removal of macro residual stresses

and to small micro residual stresses, but is associated with pronounced changes in mechanical

properties. If extensive residual stress reduction

change in yield strength or tensile strength, the

annealing temperature and time must be chosen

to correspond with the recovery stage and not

with recrystallization.

Residual stress relaxation by heat treatment is

fundamentally impossible if in a predominantly

uniaxial residual stress state rrs is smaller than

the creep yield strength. This resistance for the

onset of plastic creep deformation, designated in

the following by Rce, is characterized by the

creep strain limit at vanishingly small plastic deformation. As shown schematically in Fig. 11,

Rce decreases with increasing temperature and

load time. In contrast to relevant times regarding

the creep condition of a high-temperature component at service, only very short times are necessary for residual stress relaxation by creep processes (dislocation creep, grain-boundary glide,

or diffusion creep). With increasing T and/or t,

Rce approaches a localized residual stress peak

of magnitude rrs. For T Tti, rrs equals Rce

associated with the localized onset of creep deformation. Further increases in temperature or

time result in an increasing and measurable microplastic creep strain. As illustrated in Fig. 11,

residual stress relaxation begins at higher temperatures (Ti) the smaller the load time (ti) or the

residual stress (rrs) and the greater the creep resistance (Rce) of the material. Changes in Rce can

be achieved by deliberate alterations in the state

0

Shot peened

400 C/30 min

400

550 C/30 min

800

Ta,4

Normalized

Ta,3

6

300 C/60 min

400

HW, 2

Shot peened

rs, MPa

Ta,2

rs

log ln (0/rs)

Normalized

800

Quenched and tempered

Ta,1

log ta

400

5

4

Shot peened

800

Hardened

0

Fig. 5

log ta

0.2

Hardened

0.4 0

Distance from surface, mm

0.2

Ha/ln10

3

0.4

Macro residual stress and half-width versus distance from surface before and after annealing of shot-peened

SAE 1045 samples in normalized, quenched and tempered, and hardened states

1/kTa

Fig. 6

Avrami approach parameters

Loading stresses acting in opposition to the residual stresses in a given region of the material

delay the onset of microplastic deformation in

this region. Since the distribution of residual

stresses in a component is always inhomogeneous, there will be other regions in which superposition of rrs and rls in the same sense promotes plastic deformation. Therefore, it depends

on the superposition of rls and rrs at any point

in the component if a loading stress retards (|rls

rrs|max |rrs|max) or enhances (|rls rrs|max

|rrs|max) residual stress relaxation.

At this stage it is interesting to explain the

result shown in Fig. 1 and 2, where residual

stress relaxation in a deformed sample of SAE

1045 is delayed in comparison to that in SAE

1023 despite the compressive residual stresses

having double the magnitude. Due to its higher

carbon content, SAE 1045 contains a greater number of ferrite-cementite phase boundaries than

SAE 1023. These phase boundaries represent stable obstacles to dislocation slip. Apparently, the

restricted possibilities for recovery-driven movement of dislocations to arrangements of lower en-

(Eq 9)

equivalent residual stress is thus dependent on

the differences between the principal residual

stress components. These in their turn are proportional to the shear stresses acting on dislocations in the slip systems. Residual stress relaxation therefore does not occur for

rrs

eq Rce

ep C(T)rntm

(Ref 16, 17). C is a quantity depending on the

temperature and the condition of the material;

the exponents have the values n 1 and 0 m

1; and the creep strain increases with applied

stress and with time. Residual stress relaxation

600

rrs

eq Rce

0.4

10

500

250 C

400

300 C

300

350 C

200

400 C

100

eq Rce obtains only briefly, since the

immediate onset of creep deformation attempts

to restore the condition rrs

eq Rce.

10

102

ta, min

103

104

(a)

Stresses on Relaxation Behavior

3.2

Ta = 250 C

residual stress relaxation in machined or hard-

102

300 C

2.8 350 C

400 C

450 C

2.4

103

104

600 C

90

525 C

0

10

80

410 C

60

375 C

340 C

320 C

50

250 C

20

40

30

0.8

200 C

10

150 C

102

10

ta, min

103

104

(b)

1.4

200

Ta = 250 C

70

rs

0.4

50 C

Ta = 4

log ln (0/rs)

Ta = 450 C

0

10

Annealing time, h

1

(Eq 10)

1.2

300 C

350 C

400 C

1.0

250

450 C

200

rs

, MPa

micro

(Eq 8)

rrs

eq

ened steels occurs more quickly or at lower temperatures, the greater the magnitude of the residual stresses themselves. This is a consequence of

temperature, time, and stress-dependent processes similar to those observed in so-called primary microcreep. Empirical relationships of the

form

|rs|, MPa

larger residual stress. The delayed residual stress

relaxation in deformed SAE 1045 is thus explained by an increase in the creep resistance Rce.

In the case of a multiaxial residual stress state,

the residual stress rrs employed in the arguments

just given must be replaced by an equivalent residual stress rrs

eq. If the principal components are

rs

rs

rrs

1 , r2 , and r3 , it can be formulated to

HW, 2

that have an additional work-hardening effect

(Ref 17) shift the onset of residual stress relaxation and the entire rrs versus T curve to higher

temperatures.

Residual stress relaxation is also affected by

the superposition on a localized residual stress

rrs of an applied loading stress rls in the same

direction. Increasing rls values shift the onset of

residual stress relaxation to shorter times and/or

lower temperatures according to the relation

0.8

150

0.6

10

1.2

102

10

ta, min

103

104

(c)

1.6

1

Fig. 7

the (a) absolute values of macro residual stress,

(b) half-width, and (c) mean strain and micro residual stress

on the surface and their description by the Avrami approach for shot-peened AISI 4140 in a quenched and tempered condition

Fig. 8

2

log ta

(Eq 11)

However, macro residual stresses are inhomogeneously distributed over the cross section, and

for the residual stress relaxation in a localized

area the following relationship holds true:

rrs

0

E

e t e p e e 0

(Eq 13)

rs

constant ep

r

E

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

rs

(rrs

0 r )

ep

E

r rs

E

(Eq 14)

(Eq 12)

plastic strains of several tenths of a percent maximum are produced by complete residual stress

relaxation (rrs 0); that is, the deformation is

in the microcreep range.

Equations 10 and 11 form the basis for quantitative estimates of the residual stress relaxation.

Multiaxial and inhomogeneous residual stress

states are neglected or excluded. Although Eq 10

cannot be substituted directly in Eq 11 on account of the variable value of the stress (r

rrs), it can be seen qualitatively from a combi-

plastic strain rates, and between the residual relaxation rate and the actual residual stress to the

power of the Norton exponent n, respectively.

The experimental data shown in Fig. 12 reveal

strain rates that present a strong dependency on

temperature and residual stress values (Ref 26).

Furthermore, the data reveal strain rates that are

typical for creep processes. This finding supports

the conclusion drawn above that diffusioncontrolled dislocation creep in the residual stress

field should be the rate-controlling process for

the relaxation of residual stresses in steels.

650

200

500

T( t )

200

Ta = 450 C

rs, MPa

rs, MPa

ta = 6000 min

60 min

400

6 min

600

Measured

Calculated

400

Tsalt = 450 C

300

Surface data

450

Transient relaxation

350

T, C

550

250

0.1

0.2

0.3

Distance from surface, mm

0.4

at Ta 450 C (840 F) and different annealing

times versus distance from the surface of shot-peened AISI

4140 in a quenched and tempered condition (450 C, or

840 F, for 2 h)

Fig. 9

rs+ ls

t2, 2

20

40

60

80

100

650 C (1200 F)

450 C (840 F)

Hardened

T1

T2

Temperature, T

Fig. 11

short-time immersion in a salt bath at 450 C

(840 F) versus time. Comparison is made with curves calculated with the stress-transient method using the occurring

T(t) dependence and with modeling for isothermal annealing.

Fig. 10

Measuring property

DH, eV

B, min1

3.30

0.080

1.40 1018

Macro residual stress

Half-width

Mean strain (micro rs)

Macro residual stress

t2 >> t1

2

1

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

t, s

Table 1 Material properties DH, m, and B of the Avrami approach determined for different

heat treatment conditions of the shot-peened steel AISI 4140

Normalized

T 1

100

105

106

107

Heat treatment

t1, 1

rs

200

Isothermal relaxation

0 min

800

In certain cases in practice, uniaxial deformation is often employed in addition to stressfree annealing to relieve residual stresses. In the

case of forming, for example, the residual

stresses can be reduced by a second forming

stage using a smaller reduction in the crosssectional area. This can be achieved by redrawing, restretching, rerolling, repressing, and

straightening (Ref 16, 17). However, these techniques can be used only on simply shaped components with a uniform cross section. In the case

of welded seams, a uniaxial load is applied to

reduce or redistribute macro residual stresses.

When a critica1 value of the applied loading

stress is exceeded, directed dislocation movement converts the elastic strain associated with

the macro residual stress into micro plastic

strain.

Several typical examples will serve to illustrate the relaxation of residua1 stresses due to

joining and shot peening by uniaxial deformation. In an evaluation of the effect on the strength

of macro residual stresses set up during welding,

it is important to know the stability of these

stresses on loading the weld. Figure 13 illustrates

the considerable reduction in macro residual

stresses accompanying tensile loading of joints

r

E

Uniaxial Deformation

Strain rate p, s1

et constant ep ee ep

nation of the two expressions that increasing residual stress values lead to a more effective residual stress relaxation. As a result of the greater

driving force, shorter times and/or lower temperatures are necessary. This is in agreement

with the experimental results presented.

The influence of the magnitude of the residual

stresses can also be illustrated by describing the

residual stress relaxation with the Norton approach, which is known from high-temperature

creep. The total strain rate as the sum of elastic

and plastic strain rate must vanish according to

>>

test. It is much more like a stress relaxation experiment. In the latter case, the total strain remains constant while elastic strain is converted

into plastic strain. With a homogeneous stress

distribution over the cross section of a specimen,

the following expression would be valid:

3.38

3.29

2.48

2.64

2.33

0.080

0.122

0.116

0.096

0.110

1.90 1021

1.22 1021

1.09 1013

5.32 1012

1.20 1016

108

T = 450 C

109

400 C

350 C

300 C

1010

1011

60

100

200

400

250 C

600

1000

rs, MPa

Plastic strain rate versus mean residual stress of

shot-peened AISI 4140 in a quenched and tempered condition (450 C, or 840 F, for 2 h). Determined

from the data of Fig. 8(a)

Fig. 12

produced by electron beam welding and tungsten

inert gas (TIG) welding in a maraging steel (Ref

37). In the electron beam welded specimen, it

begins at about 70% of the yield strength Re of

the unannealed weld. Residual stress relaxation

is not complete. In the case of the TIG weld,

noticeable residual stress relaxation first occurs

above Re. By the time the tensile strength Rm is

reached, residual stress relaxation is virtually

complete.

Shot peening residual stress states under tensile or compressive loading are unstable, and relaxation occurs when critical loading stresses are

exceeded. Corresponding results of surface residual stresses of the quenched and tempered

steel AISI 4140 are presented in Fig. 14 as a

function of loading stress and in Fig. 15(a) as a

function of total deformation (Ref 38). At tensile

loading, relaxation of the peening-induced compressive residual stresses in the surface of 540

MPa (80 ksi) sets in at 95% of the yield

strength of the unpeened material (Re(t) Rp0.01(t)

1180 MPa, or 170 ksi). In compression, relaxation starts at 45% of the yield strength of the

unpeened material (Re(c) Rp0.01(c) 1300 MPa,

or 190 ksi). In the quenched and tempered condition of this material exists a typical strengthdifferential effect of Re(c) Re(t) 120 MPa (20

ksi) (Ref 38). The relaxation rate of macro resid-

200

Re

occurs much more rapidly during tension than

during compression. After a total strain et

1% (rls 1250 MPa, or 180 ksi), the

surface residual stresses rrs are completely removed (see Fig. 15a). If et 1%, the sign of

rrs changes. In the case of tensile loading, the

residual stress relaxation, however, is incomplete. Thus, the relaxation behavior is different

and anisotropic during tensile and compressive

loading. Combined with the relaxation of macro

residual stresses, only a relatively small decrease

of the half-width values, which are a measure of

the micro residual stresses, is observed at the surface. Figure 15(b) illustrates this behavior as a

function of total strain.

The effect of different strengths on the relaxation behavior of shot peening residual stresses

at the surface of AISI 4140 due to uniaxial deformation is illustrated in Fig. 16. Results of Fig.

14 for specimens quenched and tempered at 450

C (840 F) are compared with results for normalized specimens and for specimens quenched

and tempered at 650 C (1200 F) (Ref 39). The

courses of rrs versus rls are similar. However,

the absolute values of the critical loading stresses

for the onset of residual stress relaxation in tension |rlscrit(t)|, and in compression, |rlscrit(c)|, increase with increasing yield strength of this steel

in different heat treatment conditions.

Figure 17 represents the influence of loading

stresses rls on the shot peening residual stresses

Rm

4140 for different testing temperatures (Ref 40).

The rrs versus rls course at 25 C (75 F) is the

same as in Fig. 14. At higher temperatures (250

and 400 C, or 480 and 750 F), the initial macro

residual stresses are reduced by heating and

waiting for temperature compensation due to

thermal residual stress relaxation. The values of

the critical loading stresses for the onset of residual stress relaxation during tension and compression as well as the relaxation rate decrease

clearly with increasing temperature.

The relaxation of macro residual stresses by

uniaxial deformation begins at relatively small

loads or plastic strains. It may be complete or

partial, both the degree and rate of residual stress

relaxation depending on the type and state of the

material as well as on the nature of the applied

load.

The deformation behavior and relaxation of

macro residual stresses in materials can be

roughly approximated in terms of a cylindrical

rod with longitudinal residual stressesnamely,

constant compressive residual stresses (rrs

s ) at

the surface and constant tensile residual stresses

(rrsc ) in the core (Fig. 18). The surface is characterized by the yield strength Re,s and the core

by Re,c. Plastic deformation commences in the

core at the critical tensile loading stress:

rlscrit(t) Re(t),c rrs

c

(Eq 15)

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

200

200

400

400

800

Re Rm

1200

0

rs, MPa

rs, MPa

ls, MPa

Welding residual stress versus tensile loading

stress of a maraging steel. (a) Transverse residual stresses in the welding seam center of an electron beam

welded joint. (b) Transverse residual stresses 3 mm (0.12

in.) from the welding seam center of a TIG welded joint

Fig. 13

200

s|

400

tensile loading for

600

6

c

4

t, %

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

s|

15

HW, min

rs, MPa

0

200

600

1600

15

30

45

400

800

800

1600

60

6

ls, MPa

Macro residual stress at the surface versus loading stress under tension and compression, respectively, of shot-peened AISI 4140 in a quenched and

tempered condition (450 C, or 840 F, for 2 h)

Fig. 14

t, %

residual stress and (b) the half-width change at

the surface of shot-peened AISI 4140 in a quenched and

tempered condition (450 C, or 840 F, for 2 h)

Fig. 15

(Eq 17)

30

200

(Eq 16)

(Eq 18)

meaning of resistances against residual stress relaxation by uniaxial deformation.

The critical compressive loading stress,

rlscrit(c), can be determined experimentally from

corresponding loading and unloading tests. The

yield strength of the surface, Re(c),s, can be estimated from Eq 16 with measurements of the surface residual stresses and the knowledge of the

initial surface residual stress by this relationship:

ls

|Re(c),s| |rcrit(c)

| |rrs

s|

(Eq 19)

modified yield strengths of the surface occur.

Consideration of an isotropic biaxial surface re-

req |Re(c),s|

2

ls

rs

(rlscrit(c))2 (rrs

s ) rcrit(c) rs

(Eq 20)

relationship leads to somewhat smaller yield

strengths, as with a uniaxial estimation (Ref 29,

41). The application of Eq 20 for the shot-peened

AISI 4140 in different conditions results in the

Re(c),s values summarized in Table 2. The corresponding critical loading stresses for initiation of

macro residual stress relaxation rlscrit(c), the yield

strengths of the unpeened conditions Re(c), the

ratio Re(c),s /Re(c), and the half-widths ratio of surface and core are also indicated. For the normalized condition the relation

|Re(c),s| |Re(c)|

is valid. For the quenched and tempered conditions, however, the yield strengths show with

|Re(c),s| |Re(c)|

thus generate in comparison with the core a

work-hardened surface state in the normalized

condition and a work-softened surface state in

the quenched and tempered conditions. The final

observations seem to be in contradiction with the

work-hardening behavior obtained on the basis

of the measurements of the half-width values of

surface and core. The last column in Table 2

1.05 and 1.58. Therefore, in all cases a peeninginduced microstructural work-hardening effect

appears. The smallest effect occurs at the 450 C

(840 F) quenched and tempered condition. The

other heat-treated states show a considerably

larger microstructural work-hardening effect.

These apparent contradictions become clear

with a more exact consideration of the conditions

of the local deformation of the surface layers

during shot peening and the subsequent uniaxial

compression test. Besides a biaxial plastic

stretch-forming of the surface area with maximum flow stresses and work hardening direct at

the surface, the peening-induced deformations

include additional Hertzian stresses with maximum effects below the surface (Ref 42). Now,

at the determination of the surface yield strength

Re(c),s under homogeneous compressive loading,

the direct surface layer is deformed exactly inverse to the deformation direction during shot

peening. In this case, the Bauschinger effect relieves the deformations in the inverse direction

and reduces the resistance of the work-hardened

surface layers due to back stresses and an anisotropic mobility of dislocations. It is known that

the Bauschinger effect appears especially distinctive in quenched and tempered steels (Ref

43). Therefore, it is not surprising that the shot

peening treatments of quenched and tempered

AISI 4140 lead to values of Re(c),s /Re(c) 1. The

surface yield strengths are controlled not only by

the peening-induced surface work-hardening,

which is characterized by the half-width ratio

HWs /HWc, but also by the Bauschinger effect.

Obviously, the condition Re(c),s /Re(c) 1 is valid

if the peening-induced work-hardening has a

stronger effect than the Bauschinger effect. If the

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

A quantitative description of the relaxation

behavior of multiaxial macro residual stresses

with known distribution over the specimen cross

section under uniaxial deformation is possible

with the aid of finite-element modeling (Ref 28,

44). Isoparametric rectangular elements with

plane strain behavior were applied for this purpose. The corresponding depth courses of the

macro residual stresses as well as the initial parts

of the stress-strain curves under tensile and compressive loading must be known and must be input with certain assumptions in a discrete form.

Figure 19 shows corresponding results of tests

at 25 and 400 C (75 and 750 F) for AISI 4140

in a quenched and tempered condition. There is

a good agreement between experimental results

and finite-element modeling. An important finding is that besides the pure mechanical and thermal relaxation of macro residual stresses during

heating and waiting for temperature compensation, no additional relaxation effects occur under

200

0

rs, MPa

shot peening results on the basis of the von

Mises hypothesis for the onset of residual stress

relaxation at compressive loading to:

T = 25 C

T = 250 C

T = 400 C

200

400

600

1600

800

800

1600

ls, MPa

,

200

,

,

Macro residual stress at the surface versus loading stress under tension and compression, respectively, at different deformation temperatures for shotpeened AISI 4140 in a quenched and tempered condition

(450 C, or 840 F, for 2 h)

Fig. 17

Quenched and

tempered 450 C

Quenched and

tempered 650 C

Normalized

rs, MPa

Surface

Core

rs

rs

200

Surface

Cross

section

area

400

srs

600

1600

1200

800

400

400

800

1200

1600

ls, MPa

Macro residual stress at the surface versus loading stress under tension and compression, respectively, of shotpeened AISI 4140 in different heat treatment conditions: normalized; quenched and tempered at 450 C (840

F), for 2 h (qt 450); and quenched and tempered at 650 C (1200 F) for 2 h (qt 650)

Fig. 16

residual stress for residual stress relaxation due

to uniaxial loading

Fig. 18

influenced by macro and micro residual stresses.

Since in certain cases they can be considered as

locally variable mean stresses, they can lead to

substantia1 increases in fatigue strength (cf. Ref

16, 17). The interaction of macro residual

stresses with the cyclic deformation mechanism

during fatigue in the microcrack-free stage is of

particular importance. Coupled with this is the

stability or the cyclic relaxation of residual

stresses. Residual stress relaxation occurs when

a critical value of the applied loading stress amplitude is exceeded and a cyclic directed dislocation movement converts the elastic strain associated with the macro residual stress into

micro plastic strain.

Numerous investigators have shown that the

effect of macro residual stresses decreases with

increasing stress amplitude and growing number

of cycles as a result of residua1 stress relaxation

(cf. Ref 16, 17). This is illustrated in Fig. 20 for

quenched and tempered AISI 4140 in a shotpeened condition that has undergone a push-pull

fatigue test (Ref 28, 46). Figure 20(a) presents

cyclic deformation curves with a plot of the plastic strain amplitudes, ea,p, versus the number of

cycles, N, during stress-controlled cyclic loading

for different stress amplitudes, ra. Initially, the

specimens show macroscopically a quasi-elastic

cyclic deformation behavior at 400, 500, and 600

MPa (60, 70, and 90 ksi). At loading with ra

700 MPa (100 ksi), small plastic strain amplitudes occur from the beginning of the test. After

a distinct number of cycles for incubation, which

decreases with increasing stress amplitude, cyclic work-softening occurs at all amplitudes.

This augmentation of plastic strain amplitude

continues until specimen failure. The maximum

plastic strain amplitude observed just before failure increases with increasing stress amplitude.

The alterations of the shot peening-induced

macro residual stresses at the surface rrs with the

number of cycles are shown in Fig. 20(b). The

residual stress values are strongly reduced in the

first cycle. Afterward, a linear dependence of the

fashion during the relaxation of macro residual

stresses by uniaxial deformation. So far, only

isolated experimental data based on x-ray profile

analysis are available.

When a material is formed or machined, both

directed and inhomogeneous micro residual

stresses are set up. Those of the first type are the

result of back stresses due to dislocation pileups

at grain or phase boundaries and elastic strained

second phases of heterogeneous materials. If

these dislocations move in the reverse direction

during localized plastic deformation, the back

stresses and hence the directed micro residual

stresses are initially reduced. Further deformation causes renewed buildup of back stresses in

the opposing direction associated with dislocation multiplication and hardening, which must

once again increase the micro residual stresses.

Dislocation arrangements in a random distribution or in tangles or cells give rise to inhomogeneous micro residual stresses. Microplastic

deformation can lead to the rearrangement of

dislocations into arrangements of lower energy

and thus bring about micro residual stress relaxation. If new dislocations are produced, a renewed buildup of micro residual stresses is superimposed on the relaxation process.

A reduction in the micro residual stresses in

hardened steels is observed both on deformation

and on machining (Ref 16, 17, 32, 45). Two superimposed effects can operate here. After hardening, a very high density of dislocations is presenteither randomly distributed or in tangles.

Microplastic deformation brings about a rearrangement of these dislocations into configurations with lower distortion energy and therefore

a reduction in the micro residual stresses. On the

other hand, solute carbon atoms may be induced

to jump into the energetically more favorable octahedral sites in the martensite lattice under the

influence of the stress field of the moving dislocations. This causes a reduction in the tetragonality and hence the lattice distortion due to

solute carbon atoms (Ref 45).

(Eq 21)

200

T = 25 C

0

rs, MPa

according to a logarithmic creep law:

200

400

600

Calculated

Measured

T = 400 C

200

0

rs, MPa

Deformation

200

400

Calculated

Measured

600

1600

800

800

1600

ls, MPa

Measured and calculated macro residual stress

at the surface versus loading stress under tension and compression, respectively, at different deformation temperatures for shot-peened AISI 4140 in a quenched

and tempered condition (450 C, or 840 F, for 2 h)

Fig. 19

a = 700 MPa

a,p in 103

2

1

(a)

600 500

400

0

0

700

rlscrit(c)

Temperature

Re(c)

F

MPa

ksi

MPa

ksi

MPa

ksi

Re(c),s /Re(c)

25

250

400

75

480

750

275

300

340

40

45

49

345

300

275

50

45

40

480

496

516

70

72

75

1.39

1.65

1.87

1.58

1.33

1.28

650 C (1200 F)

20

450 C (840 F)

25

250

400

70

75

480

750

340

600

400

360

49

85

60

52

805

1300

950

830

115

190

140

120

685

953

758

606

100

138

110

88

0.85

0.73

0.79

0.73

1.56

1.05

1.05

1.05

Normalized

500

a = 400 MPa

400

Re(c),s

C

Heat treatment

200

Table 2 Critical compressive loading stress rlscrit(c), yield strength Re(c) of unpeened conditions, surface yield strength Re(c),s, yield strength ratio Re(c),s /Re(c), and ratio of half-width values

HWs /HWc of shot-peened AISI 4140 in different heat treatment conditions

rs, MPa

600

HWs /HWc

600

10

(b)

102

104

Number of cycles, N

106

residual stress versus number of cycles during

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

Fig. 20

tionally peened conditions and warm-peened

conditionsthe changes in the macro residual

stresses, rrs

l,s, measured at the side of the specimens primarily loaded in compression as a function of the logarithm of the number of cycles.

For both peening variants, the residual stresses

at the surface relax more rapidly with increasing

stress amplitude during the first cycle as well as

during further cycling. The linear correlation between residual stresses and the logarithm of the

number of cycles N according to Eq 20 can be

recognized for wide intervals of N 1. In the

conventionally peened samples, however, the

900

800

200

700

600

300

l,s

500

400

400

300

500

600

700

10

102

103

104

Number of cycles, N

105

106

107

(a)

0

a,s, MPa

100

1000

Tpeen = 290 C

900

200

800

700

rs , MPa

600

500

400

400

500

400

600

Top longitudinal

Bottom longitudinal

Top transverse

Bottom transverse

200

0

10

1000

100

l,s

|rs|, MPa

600

a,s, MPa

Tpeen = 20 C

300

800

warm-peened condition. Pronounced differences

can be seenespecially in the first cycle, where

the residual stresses in the warm-peened samples

show a higher stability than those in conventionally peened samples. For N 1, the relaxation

rate of the variant peened at room temperature is

also higher compared to the variant peened at

290 C (555 F).

The resulting surface values of the half-widths

HWs, related to their initial values HWs,0, are

given in Fig. 23 as a function of the logarithm

of the number of cycles. The half-widths of the

rs , MPa

due to the so-called cyclic creep, where the properties A and m can be determined for each stress

amplitude from the experimental data in the linear sections of the curves in Fig. 20(b). In the

case of the stress amplitude ra 400 MPa (60

ksi), after the first cycle no further residual stress

relaxation is observed with increasing number of

cycles N. Obviously, this ra value is smaller than

the materials resistance against cyclic residual

stress relaxation ra,crit at the surface. For the two

highest stress amplitudes, ra 600 and 700

MPa (90 and 100 ksi), at the end of fatigue life

greater residual stress relaxation rates are observed than predicted by the rrs /log N law. The

beginning of this phase can be correlated with

the onset of cyclic work softening in the ea,p versus N curves in Fig. 20(a). Hence, increasing

plastic strain amplitudes at constant stress amplitude may cause increasing relaxation rates of

residual stresses.

For shot-peened AISI 4140 in a quenched and

tempered state and subjected to alternating bending at the stress amplitude ra,s 700 MPa (100

ksi), Fig. 21 presents the course of the absolute

values of the macro residual stresses at the surface of flat specimens versus the number of cycles in the longitudinal and transverse directions

(Ref 40). While the longitudinal residual stresses

at the top side (which was loaded in compressive

direction in the first half-cycle) relax from 600

MPa (90 ksi) to about 480 MPa (70 ksi),

the longitudinal residual stress at the bottom side

shows no considerable relaxation. In the second

half-cycle, this trend is reversed and the residual

stresses measured become similar. In the course

of further cyclic loading up to 105 cycles, the

residual stresses decrease linearly with the logarithm of the number of cycles according to Eq

21. After crack initiation at 105 Ni 4 105,

further reductions in the magnitudes of the residual stresses are measured. The residual

stresses in the transverse direction always relax

more slowly than in the longitudinal direction.

For initial stress amplitudes at the surface between 300 and 1000 MPa (45 and 145 ksi), Fig.

22 summarizes for the same steelin conven-

102

104

Number of cycles, N

700

106

10

102

103

104

Number of cycles, N

105

106

107

(b)

Absolute values of macro residual stress at the

surface versus number of cycles during alternating bending tests of shot-peened AISI 4140 in a

quenched and tempered condition (450 C, or 840 F, for

2 h) for a surface stress amplitude of 700 MPa (100 ksi)

Fig. 21

Macro residual stress at the surface in longitudinal direction versus number of cycles during alternating

bending tests of quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h) at different surface stress

amplitudes. (a) Conventionally shot-peened condition (Tpeen 20 C, or 70 F). (b) Warm-peened condition (Tpeen 290

C or 555 F)

Fig. 22

conventionally peened samples relax for ra,s

700 MPa (100 ksi). With increasing stress amplitude, an increasing relaxation is observed.

The reduction of the half-widths and, hence, of

the micro residual stresses is much more pronounced for the samples peened at room temperature compared to the other peening variant.

For the warm-peened states, the measured values after cycling are often a bit higher than the

initial values. Only for ra,s 900 MPa (130

ksi) and N 104 can a significant decrease be

seen.

For the same samples peened at 20 or 290 C

stresses was investigated in detail (Ref 47, 48).

The depth distribution of the residual stresses

and the half-widths in the region close to the

surface was determined at different numbers of

cycles at a surface stress amplitude 1000 MPa

(145 ksi). The curves for the conventionally

peened samples in Fig. 24(a) show that the compressive residual stresses in the region close to

the surface are strongly reduced with increasing

number of cycles. During the first cycle, the absolute value of the residual stress of the surface

decreases from about 600 MPa (90 ksi) to 350

1.0

0.9

a,s, MPa

HWs/HWs,0

300

400

0.8

500

600

700

0.7

800

900

1000

0.6

1

10

102

103

104

105

106

107

Number of cycles, N

(a)

1.0

150 MPa (20 ksi). The depth where the residual

stresses change their sign is not significantly affected. As shown in Fig. 24(a), a shot peening

treatment at 290 C (555 F) considerably delays

the relaxation of residual stresses. During the

first cycle, the initial absolute value of the residual stress at the surface, |ra,s| 660 MPa (95

ksi), is reduced only by about 140 MPa (20 ksi).

After N 1000, |ra,s| 380 MPa (55 ksi) is

measured. The affiliated depth distributions of

the half-widths of the conventionally peened

condition in Fig. 25(a) verify that with increasing number of cycles, reductions of the halfwidths and consequently of the micro residual

stresses occur due to larger rearrangements of

mobile dislocations in the surface layers. In opposition to this finding, under the same loading

conditions the warm-peened condition shows

practically no relaxation of the micro residual

stresses. Obviously, the dislocation arrangements in the warm-peened condition under cyclic loading are considerably more stable than in

the conventionally peened condition. This is due

to dynamic and static strain aging effects that

occur during and after warm peening (Ref 47,

48). Dynamic strain aging results in a more diffuse and stable dislocation structure caused by

pinning of dislocations by solute carbon atoms

and formation of extremely fine carbides.

The effect of stabilized residual stresses on fatigue behavior is illustrated in Fig. 26, with S-N

curves at alternating bending of the conventionally peened and two warm-peened conditions in

comparison with the ground condition. The S-N

curves for a failure probability of P 50% were

determined in each case from approximately 30

tests after the arcsin P procedure (Ref 49). Shot

peening at room temperature increases the fatigue strength Rfb for about 90 MPa (13 ksi) from

440 MPa (65 ksi) to 530 MPa (80 ksi) compared

to the ground condition. Compared to that, peening at 200 and 290 C (390 and 555 F) increases

the fatigue strength to 590 and 640 MPa (85 and

95 ksi), respectively.

a,s, MPa

0.9

400

stresses can be divided into four phases (Ref 16,

17, 50):

HWs/HWs,0

500

600

0.8

700

800

deformation processes

Quasi-static relaxation during the second half

of the first cycle, which is influenced by the

load reversal

Cyclic relaxation at 1 N Ni (Ni number of cycles to crack initiation) due to cyclic

deformation processes, which usually leads to

a linear reduction of the residual stresses as a

function of the logarithm of the number of

cycles according to Eq 21 due to cyclic creep

Cyclic relaxation at Ni N Nf (Nf number of cycles to failure) for sufficiently high

stress amplitudes at the surface, which is

900

0.7

1000

0.6

1

(b)

10

102

103

104

105

106

107

Number of cycles, N

Ratio of half-width at the surface versus number of cycles during alternating bending tests of quenched and

tempered AISI 4140 (450 C, or 840 F, for 2 h) at different surface stress amplitudes. (a) Conventionally shotpeened condition (Tpeen 20 C, or 70 F). (b) Warm-peened condition (Tpeen 290 C, or 555 F)

Fig. 23

enforced in the plastic zones at the tips of

initiated surface cracks.

Assuming a predominantly uniaxial macro residual stress state, it is a fundamental principle

that relaxation of residual stresses at the surface

will commence within the first cycle, if the stress

amplitude ra reaches the critical loading stress

rs,crit. It should be noted that in the case of pushpull tests the critical loading stress rs,crit is identical with the property |rlscrit(c)| (see the section

Residual Stress Relaxation by Uniaxial Deformation in this article). If ra rs,crit, the macro

residual stresses should remain stable in the first

cycle. For N 1, relaxation of macro residual

stresses can begin gradually if the critical loading stress amplitude ra,crit is exceeded (Ref 16,

17, 40, 41, 47, 48, 50):

ra ra,crit

resistance against relaxation of macro residual

stresses. Equation 16 is equally valid for the

critical cycling loading stress ra,crit if Re(c),s is

replaced by the corresponding cyclic yield

strength for surface R cycl

e,s . That means for a predominantly uniaxial residual stress state, relax-

200

with cyclic plastic deformation in the direct surface if this condition is fulfilled:

rs

a ra,crit Rcycl

e,s m |rs |

(Eq 22)

stress sensitivity factor. In this consideration, it

is assumed that m 1. The macro residual

stresses are stable if the condition is fulfilled:

rs

ra ra,crit Rcycl

e,s |rs |

(Eq 23)

For cyclic bending tests under the conditions illustrated in Fig. 27, the critical loading stress

rs,crit can be found by plotting the residual stress

values after the first cycle from Fig. 22 versus

the loading stress on the surface. This is given

in Fig. 28 for conventionally peened and warmpeened samples. The data points at N 1 show

that the critical loading stress at which quasistatic residual stress relaxation begins is much

higher for the warm-peened condition (rs,crit

500 MPa, or 70 ksi) than for the conventionally

peened variant (rs,crit 310 MPa, or 45 ksi) due

to the above-mentioned strain aging effects. The

residual stress relaxation during cyclic loading

Tpeen = 20 C

rs, MPa

200

N=0

400

N=1

N = 10

N = 102

N = 103

600

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

(a)

200

Tpeen = 290 C

a,s = 1000 MPa

rs, MPa

200

N=1

N = 10

400

N = 102

600

N = 103

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

(b)

Macro residual stress in longitudinal direction versus distance from surface for quenched and tempered AISI

4140 (450 C, or 840 F, for 2 h) after different numbers of cycles at the surface stress amplitude of 1000 MPa

(145 ksi). (a) Tpeen 20 C (70 F). (b) Tpeen 290 C (555 F)

Fig. 24

and the initial residual stress value rrs

s at the surface enables the estimation of the compressive

yield strength of the peened surface region Re(c),s

with Eq 20 if the property |rlscrit(c)| is substituted

by rs,crit according to:

(Eq 24)

Applying Eq 24 for the shot-peened variants investigated, clear differences of Re(c),s can be recognized (see Table 3) (Ref 48). For both variants, however, the resulting compressive yield

strength at the surface is smaller than that of the

core region, which is qualitatively in agreement

with the results in Table 2. This is also due to

the Bauschinger effect. However, the

Bauschinger-induced work softening is much

smaller for the warm-peened samples than for

the samples peened at room temperature, due to

the dynamic and static strain aging effects that

occur during and after warm peening.

By using the critical loading amplitude ra,crit

as well as the residual stresses at the surface

rrss,N1 after N 1, and the assumption that the

residual stress state at the surface is still almost

axisymmetric, the cyclic yield strength at the

surface Rcycl

e,s can be calculated similarly to Eq 24

by:

2

rs

2

rs

Rcycl

e,s (ra,crit) (rs,N1) rs,N1 ra,crit

(Eq 25)

N=0

800

|Re(c),s|

2

rs

(rs,crit)2 (rrs

s ) rs rs,crit

800

reduction of residual stresses with increasing

logarithm of the number of cycles for 1 N

104 according to Eq 21. Therefore, the residual

stress values at the surface after 104 cycles from

Fig. 22 can be taken as a measure of the degree

of cyclic residual stress relaxation. These values

are shown also in Fig. 28 for both shot-peened

variants versus the absolute value of the applied

loading stress amplitude at the surface ra,s. The

linear fit of these points intersects the line fitted

through the residual stresses measured at N 1

at the critical loading amplitude ra,crit, which is

a measure for the onset of cyclic residual stress

relaxation. This procedure leads to ra,crit 514

MPa (74 ksi) for the conventionally peened condition and ra,crit 714 MPa (103 ksi) for the

warm-peened condition (see Fig. 28).

can be seen that Rcycl

e,s for the conventionally

peened samples is much smaller than for the

warm-peened state. Moreover, the ratio of the

calculated cyclic yield strength at the surface and

that for the core region Rcycl

1090 MPa, or

e

160 ksi) found in push-pull tests (Ref 28) indicates that the cyclic work softening that is typical

for quenched and tempered steels does not apcycl

pear in the warm-peened variant Rcycl

e,s /Re

1.07). The increase of Rcycl

at

the

warm-peened

e,s

Residual stress relaxation when N l can be

attributed to cyclic plastic deformation. In this

case, the dislocation arrangements present in the

as-received materials are rearranged to configu-

Summary

The stability of the mechanically treated surface state of components during the application

of thermal and/or mechanical energy is desirable

800

P = 50%

700

Tpeen = 290 C

Tpeen = 200 C

600

Tpeen = 20 C

500

Ground

400

104

105

106

Nf

640

590

530

Rfb, MPa

the given stress amplitude. Especially for steels

with high dislocation densities (cold-worked as

well as hardened or quenched and tempered conditions), dislocation rearrangements are frequently associated with work-softening processes in the microcrack-free stage of fatigue.

Rearrangement of dislocations from configurations typical of the machined or heat-treated

condition to that of the fatigued state is fundamental to the relaxation of macro residual

stresses. Residual elastic strains associated with

the macro residual stresses are once again converted into microplastic strains. If fatigue softening processes are involved, relaxation of micro

residual stresses can also be expected. In the case

of machined and heat-treated materials, this

manifests itself in decreasing hardness and

sharper x-ray interference lines (Ref 16, 17). Differences in the residual stress relaxation behavior

of cold-worked, machined, and heat-treated materials under cyclic load can be attributed to different dislocation arrangements and densities.

The closer the similarity to the dislocation configuration characteristic of fatigue, the less the

extent of dislocation rearrangement and hence of

residual stress relaxation.

a,s, MPa

strong pinning of dislocations by clouds of carbon atoms and extremely fine carbides that exist

due to strain aging effects. This pinning is so

strong that even at highest loading amplitudes,

ra,s, the pinned dislocations cannot move. Therefore, dislocations that were newly generated during cyclic loading cause the macro residual stress

relaxation. The increase of the half-widths for

these variants during alternating bending tests

(see Fig. 23) is a further hint for this assumption.

With these relationships, the effects of the different peening treatments on fatigue strength can

be interpreted. Compared to the conventionally

peened condition, the significantly higher fatigue

strength of the warm-peened variants cannot be

explained only by the slightly higher compressive residual stresses. Moreover, their higher stability seems to be the main reason for this behavior.

440

107

108

quenched and tempered AISI 4140 (450 C, or

840 F, for 2 h). Samples shot-peened at different temperatures are compared with the ground condition.

Fig. 26

Cross section

area

Surface

4.0

Tpeen = 20 C

a,s = 1000 MPa

N=0

N=1

N = 10

3.5

srs

HW, 2

N = 102

Core

N = 103

rs

rs

c

3.0

2.5

Surface

2.0

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

residual stress for residual stress relaxation due

Fig. 27

to bending

(a)

4.0

Tpeen = 290 C

a,s = 1000 MPa

100

N = 1 N = 104

Tpeen = 20 C

N = 102

200

Tpeen = 290 C

N = 103

300

l,s

3.0

rs , MPa

N=1

N = 10

3.5

HW, 2

N=0

2.5

400

500

0.05

0.10

0.15

0.20

0.25

0.30

0.35 0.40

(b)

Half-width versus distance from surface of a quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h)

after different numbers of cycles at a surface stress amplitude of 1000 MPa (145 ksi). (a) Tpeen 20 C (70 F).

(b) Tpeen 290 C (555 F)

Fig. 25

s,crit

600

700

200

2.0

a,crit

s,crit

a,crit

400

600

800

|s| or a,s, MPa

1000

Macro residual stress at the surface in longitudinal direction versus absolute value of surface

stress or surface stress amplitude of quenched and tempered AISI 4140 (450 C, or 840 F, for 2 h) after alternating

bending to N 1 and N 104 in conventionally shotpeened and warm-peened conditions

Fig. 28

Table 3 Quasi-static surface yield strength in compression Re(c),s of quenched and tempered

AISI 4140 in conventionally peened and warm-peened conditions

Shot peening

temperature

C

20

290

rrs

s

rs,crit

Re(c),s

F

MPa

ksi

MPa

ksi

MPa

ksi

Re(c),s /Re(c)

70

555

310

500

45

75

600

660

90

95

801

1008

116

146

0.60

0.78

e,s of quenched and tempered AISI 4140 in

conventionally peened and warm-peened conditions

Shot peening

temperature

C

20

290

rs,rs N1

ra,crit

cycl

Re,s

F

MPa

ksi

MPa

ksi

MPa

ksi

cycl

Rcycl

e,s /Re

70

555

514

714

74

103

520

620

75

90

895

1156

130

168

0.82

1.07

properties. In this connection, residual stress relaxation behavior is the most important aspect to

be considered. Macro and micro residual stresses

can be reduced or completely relaxed either by

heat treatment or under unidirectional or cyclic

mechanical loading.

Elastic residual strains related to macro residual stresses can be converted into microplastic

strains by suitable deformation processesfor

example, by dislocation slip or creep. The onset

of residual stress relaxation is determined by the

critical applied loading stress, rcrit, with the

meaning of a resistance against relaxation of residual stresses, which is accompanied by the beginning of plastic deformation. In the case of

thermal treatment (rcrit 0), unidirectional deformation (rcrit rlscrit(c) and rs,crit, respectively),

as well as cyclic loading (rcrit ra,crit), residual

stress relaxation occurs when under the assumption of a predominantly uniaxial stress state this

relationship is valid:

|rls| |rcrit| |Ri| |rrs|

(Eq 26)

Rce for thermal treatment, Ri Re(c),s and Re(t),s,

respectively, for unidirectional loading, and Ri

Rcycl

e,s for cyclic loading. Stability of the residual stress state exists, and residual stress relaxation does not occur, if this condition is fulfilled:

|rls| |rcrit| |Ri| |rrs|

sufficient stable obstacles to dislocation movement, which increases rcrit at the loads given. For

T 0.4 Tm (Tm melting temperature in K)

these could, for example, be grain and phase

boundaries, finely dispersed incoherent particles,

coarse secondary phases, and, as long as no diffusion can occur, solute alloying atoms and certain dislocation arrangements. The onset and extent of residual stress relaxation is influenced in

a complex manner by the combined effects of

heat treatment, unidirectional and cyclic stress,

and multiaxial applied and/or residual stress

states. The kinetic of residual stress relaxation is

essentially determined by the difference |rls|

|rrs| |Ri|. The greater the applied loading stress

isthat is, the higher the temperature, the

longer the time, and the greater the magnitude

and amplitude of the applied stress, plastic deformation, or number of cyclesthe greater the

increase in the rate of residual stress relaxation.

The rate often increases as the absolute value of

the residual stress increases. The fewer the stable

obstacles to dislocation slip, the faster the rate.

(Eq 27)

have a value that does not vary with time (e.g.,

thermal residual stress relaxation), have a steadily increasing value (e.g., unidirectional deformation), or have a value varying periodically

with time (e.g., cyclic deformation). Since, according to Eq 26, the onset of residual stress relaxation depends on the resistance of the material

rcrit as well as on the applied loading stress r,

all the parameters characterizing the state of the

material as well as temperature and time or frequency are of importance. The onset of residual

stress relaxation is delayed by the presence of

REFERENCES

1. E. Macherauch and V. Hauk, Ed., Proc. Int.

Conf. Residual Stresses 1, DGM-Informationsgesellschaft, Oberursel,1987

2. B. Beck, S. Denis, and A. Simon, Ed., Proc.

Int. Conf. Residual Stresses 2, Elsevier Applied Science, London, 1989

3. H. Fujiwara, T. Abe, and T. Tanaka, Ed.,

Proc. Int. Conf. Residual Stresses 3, Elsevier Applied Science, London, 1992

4. M. James, Ed., Proc. Int. Conf. Residual

Stresses 4, Society for Experimental Mechanics, 1994

5. T. Ericsson, M. Oden, and A. Andersson,

Ed., Proc. Int. Conf. Residual Stresses 5, Institute of Technology, Linkopings Universitet, 1998

6. A. Niku-Lari, Ed., Proc. Int. Conf. Shot

Peening 1, Pergamon Press, Oxford, 1982

7. H.O. Fuchs, Ed., Proc. Int. Conf. Shot Peen-

8. H. Wohlfahrt, R. Kopp, and O. Vohringer,

Ed., Proc. Int. Conf. Shot Peening 3, DGMInformationsgesellschaft, Oberursel, 1987

9. K. Iida, Ed., Proc. Int. Conf. Shot Peening

4, Japan Society for Precision Engineering,

Tokyo, 1990

10. D. Kirk, Ed., Proc. Int. Conf. Shot Peening

5, Oxford, 1993

11. J. Champaigne, Ed., Proc. Int. Conf. Shot

Peening 6, San Francisco, 1996

12. A. Nakonieczny, Ed., Proc. Int. Conf. Shot

Peening 7, Institute of Precision Mechanics,

Warsaw, 1999

13. E. Kula and V. Weiss, Ed., Residual Stress

and Stress Relaxation, Plenum Publishing

Corp., 1982

14. A. Niku-Lari, Ed., Advances in Surface

Treatments, Vol 4, International Guidebook

on Residual Stresses, Pergamon Press, Oxford, 1987

15. V. Hauk, Ed., Structural and Residual

Stress Analysis by Nondestructive Methods:

Evaluation, Application, Assessment, Elsevier Science B.V., 1997

16. O. Vohringer, in Residual Stresses: GenerationMeasurementValuation, Vol 1, E.

Macherauch and V. Hauk, Ed., DGMInformationsgesellschaft, Oberursel, 1983,

p 4983

17. O. Vohringer, in Advances in Surface Treatments, Vol 4, International Guidebook on

Residual Stresses, A. Niku-Lari, Ed., Pergamon Press, Oxford, 1987, p 367396

18. M.R. James, in Advances in Surface Treatments, Vol 4, International Guidebook on

Residual Stresses, A. Niku-Lari, Ed., Pergamon Press, Oxford, 1987, p 349365

19. E. Macherauch and P. Muller, Z. Angew.

Phys., Vol 7, 1961, p 305312

20. B.E. Warren, Prog. Met. Phys., Vol 8, 1961,

p 147202

21. B.E. Delhez, Th.H. Keijser, and E.J. Mittemeijer, Fresenius Z. Anal. Chem., Vol 312,

1982, p 116

22. O. Vohringer and E. Macherauch, Hart.Tech. Mitt., Vol 32, 1977, p 153168

23. B. Hoffmann, O. Vohringer, and E. Macherauch, Mater. Sci. Eng. A, Vol 234236,

1997, p 707710

24. B. Hoffmann and O. Vohringer, in Residual

Stresses and Distortion due to Thermal Effect, D. Aurich, K.-H. Kloos, G. Lange, and

E. Macherauch, Ed., DFG-Research-Report,

Wiley-VCH Verlag, Weinheim, 1999, p

160174

25. J. Hoffmann, B. Scholtes, O. Vohringer, and

E. Macherauch, in Proc. Int. Conf. Shot

Peening 3, H. Wohlfahrt, R. Kopp, and O.

Vohringer,

Ed.,

DGM-Informationsgesellschaft, Oberursel, 1987, p 239246

26. V. Schulze, F. Burgahn, O. Vohringer, and

E. Macherauch, Materialwiss. Werkstofftech., Vol 24, 1993, p 258267

27. V. Schulze, O. Vohringer, and E. Macher-

28.

29.

30.

31.

32.

33.

34.

Kirk, Ed., Oxford, 1993, p 264273

H. Holzapfel, V. Schulze, and O. Vohringer,

in Shot Peening: Compilation of Conferences, J. Flavenot and B. Scholtes, Ed., Centre Technique des Industries Mecaniques

(CETIM), 1995, p 1530

T. Hirsch, O. Vohringer, and E. Macherauch, Hart.-Tech. Mitt., Vol 38, 1983, p

229232

J. Hoffmann, Dr.-Ing. thesis, University

Karlsruhe (TH), 1985

J. Hoffmann, B. Scholtes, O. Vohringer, and

E. Macherauch, in Proc. Int. Conf. Residual

Stresses 1, E. Macherauch and V. Hauk,

Ed., DGM-Informationsgesellschaft, Oberursel, 1987, p 695702

O. Vohringer, in Proc. Int. Conf. Shot Peening 3, H. Wohlfahrt, R. Kopp, and O. Vohringer, Ed., DGM-Informationsgesellschaft,

Oberursel, 1987, p 185204

M.W. Fine, Introduction to Phase Transformations in Condensed Systems, Macmillan,

1965

M. Roth, in Proc. Int. Conf. Residual

Stresses 1, E. Macherauch and V. Hauk,

35.

36.

37.

38.

39.

40.

41.

M. Roth, Z. Werkstofftech., Vol 18, 1987, p

225228

E. Hanke, Fortsch.-Ber. VDI-Z., Vol 5 (No.

9), 1969

H. Hickel, H. Wohlfahrt, and E. Macherauch, DVS-Bericht 26, Strahltech. VI, 1973,

p 5356

H. Hanagarth, O. Vohringer, and E. Macherauch, in Proc. Int. Conf. Shot Peening 4,

K. Iida, Ed., Japan Society for Precision Engineering, Tokyo, 1990, p 327336

B. Eigenmann, V. Schulze, and O. Vohringer, in Proc. Int. Conf. Residual Stresses 4,

M. James, Ed., Society for Experimental

Mechanics, 1994, p 579588

H. Holzapfel, V. Schulze, O. Vohringer, and

E. Macherauch, in Proc. Int. Conf. Residual

Stresses 5, T. Ericsson, M. Oden, and A.

Andersson, Ed., Institute of Technology,

Linkopings Universitet, 1998, p 115120

H. Holzapfel, V. Schulze, O. Vohringer, and

E. Macherauch, in Proc. Int. Conf. Shot

Peening 6, J. Champaigne, Ed., San Francisco, 1996, p 385396

Relaxation, E. Kula and V. Weiss, Ed., Plenum Publishing Corp., 1982, p 19151921

43. B. Scholtes and O. Vohringer, Hart.-Tech.

Mitt., Vol 41, 1986, p 347354

44. H. Holzapfel, Dr.-Ing. thesis, University

Karlsruhe (TH), 1994

45. B. Hoffmann, O. Vohringer, and E. Macherauch, Mater. Sci. Eng. A, Vol 319321,

2001, p 299303

46. V. Schulze, K.-H. Lang, O. Vohringer, and

E. Macherauch, in Proc. Int. Conf. Shot

Peening 6, J. Champaigne, Ed., San Francisco, 1996, p 413423

47. A. Wick, V. Schulze, and O. Vohringer, in

Proc. Int. Conf. Shot Peening 7, A. Nakonieczny, Ed., Institute of Precision Mechanics, Warsaw, 1999, p 102109

48. A. Wick, V. Schulze, and O. Vohringer,

Mater. Sci. Eng. A, Vol 293, 2000, p 191

197

49. D. Dengel, Z. Werkstofftech., Vol 8, 1975,

p 253261

50. H. Holzapfel, V. Schulze, O. Vohringer, and

E. Macherauch, Mater. Sci. Eng. A, Vol 248,

1998, p 918

G. Totten, M. Howes, T. Inoue, editors, p70-86

DOI: 10.1361/hrsd2002p070

All rights reserved.

www.asminternational.org

Embrittlement and Stress Corrosion

Cracking

A.I. Kovalev, D.L. Wainstein, V.P. Mishina, Surface Phenomena Researches Group, Moscow

V.V. Zabilsky, Physical Technical Institute, Russia

METALLURGISTS AND METALS SCIENTISTS traditionally consider hydrogen an undesirable impurity in metals and alloys, associating its presence with the origination of

floccules, porosity, and corrosion. Due to specific behavior of the element in steel, it makes

worse technological and exploitation properties

of the material. Hydrogen embrittlement (HE),

as well as other types of brittle fracture, of metals

result from nucleation and development of microcracks caused by internal stresses. Peculiarities of HE are involved with steel physical properties and mechanisms of its interaction with

hydrogen. Considerable hydrogen solubility difference depending on external factors, a high diffusion mobility, and the capability to form molecules on interfaces that induces a high internal

hydrostatic pressure all promote various types of

brittleness.

Hydrogen in Steel

Sources of Hydrogen Penetration in Steel.

Water-containing raw charge materials and fur-

nace gases involved in the open-hearth (Martins) process are the main sources of hydrogen

penetration in steel (Ref 1). Industrial cast usually contains H2 of not less than 1.5 to 2.0 cm3 /

100 g. Use of vacuum evaporation in a ladle or

at casting is the only way to decrease the content

of hydrogen in steel down to a concentration not

affecting the mechanical properties, of the steel.

Penetration of hydrogen into a carbonic and

alloyed steel was observed in a hydrogen atmosphere at a high pressure (Ref 2, 3). Metal welding is one of the most dangerous technological

processes inducing penetration of hydrogen in

steel. During the process, hydrogen, contained in

electrode coatings as organic compounds and

water molecule, is dissociating, then being in the

atomic state it dissolves in a molten metal and

partly diffuses into neighboring zones of the basic material. Various electrochemical processes,

etching in acids, and interaction with liquid

agents containing hydro-sulfide also promote to

steel saturation with hydrogen. Therefore, hydrogen is an inevitable impurity in steel pro-

T, C

102

600

500

K K0 p1 / 2 exp(D

H / RT)

lg H 1376/T 0.665

400

(Eq 1)

of solubility in standard conditions, R is gas constant, T is temperature, and DH is endothermal

heat of solution. Those of hydrogen in pure iron,

calculated by W. Geller and T.H. Sun (Ref 4)

and then experimentally refined (Ref 57), account for 27230 J/ 0.5 mol H2 (6500 cal/0.5 mol

H2). Solubility of hydrogen in iron at various

temperatures and atmospheric pressures (Ref 5

8) is shown in Fig. 1.

From recent data, solubility of hydrogen in a

metal also depends on its purity, microstructure,

grain size, and a character of distribution of dislocations (Ref 915). The solubility curve equation for -Fe at H pressure of 0.1 MPa (0.015

ksi) is:

(Eq 2)

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

lg H 1411/T 0.468

(Eq 3)

lg H 1820/T 0.112

101

0.4

Fig. 1

Hydrogen Solubility and Diffusion in Steel.

Hydrogen solubility in steel depends on temperature and pressure to suit the following equation:

0.6

0.8

1.0

1.2

1000/T, K1

1.4

1.6

1.8

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

(Eq 4)

cm3 per 100 g of metal, can be brought into a

steel from a gas phase only at high pressures and

temperatures (Ref 16). At standard conditions, hydrogen gradually escapes from steel, its solubility

in steel then being negligible. At temperatures

higher than 200 C (392 F) a real solid solution

of hydrogen in iron becomes unstable (Ref 17).

Thermodynamic data of H diffusion in iron at

high temperatures are well reproducible in various works (Ref 1820) that cannot be told about

the data obtained below 200 C (392 F). Diffusion mobility of hydrogen in iron at low temperatures varies in a wide range and depends on

imperfection of the crystal lattice and the content

of impurities. For example, Ref 2124 there

were obtained values of the diffusion coefficient

at 25 C (77 F) varying from 109 up to

105cm2 /s. When studying diffusion in pure

iron by intensity of hydrogen emission out of

massive samples, it was found that its diffusion

mobility is sharply depressed at lower than 200

C (392 F) (Fig. 2). Above 200 C (392 F) the

diffusion coefficient is the following:

D 1 4 103 exp (3200/ kT) cm2 /s

(Eq 5)

D 0.12 exp (7820/ kT) cm2 /s

(Eq 6)

Just as impurities forming interstitial solid solutions, hydrogen interacts with various crystal

defects, creating nonuniform solid solutions.

Presence of microdeformation near the crystal

imperfections increases hydrogen solubility

there. Therefore, the solubility of hydrogen in

steel rises with an increase in density of defects.

When diffusing inside the crystal, hydrogen atoms are retained by the lattice imperfections, and

its effective diffusion coefficient essentially decreases. This results in sharply diminishing diffusion mobility of hydrogen in iron at low temperatures. Defects retaining hydrogen are known

as traps. Its concentration controls an excessive

hydrogen solubility, compared with the lattice

one (i.e., in an ideal lattice). Moreover, the number of places capturing hydrogen is inversely

proportional to the value of the effective diffusion coefficient.

The role of traps can be played by edge dislocations, boundaries of mosaic blocks or grains,

single pores uniformly distributed in crystals, inclusions of a second phase, and atoms of alloy-

ing elements. The energy of hydrogen interaction with different structural components varies

in a narrow interval (0.10.4 eV per atom). In

particular, the one with dislocations in iron was

evaluated as about 1.6 1020 J (0.1 eV per

atom) (Ref 25).

The Role of Stresses. Hydrogen atoms interact

with any anisotropic strain field, similar to its

interaction with dislocations. If concentration of

hydrogen atoms is insignificant, then distribution

of the dissolved ones in a strain field of a crystal

lattice can be found using:

C C0 exp (U/kT)

(Eq 7)

where C is a concentration of hydrogen in a certain point of the field, C0 is an average concentration, k is the Boltzmann constant, T is the kelvin temperature, and U is the interaction energy

of hydrogen atoms with the strain field in the

place of a defined concentration, C.

Accounting for only hydrostatic components

of the field, the interaction energy, U, makes:

U p dV

(Eq 8)

is, p 13(rxx ryy rzz), and dV is the lattice

volume alteration under the influence of hydrogen.

From these equations it follows that at a constant temperature the following condition is

valid:

ln (C/Co) A 13(rxx ryy rzz)

(Eq 9)

where A kT/dV.

According to (Eq 9), hydrogen in the crystal

lattice is redistributed under stress influence.

These deductions were confirmed with experiments (Ref 26, 27). It is settled that the effective

diffusion coefficient of hydrogen essentially decreases, and its solubility increases under loading the metal in a macro-elastic interval (below

yield strength). Thus, stresses promote nucleation of new reversible hydrogen traps, disappearing after elimination of an external stress.

Similar sites arise near the crack tip, in a several

where hydrogen dissolved in the metal lattice

diffuses and its concentration in steel at equilibrium conditions attains maximal values as:

C rmax C0 exp (1.9 103 ry)

of approximately 1725 MPa (250 ksi) the equilibrium concentration of hydrogen in the zone of

maximal tensile stresses exceeds its content in

volume by 26 times (Ref 28) (Fig. 3).

At absence of a crack or any other concentrator, the stresses are localized in front of a dislocation pile-up, where considerable enrichment

with hydrogen also can be observed. Both factors, concentration of tensile stresses and dissolved hydrogen, initiate nucleation of cracks

and promote its propagation.

Hydrogen Effect on Structure and Phase

Transformations in Steel. The basic consequence of hydrogen presence in iron and steels

at solidification is the development of porosity,

due to a gas-eutectic transformation (Ref 29)

similar to the well-investigated eutectic one. Total amount of pores is proportional to those of

dissolved hydrogen and is inversely proportional

to the pressure of crystallization. Hydrogen in

steel stabilizes austenite and decreases the critical point A3, the temperature of martensite transformation, and the critical rate of cooling for

quenching (Ref 29). This is the cause of martensite formation even in a carbon-free iron at standard rates of cooling. Unlike a carbonic martensite, at low temperatures of tempering there

arises a hydrogen-free martensite and a gaseous

hydrogen escaping from a sample or accumulating in micropores, creating high pressure and microcracks like floccules (Ref 29).

States of Hydrogen in Steel. Hydrogen absorbed by metal can be present in various states:

An interstitial solid solution

A segregating one on imperfections of the

crystalline structure

Adsorbing one on the surface of micropores

T, C

107

800

400

200

100

25

D, cm2/s

107

107

107

107

0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6

1000/T, K1

Fig. 2

Source: Ref 1825

H

H2

HH

H2

H2

H

HH

HH

H

HH

H

H

H2

H2

H2

H

H

Fig. 3

(Eq 10)

H

H

H

In a molecular form, accumulating in micro-

pores

Forming hydrides

Interacting with alloying elements and second

phases

When hydrogen solutes in transition metals,

its s-electrons transfer to a vacant d-shell of a

metal with the lower energy level. The free proton of hydrogen is partly screened by electrons,

which as electronic gas can be involved in its

moving in the metal lattice (Ref 28). Hydrogen

reduces a shear modulus of -iron; at 100 K temperature, this is approximately an 8% decrease

per 1% of hydrogen atoms (Ref 30). At a room

temperature and standard atmosphere pressure

the solubility of hydrogen in steel is insignificant; its supersaturated solid solution is unstable

and decomposes over time. One part of hydrogen

dissolves into the atmosphere and another is

caught by crystalline lattice imperfections. It is

necessary to distinguish two types of defects active in relation to hydrogenpores (or collectors) and traps. Collectors represent volumes

where hydrogen molization is observed and a

considerable hydrostatic pressure develops.

Traps represent defects of smaller sizes than collectors, such as vacancies, substitution and interstitial atoms, edge dislocations, its pile-ups

and thresholds, high-angle boundaries, and fields

of a volume tension (i.e., near the crack tip).

There is a principle difference between traps and

collectors. Transition of hydrogen from a crystal

lattice into traps is reversible, while transition

into collectors is accompanied by formation of

H2 molecule and is nonreversible. The considered features depend on a steel structure.

Ferrite-Pearlite Steels. Nonmetallic inclusions, such as manganese sulfides, are traps of

hydrogen in these soft steels. The maximum solubility of hydrogen in such steel depends not

only on the content but also on the shape of manganese-sulfide particles. The rolled-out prolonged ones increase hydrogen solubility and decrease its effective diffusion coefficient in

the inclusions.

Globularization of the sulfides decreases hydrogen solubility and increases its effective diffusion coefficient. Sulfide-matrix boundaries are

the sites of a preferential accumulation of dissolved hydrogen, essentially weakening the cohesive forces and promoting disruption of the

interface continuity.

Martensitic Steels. High interior residual

stresses inducing an elastic bend of the martensite crystals and localizing in places of its junction with boundaries of primary austenitic grains

are characteristic for these steels having the

structure of an untempered martensite. In addition, martensitic steels feature a high density of

crystalline structure defects, which are boundaries of grains and twins, single dislocations. All

these elements of structure affect solubility and

diffusion of hydrogen in steel. Dispersed cementite particles observed after tempering also

act as traps for hydrogen. In maraging steels such

a role is executed by intermetallic-matrix interfaces and elastic deformation fields around the

inclusions.

Types of Hydrogen Embrittlement. In general, hydrogen affects metals in two ways:

Physical action at low temperatures when

components of an alloy on the surface and in

the bulk is observed.

The attempt to enumerate all the phenomena

tied with harmful influence of hydrogen in metals was made in 1972 (Ref 31). Reference 28

specifies the following results.

Hydrogen corrosion: Develops in carbonic

Fig. 4

A.I. Kovalev

in a high-pressure hydrogen at high temperatures. The cause of this brittleness is hydrogen-carbon interaction accompanied by steel

surface decarbonization and methane formation (Ref 32).

Primary gas porosity: Develops due to hydrogen precipitation in a molecular form in a

smelt or in the front of crystallization (Ref 33)

Secondary porosity: Develops at decomposition of a supersaturated solid solution with

formation of small submicroscopic pores

filled by hydrogen (Ref 33, 34)

Decreasing plasticity under tension: Observed at low-rate deformation of steel containing hydrogen (Ref 33, 34)

Delayed fracture: Steel cracking under constant loading below the yield limit

Floccules or disruption cracks: Defects

mainly found in large forged pieces. Floccules are the result of gas hydrogen precipitation in pores and micro flaws during steel

crystallization and cooling of the slab or the

forged piece (Ref 35, 36).

Blisters or bubbles: Develop at electrolysis,

metal exposing in a water vapor or in corrosion agents containing H2S (Ref 33, 35)

with precipitation of an atomic hydrogen during corrosion reactions, its adsorption on the

surface and dissolution in metal with development of hydrogen embrittlement

Main Aspects of the Hydrogen Embrittlement Theory. Definite kinds of HE involve two

basic aspectsthe mechanism of hydrogen

transfer to a fracture place, and 2-H influence on

the steel fracture toughness.

Usually hydrogen is transported by moving

dislocations or diffuses (itself) through the crystal lattice under inhomogeneous fields of elastic

stresses (Ref 3740). Being in steel, hydrogen

decreases the fracture work by influencing mechanisms of plastic deformation and reducing the

energy of disruption of interatomic bonds near

the crack tip.

In ferrite-pearlite steels, HE is accompanied

by nonreversible structural damages in the form

of micropores containing molecular hydrogen

under high pressure. Those conditions, however,

are not enough for steel fracture that occurs due

to hydrogen migration together with moving dislocations into the zone of deformation near the

tip of a developing crack. Fracture of soft steels

is always accompanied by a considerable plastic

deformation essentially affected by hydrogen

(Ref 37) facilitating the beginning of plastic deformation of ferrite due to adsorption effects.

However, hydrogen increases the hardening coefficient at higher deformation degrees (Ref 41).

As a result, it localizes the plastic deformation

in the zone of fracture, essentially reducing the

plasticity of steel and diminishing the total fracture energy.

In steels with martensitic structure, hydrogen

migrates through the field of elastic deformations into the zone of three-axial stresses before

the front of a crack (Ref 42, 43). When the content of external and dissolved hydrogen reaches

a critical value, the crack immediately propagates by the mechanism of disruption of interatomic bonds (Ref 44). In this case, the basic

mechanism of HE is tied with breaking weakened bonds in grain boundaries and decreasing

the plastic deformation energy in local sites of

ductile fracture.

Forms of HE of steels are numerous, but only

in some cases (floccules, delayed fracture, and

stress corrosion) the role of macro and micro

stresses influencing hydrogen behavior is clearly

defined.

Floccules

Floccules are interior cracks (flaws) observed

in steel: in the central zone of deformed preforms and, as an exception, in some sites of

molded wares. In polished (not etched) cross

macro-grinds these defects look like hairline

cracks (Fig. 4) opened along a plastic flow direction of forged pieces and rolled products.

Only when considerable axial stresses are present in a pre-form, the orientation of floccules al-

ters and can be perpendicular to the axis. By

macroscopic investigation of fracture surfaces of

steel affected by floccules, the light brilliant

spots of an oval shape are observed. Fractography research of those sites, using an electronic

microscope, allows the detection of the micro

quasi-chips involving a slightly expressed riverpattern picture, pores, and plastically opened

hairline cracks (Ref 44). These defects can be

found in all steels, but mostly they are observed

in chromic, chrome-nickel, chrome-nickel-tungsten, and chrome-manganese steels. Floccules

originate in forged pieces or rolled stock of large

section some distance apart on the surface under

accelerated cooling at temperatures below 150 to

200 C (302 to 392 F), sometimes at 300 C

(572 F), but more often at room temperature.

One of the main reasons of floccules formation

is a dissolved hydrogen that does not have

enough time to escape large forged pieces at

temperatures below 200 C (392 F) and diffuses

into neighboring pores, creating molecules and

considerable gas pressures (Ref 45, 46). However, hydrogen present in metal is a necessary,

but not the only, condition for origination of the

defects under consideration. Internal tensile

stresses developing under plastic deformation,

non-uniform cross section cooling of steel wares,

and structural transformations stimulate formation of floccules (Ref 4749). Decrease of rupture resistance in sites enriched by nonmetallic

inclusions or in liquation zones promotes this

kind of brittleness. The most favorable conditions occur when the flocculation temperatures

coincide with an interval of development of martensite transformation. Dissolving in austenite in

a higher quantity than in martensite, hydrogen

during martensite transformation moves from a

solid solution into pores. Resultant high hydrostatic pressures summarize with stresses caused

by volume alteration under c phase transformation. For this reason, a quenched-in-air

chrome-nickel steel is very susceptible to flocculation. This is the result of total action of the

molecular hydrogen pressure in pores and the

field of tensile stresses existing in the metal (Ref

50). Floccules are developing during the time,

and the process quite often does not terminate

when cooling finishing and continues for many

hours at room temperature.

At high contents of hydrogen and temperatures approximately 400 to 600 C (752 to 1112

F), formation of H2 molecules in pores results

in intergranular brittle failure of steel, without

participation of external (relative to the pores)

stresses. For fracture at lower hydrogen concentrations and temperatures below 200 C (392 F),

the combined effect of tensile stresses and gas

pressure in pores is necessary. In that case, a

brittle transgranular failure develops.

Processes of formation of floccules are nonreversible. An extremely slow cooling after deforming a preform made of steel susceptible to

flocculation prevents its development. The same

effect was observed at a prolonged isothermal

annealing at heightened or high temperatures.

However, heating in an atmosphere containing

susceptible to flocculation. That tendency of

steel can develop at welding or cementation

when wares are quenched directly after treatment

heating. Floccules are not observed in a number

of superalloyed steels such as ferritic, semiferritic, and austenitic. High-speed steels are not

susceptible to flocculation, in spite of the fact

that very high internal stresses at accelerated

cooling develop there. Influence of alloying elements on steel susceptibility to flocculation is

tied with action on austenite transformation under cooling, anisotropy of properties, and a quantity of generated hard inclusions. At Ac3 temperature decreasing and Ac3-Ar3 hysteresis

increasing the diffusion of hydrogen and its escaping off steel pre-forms becomes slower. This

explains a high susceptibility to flocculation of

a chrome-nickel steel after its alloying with molybdenum or tungsten. Manganese and silicon,

promoting anisotropy of mechanical characteristics of a rolled stock, favor its susceptibility to

flocculation. Increasing concentration of carbon

in a hydrogen-containing steel promotes formation of methane and floccules.

Preventing Flocculation in Steel. The main

condition of preventing floccules in steel is decreasing its supersaturation with hydrogen and

elimination of H2 molecule formation at temperatures below 200 to 300 C. Also, it is necessary to minimize the anisotropy of mechanical

properties of deformed pre-forms. Melting and

casting of steel in a vacuum is the basic and the

most effective way to eliminate of floccules.

Transportation to forging and rolling of ingots

in a hot state, being in common practice, is undesirable for a steel susceptible to flocculation,

because a high quantity of hydrogen remains.

Special heat treatment schemes preventing flocculation are available for large pre-forms made

of steels of various classes.

Delayed fracture, or the real hydrogen embrittlement, is an evolving-in-time process of

cracking (fracture) materials exposed to hydrogen and static stresses. This type of hydrogen

embrittlement is the greatest practical problem,

especially for high-strength steels, as far as the

initiation and the growth of crack with hydrogen

present are observed at stresses much less than

the yield strength of steel. The crack growth up

to the critical value with hydrogen present begins

accordingly at the threshold value of the stress

intensity coefficient, K1th, which has a much

lower value than the fracture toughness in the air

or inert medium, K1C. Thus, the danger of the

delayed fracture process exists in that the fracture of constructions occurs under loads that are

entirely undangerous under normal operating

conditions (i.e., in the absence of hydrogen). Another important point is that the process of delayed fracture develops most intensively at temperatures close to room temperature, at which the

overwhelming majority of materials and products are used.

The main factors determining the kinetic features of delayed fracture are the chemical composition and the structure of steel, a hydrogen

content, the presence of stress concentrators, the

test temperature, the amount and the distribution

of impurities and nonmetallic inclusions, and the

presence of inherent stresses (Ref 29). In this

case the delayed fracture can develop both under

the effect of external hydrogeneous media and

inherent hydrogen having entered the metal as

the result of melting, welding, etching, anticorrosive electroplating, and other production

operations. An example is in the formation of

so-called cold cracks in the heat-affected zone

while welding alloyed steels (Ref 51). The increased hydrogen content in the near-weld zone

is the result of hydrogen diffusion from the melt

where it enters due to dissociation of moisture

in electrode coating (welding flux) (Ref 52, 53).

The effect of the steel-strength level on the

resistance to delayed fracture has been supported

by many studies. It has been shown that at the

low-strength level (up to 750 MPa, or 110 ksi)

the influence of hydrogenation is comparatively

moderate. At strength more than 1000 MPa (145

ksi), the resistance to delayed fracture reduces so

that the value of the threshold stress, rth, decreases to 200 MPa (29 ksi) (Ref 53). It has

been found also that as to the high-strength

steels, especially in the presence of stress concentrators, delayed fracture can develop even

with the low-hydrogen content not exceeding approximately 10 4 at.% (1.1 cm3 per 100 g of

metal) (Ref 1). Numerous examples of delayed

fracture of products fabricated from highstrength alloyed steels (shafts made of Fe-0.45C2Ni-Cr-Mo-V steel, axles made of Fe-0.2C-2Cr4Ni steel in the carburized state, etc.) are

presented in a monograph (Ref 54). The presence

of internal hydrogen in association with high residual stresses with a value reaching 1000 MPa,

(145 ksi) is the most commonly occurring cause

for delayed fracture in the case under consideration.

A great deal of publications in which various

methods of hydrogenation are presented are devoted to the study of delayed fracture of steels.

A main body of experimental data has been obtained with the use of the cathode polarization

method. Therefore, in analysis of the influence

of chemical composition and structure, it is necessary to take into consideration that, because of

electrolytic hydrogenation of large-size samples,

the occurrence of nonuniform distribution of hydrogen over the section can give rise to the largescale effect (Ref 55). Hydrogenation from the

gas phase under conditions of elevated temperatures (above 473 K) may cause irreversible damages in the form of bulgings and pores appearing

because of high pressure during the formation of

inner molecular hydrogen or methane. Additionally, in this case the probable surface decarburization can also have an effect on the initiation

and growth kinetics of cracks (Ref 56). From a

practical standpoint, on the contrary, these im-

portant experiments make it possible to simulate

service conditions for the purpose of developing

adequate measures to eliminate delayed fracture.

Main Criteria of Delayed Fracture of

Steels. In general, steel delayed fracture takes

1

1

Stress,

th

Curves of long-term strength of hydrogenated

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

Fig. 5

Stress, , MPa

900

700

3

4

500

300

1

100

0

1

20

10

30

40

50

Time to failure, , h

60

100

1), steel Fe-0.2C (2, 2), steel Fe-0.45C (3, 3)

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

Fig. 6

IV

III

II

K1c

K1t

Fig. 7

cracking of hydrogenated steel. Source: Ref 57

up to the critical value; and the stage of the unstable, accelerated crack growth. It is common

practice to characterize the sensitivity of steels to

delayed fracture depending on their chemical

composition and structure by so-called long-term

strength dependencies (Fig. 5), which usually are

represented in coordinates: applied stress r, logarithm s, where s is the time up to fracture. Dependencies mentioned are typical for both smooth

samples and samples with the stress concentrator.

The preliminary hydrogenation in constructing

dependencies, s f (r), can be performed by any

of the known techniquescathode polarization,

for example. These dependencies take the analogous character on tests performed just in hydrogeneous media (e.g., on tests in gaseous hydrogen

medium, hydrogen sulfide medium, or in the process of the electrolytic hydrogenation) (Ref 35,

56, 57).

In an analysis of dependencies s f (r), the

following characteristics are used as the main parameters:

The upper critical stress, r1, which corre-

hydrogenated samples

The lower critical (threshold) stress, rth (i.e.,

the stress lower which the delayed fracture

process does not develop)

The time before the crack initiation, si (an incubation period)

The time up to final failure, s, which characterizes the prolonged strength of steel (Ref

35). Thus, a dangerous area of stresses

where the delayed fracture process develops

is an interval between the upper and lower

critical stresses. In this case the value of the

lower critical (threshold) stress, rth, represents a crucial important characteristic,

namely, the prolonged strength limit of steel

under the specific conditions of hydrogen embrittlement.

It is apparent that as the hydrogen content in

steel increases, all of the parameters mentioned

are changed. This especially affects the value of

the threshold stress, rth; the duration of the incubation period, si; and the time up to fracture,

s. From the data given in (Ref 58), for example,

it follows that as the hydrogen content in highstrength steel (1600 MPa, or 232 ksi) increases,

the duration of the incubation period successively decreases from several tens of hours to

several seconds. The time up to fracture in the

limiting case does not exceed 0.1 h. The duration

of the incubation period, si, is influenced by the

preholding under load including holding under

stresses r rth. Reduction of si, observed in

this case is due to the formation of sites with the

increased hydrogen concentration, which then

serve as points of the crack initiation (Ref 35).

Curves of the long-term strength of iron, and

steels Fe-0.2C, Fe-0.45C, and Fe-0.8C are given

in Fig. 6 (Ref 59). Samples have been polished

mechanically and annealed in vacuum. Delayed

fracture tests have been carried out while holding

the samples under load in the conditions of con-

26% aqueous solution H2SO4 (the current density i 10 mA/sm2). Smooth thin plates with

the cross section of 0.8 3 mm (0.03 0.12

in.) that ensures the through saturation of samples with hydrogen have been used as samples.

From the results obtained it is evident that the

higher the hydrogen content in steel is, the more

tangible the decrease of the long-term strength,

and it does not practically take place in the case

of armor-iron and low-carbon steel Fe-0.2C. The

value of the threshold stress for steel Fe-0.45C

is of 150 to 170 MPa (20 to 25 ksi) that is several

times lower than the yield strength of this steel.

The lifetime of samples with the stress concentrator under load is mainly determined by the

time before the crack initiation (i.e., the duration

of the incubation period) (Ref 55). In the study

of crack growth it is worthwhile to construct socalled kinetic diagrams of cracking (i.e., dependencies of the crack growth rate, v, on the stress

intensity factor, K1. To construct kinetic diagrams of cracking, fatigue precracked samples

are used. The character of kinetic diagrams of

cracking is shown schematically in Fig. 7. Some

characteristic sections can be distinguished in

the diagram. At section I, which corresponds to

the threshold value of the stress-intensity factor,

K1th, the crack does not practically grow or

grows very slowly. Therefore, it is assumed that

the value of K1th characterizes the condition of

crack nonpropagation in metal. In section II, the

crack growth rate increases rapidly, whereupon

in region III, the crack growth rate over a wide

range, K1, is retained practically unchanged. Unlike the stage of accelerated, unstable fracture

(section IV), the stage of the slow, stable crack

growth can embrace the greatest part of longevity of a product. It is apparent that the delayed

fracture development at this stage is controlled

by the rate of transferring hydrogen to the region

of the crack top. At section IV, the crack growth

rate increases rapidly again, and at K1 K1C

(plane-strain fracture toughness) the final fracture occurs. The characteristic most sensitive to

the hydrogen content is the value of the threshold

coefficient of stress intensity, K1th. According to

data given in Ref 60, an increase of diffusely

moving hydrogen content in high-strength steel

Fe-15Cr-5Ni-2Cu-Ti from 0.14 104 to 2

104 at.% causes the value of K1th to reduce by

77%.

Besides the main characteristics just discussed, a number of other characteristics also

have been suggested for defining the sensitivity

to delayed fracture. For example, dependencies

of the initial coefficient of stress intensity, K10,

the time up to fracture, s, type are suggested to

be used (Ref 60). In this case long-term strength

curves can be described by the logarithmic relations of the K10 a ln s b type bounded

by the value K1C at the top and the threshold

value of the stress intensity coefficient K1th at the

bottom. In work (Ref 61) it was suggested to

construct the long-term strength curves in coordinates r 1/Zs, but not in conventional coordinates r lg s. In this case dependencies of

the long-term strength s f (r) are transformed

into straight lines convenient for further treatment according to the equation r rth

c(1/ s), where c is a constant. And finally, the

material sensitivity to delayed fracture may be

estimated by its sensitivity to the deformation

rate, namely, by the degree of their plasticity decrease under static loading at the rate of e

104 106 sec1. Such tests, for example,

are in common practice among Japanese researchers in making an estimate of the sensitivity

to the delayed fracture of high-strength maraging

steels (Ref 6264).

Mechanism of Crack Initiation and Its

Growth. It is the opinion of the authors who

studied a phenomenon of delayed fracture, that

in a qualitative sense the mechanism of the crack

initiation when the inner hydrogen embrittlement is as follows. For a crack to be initiated,

the critical concentration of hydrogen C* must

be created in local sites. The hydrogen redistribution in the solid solution is initiated by the

stress field produced by outer load. With the uniform distribution of tensile stresses, the value of

the chemical potential of hydrogen, l0, is the

same in the whole iron-hydrogen system; therefore, the hydrogen redistribution does not take

place. With the nonuniform distribution of

stresses, for example, because of the presence of

a concentrator, the pattern is essentially changed.

At the top of the concentrator (in the region of

maximum tensile stresses), the hydrogen chemical potential, l1, becomes less than that one in

the rest of the regions of the system, l2. The

appearance of the chemical potential gradient (l1

l2) brings about the diffusion of hydrogen into

the tensile stress region that with time gives rise

to the critical concentration of hydrogen C* and

the crack initiation. The critical concentration of

hydrogen C* is directly proportional to the

square root of the time of diffusion (C* C0

qt, where q is the coefficient of proportionality)

and must be so much higher than the level of

applied stresses is lower (Ref 61).

It is apparent that the increase of the hydrogen

content causes a decrease of the duration of the

incubation period, si, which is necessary for creating the critical concentration of hydrogen.

Lowering of the temperature gives the reverse

result (i.e., an increase of si caused by a decrease

of the hydrogen diffusion rate). Since to create

the critical concentration of hydrogen it is essential to localize tensile stresses, the crack initiation depends on the type of samples used. In

samples with no concentrator, the crack initiation occurs, as a rule, in the axial part of a sample

(i.e.. in the zone of maximum three-axis stresses

where the hydrogen concentration is also maximum). In samples with the sharp stress concentrator, the crack initiation takes place adjacent to

its top; in the case of the less sharp and deep

notches, the failure begins at a certain distance

from the notch top (i.e., under the surface) (Ref

35). At the moderate temperatures, hydrogen

gathers mainly on phase boundaries, microdiscontinuities, grain boundaries, piling-up dislocations, and other defects. Therefore, sites of or-

boundaries between ferrite and cementite and

nonmetallic inclusions as wellfor example,

sulfides, which are the most efficient traps for

hydrogen (Ref 56).

The crack growth kinetics upon the internal

hydrogen embrittlement also is controlled by the

hydrogen diffusion rate into the region of maximum tensile stresses ahead of the crack top. It

is suggested that this region is on the boundary

of the plastically deformed zone or at the much

less distance equal to the double opening of the

crack, d. As soon as the hydrogen concentration

becomes critical in the weakest element of a

structure, fracture of this element occurs, following which the crack formed merges into the main

one (Ref 65). The crack propagation is under

progress due to the multiple recurrence of this

process. Therefore, the crack growth can be considered as the process consisting of a sequence

of incubation periods, followed by the restricted

advance of the crack. The process of the crack

stepwise growth is displayed evidently under the

testing conditions at lowered temperatures when

the process of hydrogen diffusion is retarded

(Ref 35). The occurrence of the incubation period of the crack start and the stepwise advance

of separate sections of the crack front is evident

from the results of the registration of the acoustic

emission, which have been obtained from testing

of low-alloy steel with 0.4% C content (Ref 66).

In addition, the registration of the acoustic emission is indicative of the realization of more complicated mechanisms of the crack growth. For

example, when performing work (Ref 67) for

testing the high-strength maraging steel, a

growth of a great quantity of microcracks, 30 to

50 mkm in size, which appeared at 1 to 3 mks

intervals, was revealed.

Contrary to the internal hydrogen embrittlement, in the external hydrogen embrittlement the

crack growth kinetics is controlled mostly by the

following surface phenomena: adsorption, dissociation, and chemosorption of hydrogen on the

metal surface. The following experimentally observed effects serve as a guide for this statement:

The crack growth in the gaseous hydrogen

after applying an external load with no

marked incubation period.

The activation energy of the process at the

stage of the stable crack growth (region III,

Fig. 7) differs noticeably from the activation

energy of the hydrogen diffusion in a solid

solution and is comparable to the activation

energy of surface processes.

When gaseous hydrogen is doped with impurities, which activate (H2S) or suppress

(SO2), the process of the surface interaction

of metal with hydrogen, the crack growth rate

is also increased or reduced, respectively (Ref

57). The location of the failure zone ahead of

the growing crack front depends also on the

type of hydrogen source. In case of the external hydrogen embrittlement, the failure

zone, as a rule, does not reach the point of the

much nearer to the crack top than it is at the

internal hydrogen embrittlement) (Ref 65).

The extent of the embrittlement in the medium

of gaseous hydrogen depends both on the pressure and the chemical composition of a medium,

in part, of the steam content and oxygen, as well.

In the case of high-strength steels a dramatic effect of the hydrogen pressure on the crack

growth rate has been revealed at pressures not

exceeding 1.3 103 MPa (Ref 68). The presence of steams in the hydrogen atmosphere has

a quite profound initiating effect on the delayed

fracture development. Oxygen has the opposite

effect of efficacious retarding of the crack

growth up to the critical value. On data given in

Ref 69, for example, a mere 0.6% of O2 is sufficient for the complete stop of the crack development in the medium with humidified hydrogen

(steel with the yield strength of 1600 MPa, or

232 ksi). The crack stopped by oxygen begins to

grow again only after the complete removal of

oxygen from the medium of tests. It is suggested

that oxygen forms oxides that prevent the hydrogen from penetration into the crack top, and after

the removal of oxygen the reverse process occurs

(i.e., oxides are restored by hydrogen) (Ref 70).

The temperature factor should be set off from

other factors that affect the character of delayed

fracture. According to the data in hand, the maximal sensitivity of steels to delayed fracture takes

place at temperatures close to, room temperature

(Ref 29). But in a number of cases this regularity

is not obeyed. In Ref 71, the temperature dependence of the delayed fracture of the high-strength

steel Fe-15Cr-5Ni-2Cu-Ti containing 5 104

at.% of hydrogen was studied. The threshold

value of the stress intensity coefficient for this

steel is K1th 46 MPa m1 / 2. At 20 C (68 F)

the crack growth occurs at the rate of 2 107

m/s. At 100 C (212 F) the crack growth rate

up to the critical value increases by more than

300 times and reaches 6.6 105 m/s. In this

case the time up to fracture (s) reduces from 500

to 600 min at 20 C (68 F) to s 5 to 6 min

100 C (212 F). At the temperature of 180 C

(356 F) the development of delayed fracture

stops completely. The calculation of the activation energy of the crack growth Q has shown

good agreement between the obtained value

(64.8 kJ/mol) and the activation energy of the

hydrogen embrittlement (60.9 kJ/mol).

Chemical Composition and Structure Influence. The resistance of steels to delayed fracture depends on the carbon content, alloying, and

impurity elements. From the data given in Fig.

6, it is evident that a buildup of the carbon content is attended by the significant reduction in

the resistance to delayed fracture. This effect derives from the fact that with increased carbon

content the strength properties of steel are improved, the level of residual internal microstresses is raised, and the amount and volume of

the carbide phase are increased. The interphase

surface area (i.e., the number of places, or traps,

for the accumulation of hydrogen and the crack

initiation) is correspondingly increased (Ref 29).

The influence of alloying elements is also

quite profound. The effect of noncarbide-formative alloying elements (Mn, Al, Si, Co, Ni) is

mainly connected with their influence on the

properties of matrixthat is, with the change of

its resistance to the viscous, brittle, and grainboundary fracture. These elements (but Ni) decrease the long-term strength of steels subjected

to quenching followed by high tempering (Table

1). Carbide-formative elements (Cr, Mo, Ti, Nb,

V) have an influence mainly through the change

of the amount of the carbide and carbide-nitride

phase, its distribution, and morphology. The

positive influence of molybdenum, niobium, and

vanadium, which bring about to an abrupt increase of the time up to fracture, s, is most pronounced (Table 1). In the case of noncarbideformative elements there is, as a rule, a direct

relationship between the resistance to the hydrogen embrittlement and standard characteristics

of brittle fracturefor example, the critical temperature of brittleness, T50. In the case of carbide-formative elements such an unambiguous

relationship usually is not observed (Ref 72).

As to impurity elements, phosphor and sulfur

have the most negative influence (Table 1). The

negative influence of sulfur is connected with the

formation of manganese sulfides MnS, which,

just as indicated, are the most effective traps for

hydrogen. Manganese sulfides assist in initiating

cracks at more low critical concentration of hydrogen C* in the region of maximal tensile

stresses as evidenced by the results of the reg-

73). In this case, a decrease of sulfur to a degree

less than 0.005 wt% does not protect against the

development of the hydrogen embrittlement.

The form of nonmetallic inclusions also has a

dramatic influence, especially in steels with the

ferrite-pearlite structure. Inclusions of the spherical form are considered to be the safest ones.

The negative role of sulfur can be connected in

that this element has a retarding effect on the

recombination of hydrogen atoms on the steel

surface and so brings about an increase of the

amount of absorbed hydrogen (Ref 74).

Phosphorus and its analogs (Sb, Sn) are the

most dangerous on formation of segregation at

grain boundariesthat is, in the process of thermal treatment, which results in the development

of the temper embrittlement (Ref 7577). In the

absence of segregations the steels fracture in hydrogeneous media connected with the intensive

plastic deformation in the crack top. Fracture

mainly occurs transgranularly (i.e., along a

grain). In this case, in fractograms of the crack

growth up to the critical value, either a viscous

cup fracture or a quasi-spall takes place. During

the formation of segregation another mechanism,

the brittle fracture along grain boundaries, is realized with a consequent abrupt decrease of K1th.

In this case, an intensive reduction of K1th takes

place already at initial stages of the segregation

formation when temper embrittlement has hardly

shown up. It is considered that one of the reasons

Table 1 Influence of alloying elements and impurities on transition temperature, T50; work

of propagation of a ductile crack, A; plasticity loss after hydrogenation, Fw and resistance

to hydrogen embrittlement, s (chromium-molybdenum steel, 0.2% C)

Element (limits of

alloying, mass, %)

element concentration (per 0.1% of each)

Optimal content of

the element, mass, %

T50, C

A, MJ/m2

Fw%

s, h

for such a strong influence of phosphorus, antimony, and tin segregation is a buildup of hydrogen concentration on the grain boundaries as a

consequence of the chemical interaction of hydrogen with these impurities (Ref 75).

The steel structure has a determining effect on

the resistance to delayed fracture. The least resistance is typical for the martensitic structure,

especially at low tempering temperatures. An increase of the tempering temperature to 420 C

(788 F) causes the value of rth to increase by

several times. The decrease of the volume fraction of martensite to 20 to 30% means to that

even in the hardened state delayed fracture does

not develop (Fig. 8) (Ref 78). In Ref 54 attention

is drawn to the increased sensitivity of steel Fe0.4C-Cr-Si with the lower bainite structure to delayed fracture, although this structure is characterized by high-impact toughness and adequate

plasticity. Unlike the martensitic structures,

whose failure process develops mainly along

boundaries of former austenitic grains, the bainite structure fails along the grain body (by mechanism of quasi-spall).

On evidence given in Ref 74 the resistance to

the hydrogen embrittlement of low-alloyed pipe

steels increases in the direction from normalization to the controlled rolling, but the best results are ensured by the thermal improving

(quenching high tempering) and quenching

from the intercritical temperature interval. A

comparison shows that the thermal improving

and quenching from the intercritical interval ensure about the sixfold enhancement of the longterm strength as compared with the normalization. Water quenching at 775 C (1425 F) when

the structure contains approximately 30% of

bainite affords the best combination of properties. The marked enhancement of the resistance

to delayed fracture is provided by the high-temperature thermomechanical processing of the

C (0.2 to 0.8)

0.20.3

20 to 60

0.1 to 0.2

32

Solid solution

Si (0.2 to 1.8)

0.40.7

4 to 6(a)

5 to 8(b)

3(a)

7 to 16(b)

4 to 10

1 to 1.5

9

0.12

0.05

0.05

0

0

0.17

3

2

2

7(a) to 1.3(b)

1.2

4

5

15

23

6.5

5.7

7.5

0.04

0.01

0

2.9

1.5

8

80

1

260

0

0

0

36(a) to 10(b)

12

10

250

250

400

0.04

0

35

70

2.4

1.0

2.5

2.5

0.04

0

80

360

500

32

1.2

Ni (0 to 3.0)

Co (0 to 3.0)

Al (0.03 to 0.7

0.51

0.5

0.25

Carbide-forming

Cr (0.5 to 3.0)

Mo (0 to 1.0)

0.4 to 0.5

Ti (0 to 0.15)

Nb (0 to 0.20)

V (0 to 0.4)

0.05

0.02 to 0.06

0.1

1(a)

2 to 5(a)

15 to 30(a)

3 to 5(b)

60(a) to 14(b)

0(a) to 40(b)

20(a) to 15(b)

Modifiers

Rare-earth (Ce)(0 to 0.5)

AIN, VN, NbN (0 to 0.2)

0.1 to 0.3

0.2

10

40

Impurities

S (0.015 to 0.045)

P (0.004 to 0.026)

Sb (0.0006 to 0.027)

Sn (0.0005 to 0.03)

Cu (0 to 1.5)

11.5

1400

5

1200

Stress , MPa

Mn (0.6 to 2.5)

1600

1000

800

600

2

200

1

0.01

0.015

0.01

0

0.5

0

230

180

180

1

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

2

3

Time to failure , h

temperature on delayed failure of quenched carbon steel from 0.35% C. 1, 100% of martensite, 0.023% of

S, without tempering; 2, the same, tempering at 200 C; 3,

the same, tempering at 420 C; 4, 7080% of martensite,

0.025% of S, tempering at 200 C; 5, 7080% of martensite, 0.005% of S, tempering 200 C; 6, 2030% of martensite, 0.026% of S, without tempering. Source: Ref 78

Fig. 8

high-strength steel Fe-0.3C-2Ni-Cr-Mn-Si (Ref

79).

Some Specific Cases of Delayed Fracture.

In the scientific literature, including the educational one, it was a widespread opinion that delayed fracture of steels can develop involving no

hydrogenthat is, under the effect of only structure factors (Ref 28, 54, 80). In particular, the

question about the nature of the enhanced sensitivity to delayed fracture of steels in the socalled newly quenched state is long discussed.

Cases of quenched products cracking, which are

met in practice of thermal processing (Ref 52,

54), have been associated with the development

of delayed fracture of untempered martensite.

Initiation and growth of crack provoking delayed

fracture occurs mostly over the boundaries of

primary austenitic grains. The short-term tempering or resting of quenched samples at room

temperature enhances resistance to delayed fracture; therefore, its development calls for further

hydrogenation. Within the span of more than 50

years of investigations, alternate hypotheses for

explaining this phenomenon have been proposed. The structure hypothesis based on the

prevailing role of residual microstresses inherent

in the untempered martensitic structure has received the most acceptance (Ref 79, 80). The

role of internal hydrogen, which is always present in steel as an inevitable impurity, has not

been discussed practically in the literature.

The phenomenon of delayed fracture of

quenched steels has been studied in detail in Ref

8183. Steels Fe-0.3C-Cr-Mn-Si and Fe-0.4CCr-Mn quenched in oil at 860 C (1580 F) have

been chosen as objects of investigation. The pro-

testing on a delayed fracture in a freshly hardened state is quite the same as that in the grain

boundaries. The typical initiating crack found in

the stress concentrator top is shown in Fig. 9.

The work has been methodically performed in

the following way. First, regularities of the initiation and crack growth have been studied thoroughly by methods of mechanical tests in conjunction with the acoustic emission. In

particular, the effect of the deformation-rate inhibition of the crack growth has been established, durations of the incubation period, si, under different conditions of loading have been

determined and so on. Second, the simulation of

the newly quenched state has been performed.

For this purpose, after approximately 70 h exposure at room temperature (to eliminate the sensitivity to delayed fracture) quenched samples

have been hydrogenated by the special procedure.

Subsequent tests have shown that all the characteristics of delayed fracture of quenched

steelsthe threshold stress, rth, the duration of

the incubation period, si, and so on, as well as

the mechanism of fracture, are analogous with

the case of delayed fracture caused by hydrogen.

Therefore, it is beyond the reason to consider

that in the case of the untempered martensite

structure a certain different type of delayed fracture that is not connected with the presence of

hydrogen in steel takes place. The fact that the

kinetics of relaxation of interior microstresses in

the process of resting at room temperature does

not correspond to the kinetics of the change of

the threshold stress, rth, bears witness to this

also. Judging by the change of the interior fric-

80

3

, %

60

40

1

20

2

0

2500

3

B, MPa

2000

1500

2

1

1000

500

340

380

420

460

500

540

T, C

Nucleating grain boundary crack revealed at the

top of the stress concentrator on testings of

quenched steel Fe-0.3C-Cr-Mn-Si for delayed failure.

1100. Source: Ref 81

Fig. 9

steel Fe-18Ni-9Co-5Mo-Ti. 1,1 testings in the

air, u 4045%; 2.2, the same, u 6570%; 3,3,

vacuum testings, P 1.3 Pa. Source: Ref 86

Fig. 10

tion, the microstress relaxation process completes in a time of no more than 10 to 20 h since

quenching. In this state a material still possesses

the sensitivity to delayed fracture. In the process

of holding samples at room temperature, the evolution (desorption) of hydrogen from them into

the atmosphere has been found to occur. The

sensitivity to delayed fracture is just concurrently with termination of hydrogen desorption

(i.e., about 70 h after quenching has been

elapsed). According to the data of the laser spectroscopy, the amount of desorbent hydrogen over

the period of holding comes to approximately 2

106 at.% (Ref 84).

The source of diffusive-mobile hydrogen in

the structure of the quenched steel is inner hydrogen, whose amount at conventional process

of melting makes up no less than 2 to 3 104

at.%. The lattice solubility of hydrogen increases

from approximately 107 at 20 C (68 F) to 4.4

104 at.% at 860 C, (i.e., by about 3 orders

of magnitude), as a result of heating to the austenization temperature. A reverse decrease of

solubility in the process of the fast quenching

results in a part of hydrogen being kept for some

time in the oversaturated state in the -solid solution and has the increased diffusive mobility

and the property of localizing in the region of

maximal tensile stresses. High interior microstresses inherent of the quenched steel structure

make the process of the crack initiation and the

crack growth easier but are not the factors that

control kinetic parameters of delayed fracture.

Another interesting case is delayed fracture of

high-strength maraging steels, which develops

over the certain temperature interval of aging

400460 C (Ref 85). The development of delayed fracture in these steels requires no special

hydrogenation. Therefore, within a long period

of time this phenomenon was also connected

only with the peculiarities of the structure state

forming at indicated aging temperatures. Furthermore, this problem is urgent because aging

in the indicated interval provides for obtaining

the maximum elastic limit of steel. However,

owing to the low resistance to delayed fracture,

this mode of aging usually is not practiced. An

analysis of data in hand has shown that the key

question in explaining the phenomenon of delayed fracture of maraging steels is a question on

the role of the testing environment (air, vacuum)

as well as the role of titanium (Ref 85).

The investigations have been carried out with

steels Fe-16Ni-10W-Mo-Ti and Fe-18Ni-9Co5Mo-Ti (Ref 86, 87). The tests have been performed under static loading at the reduced rate

of e 2.8 105 s1. Dependencies of the mechanical properties of steel Fe-18Ni-9Co-5MoTi on the aging temperature when testing in the

air and in vacuum are presented in Fig. 10. A

crevasse in curves of the change of mechanical

properties on tests in the air (curves 1,1) corresponds to the aging temperature of 420 to 460

C. Plastic properties of steels in this range come

practically to the zero level. An effect of delayed

fracture for steel Fe-16Ni-Mo-10W-Ti is observed within a more narrow temperature inter-

val at 420 C. On testing in vacuum (residual

pressure of 1.3 Pa) the indexes of mechanical

properties abruptly increase (curves 3,3). An effect of the delayed fracture elimination is due to

the change of the crack-initiation mechanism.

On testing in the air the brittle grain-boundary

fracture occurs, on vacuum testing the viscous

inter-granular fracture takes place (Fig. 11). In

the study of the titanium influence it has been

found that the influence of the testing environment takes place only with a definite content of

titanium in steel exceeding 0.5%. When titanium

content is 0.5%, mechanical properties tests in

the air and vacuum tests do not practically differ

from one another (i.e., delayed fracture does not

develop).

It has been found also that resistance to delayed fracture depends on moisture of the environmental atmosphere (the laboratory air),

namely, on a season in the course of which tests

are carried out. The figure shows that an increase

of relative humidity (from 4045% to 6570%)

Fig. 11

reduction of the tensile strength by 30% (curves

1,1, 2,2). Thus, tests have been carried out in

the air-dried chamber, for which purpose the special procedure has been devised. On testing in

the air-dried atmosphere (u 0.04%) plastic

and strength properties of steel Fe-18Ni-9Co5Mo-Ti come to such a degree that corresponds

tests in vacuum (i.e., the effect of delayed fracture disappears). This means that the structure

state formed on aging within the interval of 400

to 460 C possesses the increased sensitivity to

the presence of water vapors in the air environmentthat is, delayed fracture of high-strength

maraging steels is caused by the effect of atmospheric moisture adsorbed on a surface.

Investigations of grain boundaries of fractured

samples by x-ray electron spectroscopy have

shown that on the grain boundaries there are hydrides, first of all, titanium hydrides, TiHx, apart

from apparent components (segregations of alloying elements, oxides, hydroxides, and inter-

Mechanism of failure of steel Fe-16Ni-10W-Mo-Ti at crack initiation stage (aging at temperature of 420 C).

(a) Testings in the air. (b) Vacuum testings (P 1.3 Pa). 1050. Courtesy of A.I. Kovalev

H3O

102

H2O

10-20

Oxide-hydroxide layer

P

Grain 1

~10

H3O

Ni3Ti

Hads

Grain 2

Grain 1

Oxygenless zone

Grain boundary

(a)

Fig. 12

(b)

Grain 2

Ti segregation

Ni3Ti

H2O

Air

Habs

TiHx

Grain boundary

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

of crack growth. Source: Ref 87

of the Ti-H type in the air is practically excluded

(titanium, by virtue of its enhanced activity, in

its interaction with oxygen and moisture in the

air forms first of all the hardly permeable oxide

phase TiO2), the conclusion has been drawn that

titanium hydrides are formed on inner interfaces

(grain boundaries). In other words, the TiHx formation occurs in the zone of tensile stresses

ahead of the crack top as diffusive-mobile hydrogen enters there.

According to the mechanism suggested (Ref

86), the process of grain boundary embrittlement

includes the following successive stages: adsorption of atmospheric moisture at the metal surface, oxidation of the most electronegative steel

components (iron atoms), transport of H3O

ions in the film of adsorbed water toward the top

(mouth) of the growing crack, formation of

atomic hydrogen H0ads on the freshly formed surface by cathodic reduction of H3O ions on

nickel atoms, transition of atomic hydrogen

H0ads into the absorbed state and subsequent diffusion of hydrogen to the tension-stressed zone

ahead of the crack top, and embrittlement of

grain boundaries due to both the formation of

titanium hydrides and segregations of hydrogen

solutes ahead of the crack top. Thus, the final

stage of the process is the formation of diffusivemobile hydrogen, which causes embrittlement

owing to both the formation of titanium hydrides, TiHx, and the direct action (decohesion).

This is the main difference of the studied type of

brittleness from the delayed failure of the

quenched steel, which is caused by internal diffusive-mobile hydrogen formed in the process of

steel heat treatment.

The mechanism of the crack growth, which

illustrates the sequence of stages under consideration is represented schematically in Fig. 12

(Ref 87). The formation of titanium hydrides in

the crystal lattice is attended with the considerable volume effect, which is about 3 times higher

than for the martensite transformation in steel.

Besides, hydrides favor the crack growth owing

to the reduced strength of the hydride-matrix interface (Ref 28). Atomic hydrogen H0ads formed

in the crystal lattice is ionized producing ions

H and H, which quantitative ratio, on data

given in Ref 88, makes up 1:9, approximately.

The rate of diffusion of ionized hydrogen is 1 to

2 orders of magnitude higher than that of atomic

hydrogen. According to data obtained by the

method of the secondary ion mass-spectrometry

(Ref 88, 89) more mobile ions H, which serve

as main carriers of diffusive-mobile hydrogen in

metal, are responsible for the formation of the

hydride phase and the process of embrittlement

of the grain boundaries. Ion H possesses the

lesser mobility in the crystal lattice, and its contribution to the embrittlement is not so considerable essential.

In Ref 90 and 91, delayed failure of high-alloy

steel Fe-24Cr-7Ni-3Mo-Al-Ti containing no

martensitic structure has been studied. This steel

displays the enhanced sensitivity to delayed failure in the so-called newly-smelted state (i.e.,

under the effect of metallurgical hydrogen determining the role of which in the development

of delayed failure is of great practical importance). The studied steel structure is embrittled

d-ferrite with austenite c-phase interlayers uniformly distributed in it. Unlike the cases discussed previously, the initiation and the crack

growth in the structure under study occur in accordance with a transgranular mechanism, with

defects of the cast structure (pores) of 20 mkm

in size being the source of fracture. As evident

from the registration of the acoustic emission,

the crack grows according to the mechanism of

the discrete (stepwise) growth of separate sites

of the crack front followed by the stop at viscous

sites (austenitic inter-layers). According to observations, the resistance to delayed fracture depends on the period of the natural recreation after

smelting. Within the span of 4 years the metallurgical hydrogen content drops from approximately 18 104 to 6 104 at.%, and it is

a limiting one at which the development of delayed fracture is still possible.

The object under study is of interest as far as

it allows the different mechanisms of the embrittlement of grain boundaries or their volume

to be realized. The thermal treatment has been

carried out under operating conditions including

the high-temperature preheating to 1250 C

(2282 F) for obtaining the one-phase structure

of d-ferrite and postaging (annealing) at different

temperatures. The hydrogen content in thermally

treated samples does not exceed 1.4 104

at.%. Annealing at 680 C (1256 F) makes it

possible to obtain the brittle fracture along the

grain boundaries, which, on data of the auger

spectroscopy, arises from the segregation of sulfur and phosphor impurities. After the 850 C

(1562 F) annealing on the grain boundaries of

ferrite, the abundant precipitation of embrittling

r-phase responsible for the quasi-brittle grain

boundary fracture takes place. Aging at 440 C

(824 F) makes it possible to obtain the brittle

intergranular fracture in accordance with the

transgranular mechanism.

The tests performed on embrittled samples

have once again favored the view that structure

factors bringing about the embrittlement of the

grain boundaries or their volume cannot in themselves result in the development of delayed fracture. As is obvious from this data obtained from

steels with the different structures, the development of delayed fracture is only possible with

the participation of diffusive-mobile hydrogen

independently of the source of its origination,

namely on quenching, smelting, from atmospheric moisture, and so on.

Stress-corrosion cracking (SCC) is a specific

type of delayed fracture that develops under the

concurrent action of tensile stresses and the corrosion medium. The given type of fracture occurs in chemical, gas-oil producing, metallurgic,

accounts for approximately 30% of material

damage caused by corrosion. The most dangerous media that can cause SCC under the certain

conditions have been established for most steel

classes used. The greatest body of information

accumulated to date refers to high-alloy steels,

in part, to corrosion-resistant steels with austenitic structure. Stress-corrosion cracking of lowalloy steels has been studied least, although such

steels, especially in the high-strength state, have

an enhanced tendency to corrosion-mechanical

fracture. Low-alloy steel cracking can occur under the action of various actuating media including solutions of acids, nitrates, alkali, and sulphur-hydrogeneous media as well. The low

resistance of high-strength steels to SCC makes

itself evident even in such, at first sight, undangerous media as pure water, moist air, and water

vapors (Ref 92, 93).

One of the characteristic peculiarities of SCC

is a pronounced selectivity of a material relative

to a corrosion medium. Therefore, the results of

one metal-medium system investigation cannot be extended to another system. Besides, the

noncorrodive media, in common conception of

this term, may appear to be the most aggressive

ones, because the corrosion cracking process is

actively developing in them under the effect of

tensile stresses. Another interesting peculiarity

lies in a usual insensitivity of high-pure metals,

including iron, to SCC in any environment. The

susceptibility of iron to cracking under the effect

of stresses appears only in the presence of introduction impurities, namely, carbon and nitrogen.

The response to the cold plastic deformation is

also intriguing. Unlike austenitic corrosion-resistant steels, the cold plastic deformation of carbon steels makes them relatively insensitive to

SCC. Annealing of the cold-deformed 0.15% Ccontaining steel at 600 C (1112 F) restores this

sensitivity to the level of nondeformed steel (Ref

70).

Peculiarities of Crack Initiation and

Growth Mechanism under SCC. The mechanism of initiation and growth of cracks under

SCC depends on the steel structure, the value of

applied stresses, and the testing medium. Fracture along the primary austenitic grain boundaries is the most dangerous from the standpoint

of loss in corrosion-mechanical strength. Carbon

steels, as well as many of the low-alloy steels,

fail in alkaline, nitrate, phosphate, carbonate solutions, and also in aqueous acid solutions by

such a mechanism (Ref 75). Depending which

reaction controls the cracking processelectrochemical, anodic, or cathodictwo main SCC

mechanisms are differentiated: the active-part

corrosion (APC) or the hydrogen-induced corrosion cracking (HISCC). Kinetics of the anodetype cracking is determined largely by the conditions of passivation (i.e., by the peculiarities of

the protection corrosion films formation and by

their properties). This is reflected in the fact that

the region of danger of the given type of SCC

corresponds to the so-called active-passive region of potentials in the polarization curves. The

no clearly defined boundaries and extends from

the potential of free corrosion to the potential of

cathodic protection (Ref 94).

The mechanism of part cracking presupposes

the accelerated metal dissolution, which is considered to be connected with the local disruption

of passivity at the crack top i.e., with the protective film breakdown and occurrence of the newly

formed (juvenile) surface. The mechanism of

HISCC involves the following main reactions:

discharge of hydrogen H3O ions from water

solution onto the steel surface, the transport of

hydrogen to the region of the elevated stress concentration ahead of the crack top, and the crack

propagation caused by the local hydrogen embrittlement. The concrete mechanism of the hydrogen embrittlement depends on the structure

and properties of steel, the loading conditions

and the ability of a medium to hydrogenation

(Ref 94).

Whatever the mechanism of corrosive-mechanical fracture is, the three stages of the process are distinguished: incipient defects (microcracks) formation, microcracks growth, and fast

fracture as soon as the main crack reaches the

critical size. Both metallurgical defects of the

surface and defects of corrosion origin may present as incipient defects (Ref 93). In the SCC

theory, particular attention is given to the processes proceeding in the crack top, which are

much different from the processes proceeding on

the surface. In particular, an important role is

given to the newly formed (juvenile) surface,

which appears either under the effect of the external loading or in the process of APC. All the

particular SCC reactions, namely, APC, cathode

precipitation of hydrogen, its absorption, and so

on, are sharply speeded up on such a surface. The

newly formed surface possesses a more negative

potential compared with both the previous surface and the passivated surface of crack edges.

It is suggested that in the corrosion crack there

is a local galvanopair of the newly formed surface-crack edges surface type that contributes to

the accelerated selective crack propagation (Ref

95).

Criteria and Methods of Estimating Sensitivity to SCC. A great body of different criteria,

namely, mechanical, electrochemical, physical,

and so on, has been suggested to estimate the

sensitivity of steels to SCC (Ref 92). All the

known methods of corrosion-mechanical testings can be divided into three main groups:

Testings under constant external load (P

constant)

Testings under constant strain (e constant)

Testings at the constant load rate (e, v constant).

Samples without a notch, with a notch, and with

the precreated fatigue crack can be used in each

of these above-stated methods.

The service conditions are best reproduced by

corrosion-mechanical testings under constant external load. In this case, as the crack initiates and

grows, the stress rises in the rest part of a sample

that contributes to the development of fracture.

By analogy with delayed fracture (see the section

Delayed Fracture in Steels), the sensitivity to

SCC is characterized by the time till a crack initiation, si, durability till complete failure, s, and

the value of the threshold stress, rth. Unlike delayed fracture, it often occurs that dependencies,

s f (r), are not brought out onto the plateau,

when testings for corrosion cracking are performed (i.e., there is no threshold stress). The

conventional threshold stress, rth, which should

be determined in this case, is determined at the

specified (base) duration of testings. The duration of the incubation period si during which the

crack initiation occurs takes usually 80 to 90%

of the whole time prior to fracture (s) (Ref 96).

Testings under constant (fixed) strain are usually carried out according to the scheme of threeor four-point bending of samples fixed in the

rigid attachment. On such testings the external

load is reduced gradually because of the crack

formation and the residual bending strain accumulation. Testing under constant strain render

well the cracking processes in the material structure that develop under the action of residual

stresses.

Tests at constant load rate are mainly performed by the slow strain rate testing procedure

(SSRT). In this case the resistance of a material

to cracking is characterized by the value of tensile strength in the corrosion medium, characteristics of plasticity, W, d, the critical fracture

stress, r*, and the area below the strain curve, r

e, characterizing the fracture work. With the

plasticity characteristics taken into accounts, the

relative (as compared to the air) sensitivity of a

material to corrosion cracking is determined.

Used as criteria in this case are the so-called coefficients of medium effect of the (W0 W1)/

W0 type, where W0 and W1 are the relative narrowing on testings in the air and in the corrosion

medium, respectively. The strain rate e with the

use of the SSRT method is usually selected

within the range of 104 to 106 sec1.

When using samples with a crack, a load, the

load line displacement, the crack length, and the

crack opening displacement (COD) are registered as per the SSRT method. For analyzing the

test results, the criteria of fracture mechanics are

used. The threshold (critical) stress-intensity factor, KISCC, which makes it possible to estimate

quantitatively the steel resistance to cracking in

the corrosion medium is taken as the main criterion. The value of KISCC corresponds to the beginning of the crack growth before the critical

value (Ref 97). It is apparent that if K1 KISCC

or r rth, the fracture of the construction element inevitably takes place. When the criterion

KISCC is used, the sensitivity of material to the

action of the corrosion medium is characterized

by the ratio KISCC /KIC, where KIC is the stressintensity factor when testing in the air (planestrain fracture toughness) (Ref 98100). In all

the schemes of testings the maximum length of

a crack resulting in fracture, the crack growth

rate, the number of cracks per unit surface, etc.,

can be used as characteristics of a material (Ref

101, 102).

on the basis of standard kinetic cracking resistance diagrams analogous to that represented in

Fig. 3. The plateau-figurative regions in kinetic

diagrams are connected with the branching and

blunting of a crack. If relative resistance of materials to the action of corrosion media is to be

determined quickly, the SSRT method has an advantage over the constant load testing method for

determining KISCC, although it is considered the

basic one. The fact of the SSRT method high

efficiency is evidenced by the data of many scientists. The console samples made of Fe-0.3C2Ni-Cr-Mn-Si steel, for example, have been

tested in the NaCl 3.5% solution (Ref 103). The

values of KISCC obtained at a constant rate of load

and on standard testing (under constant load)

have turned out practically to be equal to 11.69

and 1196 MPa m1 / 2, respectively. To choose the

proper load rate is of great importance in this

case. At the high load rate the corrosion process

has no time to produce changes at the crack top;

at low rate, on the contrary, the passivation process develops at the crack top, thus preventing

from corrosion cracking propagation (Ref 96).

The linear mechanics of fracture are unsuited

for high-ductility steels. In this case, resistance

to crack growth is characterized by the value of

the J-integral, in particular, its threshold value,

JISCC. A tendency to corrosion-mechanical

cracking is estimated also by electrochemical parameters, namely, by the critical value of the hydrogen index of the medium pH, the electrode

potential, E and the density of anodic current at

which the cracking process begins. When testing

the carbon steel in the 3% solution of NaCl,

strong jumps of the electrode potential (300

mV) and the anodic current at the moment of a

crack propagation have been detected (Ref 104).

This effect is connected with revealing the juvenile surface, which has an enhanced electrochemical activity and it can also be taken as one

of criteria for the sensitivity of steels to corrosion-mechanical fracture. The important information on kinetics of SCC, especially at initial

stages of the process, can be obtained from

acoustic emission measurements (Ref 105107).

Effect of Alloying Elements on Resistance

to SCC. The effect of most alloying elements on

steel resistance to SCC depends on their concentration, the presence of other alloying elements,

impurities, the corrosion medium content, the

value of electrode potential, and other factors

(Ref 108). For high-alloy austenitic steels there

is a certain relationship between the effect of an

element and its position in the periodic system.

The analogous relationship is not observed for

low-alloy steels (Ref 109). The theory of alloying that makes it possible to predict the resistance of low-alloy steels to SCC in different media also has not been developed. Hypotheses

suggested provide, as a rule, an explanation for

the role of alloying elements by only one of

some factors, namely, effect on strength of interatomic bonds, the rate of forming protective

films, the effect on the carbon distribution in

steel, and so on (Ref 108).

on mechanisms of the crack growth in different

corrosion media is necessary to develop successfully new grades of steels with an enhanced

resistance to corrosion cracking. For lack of this

information, the empirical data are usually used

which are often contradictory. The findings on

the individual and combined effects of 27 alloying elements and impurities on the resistance to

SCC of iron and low-alloy steels, including highstrength steels, have been generalized in Ref

108. The effect of elements depending on the

medium (aqueous solutions of chlorides, nitrates, carbonates, alkali, hydrogen sulfide, etc.),

the temperature, the electrode potential, and the

pH value have been considered. The following

elements have been analyzed: N, Al, B, V, H,

Ge, Ca, O, Co, Si, La, Mn, Cu, Mo, As, Ni, Nb,

Sn, Pb, S, Sb, Ti, C, P, Cr, Ce, Zn. It has been

shown that the majority of these elements, especially in the case of combined alloying, ambiguously affects resistance to SCC. The same

element can have both a positive and negative

effect depending on the corrosion medium. Silicon, for example, enhances steel resistance to

cracking in nitrate solutions but reduces it in alkali ones. However, the previously listed elements can be divided arbitrarily into four main

groups: those that enhance the resistance to SCC

(Ti, Nb, Mo, Al, Co, Cu, Pb, B, Ce, Ca, La),

those that reduce it (N, O, H, S, P, V, Sb, Cr,

Mn, Ge), elements of the alternating effect (C,

Si, Ni), and neutral ones (Sn, As, Zn).

The characteristic properties of distribution of

elements and impurities in the steel structure are

of great importance. Segregations of sulphur and

phosphor impurities, for example, change the

electrochemical characteristics of the grain

boundaries and initiate fracture by the local parth

corrosion mechanism (Ref 108). Steels produced

by vacuum melting possess the enhanced resistance to SCC in so far as they contain 2 to 3

times as few detrimental impurities as standard

steels do (Ref 110). It has been noted that the

effect of alloying elements can be judged by the

mechanism of fracture. Elements with a positive

effect assist in realizing the transgranular mechanism of the crack growth. Those that have a

negative effect cause cracking along the grain

boundaries. The effect of the chemical composition may manifest itself indirectly, namely,

through the change of the structure and mechanical properties of the steel. Therefore, the thermal processing can drastically change degree

and even the direction of some alloying elements

effect.

Role of Structure and Thermal Processing

in the SCC Process. The basic parameters of the

SCC, namely, the threshold stress, rth, corrosion

crack-toughness, KISCC, the crack growth rate,

and so on, depend to a great extent on the following structure factors: phase composition,

austenite grain size, state of grain boundaries,

and nature and amount of nonmetallic impurities

(Ref 111). Figure 13 shows that the crack-tough-

ness reaches its peak values KISCC in hightempered steels. In the region of high-strength

states realized at low tempering, the parameter

KISCC takes quite low values not exceeding 20

MPa m1 / 2.

As for steels of mean and low strength, the

structures of fine-grain ferrite, martensite, and

bainite possess the best properties. The spheroidized structure with uniformly distributed finegrain carbides possesses somewhat worse properties, and the lamellated pearlite structure

possesses the worst ones (Ref 112). There is also

evidence that compound structures, for example,

consisting of fine-grain ferrite and bainite

wherein the intergranular ductile mechanism of

failure is realized (Ref 113, 114) show the highest resistance to SCC. For the steel Fe-0.3C-MnSi-Ni with the compound structure of martensite

lower bainite, the value KISCC in the NaCl

3.5% solutions is 21 MPa m1 / 2. In the case of

the martensitic structure this index is reduced by

half (i.e., steel with this structure possesses the

much lower resistance to SCC) (Ref 115).

The positive role of residual austenite is observed frequently. In the scope of the SCC model

by the hydrogen embrittlement mechanism this

derives from the fact that the saturability of austenite with hydrogen is essentially higher, and

the hydrogen diffusion factor is three to four orders of magnitude lower than that in the martensite structure. Cracks propagating in martensitic

slow down in the region bordering hydrogen-resistant austenite. However, with the insufficient

stability of residual austenite its negative effect

is a possibility. In this case, as residual austenite

is turned to martensite under deformation at the

crack top, the resistance to SCC reduces (Ref

116).

The use of high-temperature thermo-mechanical processing which raises both the parameters

100

1

2

3

4

5

6

7

80

60

40

20

1000

1400

B, MPa

1800

Strength level effect of steels Fe-0.3C-2Ni-CrMn-Si (1,2,3), Fe-0.4C-Cr-Ni-Mn (4,5), and Fe0.3C-Cr-Mn-Si (6,7) on value of threshold stress-intensity

factor KISCC in corrosion medium. 1,4,6,8, tests in water;

2,5,7, tests in H2S solution; 3, tests in gaseous H2S. Source:

Ref 93

Fig. 13

KISCC and KIC in the case of high-strength combined-alloy steels (when testing in the NaCl

3.5% solution), holds promise (Ref 117).

Quenching from the intercritical temperature interval can also be profitable. The formation of

the ferritic-martensitic structure as the result of

quenching from the intercritical temperature interval (740 C, or 1364 F) makes it possible to

enhance vastly the resistance of steel Fe-0.5CCr-Mo to the hydrogen SCC in salt water (Ref

118). The two-stage austenitization (high-temperature conventional one) forms a toothed

structure of grain boundaries, which also inhibits

the grain boundary failure by enhancing the resistance to SCC.

The effect of grain size on the resistance of

steels to SCC is then ambiguous. In many works

it was pointed out that the finer the austenitic

grain, the higher resistance to SCC is observed.

For example, on the data given in Ref 119 and

120 the time before failure in the NaCl 3.5%

solution increases by 2 to 3 orders of magnitude

with decreasing the austenite grain in the steel

Fe-0.4C-Cr-Ni-Mo from 172 to 12 mkm in size.

However, there is other evidence. For example,

overheating in the process of the austenitization

of the high-strength steel Fe-0.45C-2Ni-Cr-MoV favors a great increase of KISCC when testing

in distilled water (Ref 121). The high-temperature austenitization is most profitable for cast

steels in so far as they are less prone to the grain

growth as compared to hot-strained steels (Ref

122). Unlike quenching, the overheating in normalizing, as a rule, has a negative effect, especially in the case of low-alloy steels containing

manganese and vanadium (Ref 123).

The problem on the effect of initial structure

is of great importance. It is found that, compared

with hot-strained steels, cast steels have higher

values of the corrosion cracking resistance, although standard mechanical properties of cast

steels are much lower (Ref 122). Studied in Ref

124 is the effect of different methods of melting

(an electric furnace, an arc-vacuum furnace, a

vacuum induction furnace, electroslag remelt)

and their combinations on the corrosion mechanical properties of maraging steel. It has been

shown that combined methods of melting make

it possible to increase considerably (by 25%)

KISCC when testing in the NaCl 3% solution. The

analogous data have been obtained in work (Ref

110). Commercial Fe-0.4C-2Ni-Cr-Mo and Fe0.4C-2Ni-Cr-Mo-Si steels that have been melted

in the air and vacuum were tested in the chloride

medium at K 0.9KI. Vacuum-melted steels

possessed the enhanced SCC resistance, whereas

in air-melted steels, cracks developed practically

at once.

In corrosion-mechanical steel failure one of

the key problems is the problem of nonmetallic

impurities. In this case, both morphological

characteristics of impurities (the quantity, form,

sizes, and distribution) and their chemical activity in the given corrosion medium are of importance. The impurities of sulphides, silicates, and

oxides as well, which play the main role in the

process of the crack initiation, are the most ac-

tive (Ref 111). The local electrochemical corrosion processes developing in the corrosive medium around the impurities, resulting in

inevitable formation of aggressive medium-filled

microcavities. The role of MnS impurities in the

course of corrosion-mechanical failure of the Fe0.2C-Cr-Ni-Mo type reactor steel intended for

the production of high-pressure containers was

studied in (Ref 125). The suggested mechanism

of initiation and growth of cracks includes, as

the first stage, the appearance of a slot between

an impurity and a matrix owing to dissolving

MnS in the electrolyte. Atomic hydrogen is adsorbed at crack lateral surfaces, following which

it diffuses into the metal toward the crack top

and causes local cracking.

In acid media sulfide impurities dissolve and

yield H2S that greatly reduces the resistance to

SCC (Ref 126). Dissolution products of some

impurities enhance the adjacent metal corrosion

that also promotes the cracking process. The enhancement of cracking in dissolving nonmetallic

impurities is supported by the data given in Ref

127 devoted to the study of SCC of the Fe-0.2CMn-Ni-Mo steel in the medium intended for

cooling atomic reactors. As the medium has been

contaminated by dissolution products, the MnS

samples have broken down due to the development of the corrosion cracking process, whereas

in the medium free of the corrosion products, this

process has not occurred. At the same time, as

shown in Ref 128, with cathode deposits present,

the SCC process of pipe steels of Fe-0.1C-2MnSi-V-Nb and Fe-0.1C-2Mn-Si-V, and so on type,

which have been chosen from zones of main

pipeline failure cannot be explained in the context of the suggested model of the sulphide impurity dissolution (Ref 128). On authors data,

the negative effect of sulphides is connected

mainly with their role as concentrators of internal stresses (Ref 129). Nonetheless, in spite of

distinctions between interpretations, it is believed that the enhancement of the corrosion-mechanical steel strength can be attained by reducing the nonmetallic impurity content that is

achieved in part by rational deoxidizing and by

modifying the steels by the addition of rare-earth

and alkaline-earth elements.

Stress Corrosion Cracking of Steels for

Main Gas-Pipe Lines. In many countries,

namely, the United States, Canada, Australia,

and Russia, there are recorded emergency failures of subterranean buried gas-pipe lines because of pipe steels SCC (Ref 130, 131). The

summary data displaying the time of trouble-free

operation of main gas-pipe lines used by three

major Russian companies are given in Fig. 14.

Pipes of different diameter (1020, 1220, 1420

mm) and different chemical composition have

been analyzed. It has been found out that pipes

put into service from 1967 to 1980 have the resistance to SCC much higher than those which

have been laid recently in a period from 1981 to

1989. The time of safe service of the former is

from 9 to 23 years, that of the latter is only from

2 to 14 years. It is significant that all the shortlived pipes have been fabricated from steels of

the Fe-0.1C-2Mn-Si-V-Nb and Fe-0.1C-2MnSi-V type produced in France, Italy, Germany,

and Japan. Also, it has been found that in the

most cases (more than 60% of emergencies) the

failure happens in places with the increased

ground humidity. Cracks are revealed most often

in the lower part of pipes. The mean rate of crack

growth is about 1.15 mm/year (0.05 in./year).

The critical length of cracks that cause failure is

in the range of 140 to 150 mm (5.55.9 in.) (Ref

129).

According to current concepts, main gas-pipe

line steels are susceptible to two types of SCC:

APC and HISCC. The first (classical) type of

SCC is developed at medium acidity of pH

9.5 (high pH SCC); the second one is developed

at pH 6.5 (low pH SCC). In both cases crack

groups (colonies) orientated along the axis of the

pipe arise (Ref 94). More high density of cracks

in the colony is inherent in cracking by the APC

mechanism. Furthermore, it has been found that

in the overwhelming majority of cases, SCC at

a high pH is observed within the limits of regions

in extent of 20 km (12.4 miles) from the compressor stationthat is, where the temperature

and the pressure of gas have their top values (Ref

132134). As to cracking by the HISCC mechanism, the analogous regularity is not fulfilled,

but in this case, the well-defined connection with

the weld location becomes evident. A major part

of failures at low pH begins in centers located at

a distance of 200 to 250 mm (810 in.) from the

weld that is connected with the increased level

of processing residual stresses. The failure at

high pH develops mainly in the intercrystalline

manner (i.e., along the grain boundaries), at low

pH, it develops in the transcrystalline manner

(Ref 94).

30

25

20

15

C

T

Y

T

10

T

5

0

1966

1971

1976

1981

1986

1991

Period of gas main pipes service till first emergency failure depending on year of puttingthem into operation and manufacturing country. Closedcircle data points, steels Fe-0.1C-2Mn-Si-V-Nb and

Fe-0.1C-2Mn-Si-V made in Germany, Italy, France, and Japan; open circle data points, steels Fe-0.17C-Mn-Si, Fe0.15C-2Mn-Si made in Russia; rhomb, steels Fe-0.14C2Mn-Si-N-V, Fe-0.17C-2Mn-Si-N-V made in Russia. Size

of small circle corresponds to 1020 mm pipe diameter;

middle one, to 1220 mm pipe diameter; and large one, to

1420 mm pipe diameter. Y, T, and C denote companies

Ultransgas (Y), Tyumentransgas (T), and Severgasprom (C),

respectively. Source: Ref 129

Fig. 14

forming under the peeled-off insulation is the

main factor on which the realization of the given

SCC mechanism depends. High pH media are

concentrated carbonate-bicarbonate earth solutions with temperatures up to 75 C (135 F).

These solutions are formed under conditions of

the cathode protection that initiates an increase

of pH (acidifying) of the corrosion medium on

the pipe surface. The crack grows as a result of

the rapid electrochemical metal corrosion at its

top where the main processes stimulating the

crack growth, namely, the stress localization, the

plastic strain, emerging of slip strips on the

newly formed surface, are grouped. Low pH

SCC develops in deleted earth electrolyte media

containing dissolved CO2. It is suggested (Ref

130, 133137) that hydrogen arises in microregions filled with the electrolyte (i.e., in slots, pittings, etc.). As a result of corrosion product hydrolysis, the electrolyte in these microregions is

acidified to a level sufficient to discharging ions

of hydrogen, which then diffuses into the region

of tensile stresses ahead of the crack top. As in

the classic model (high-pH SCC), the plastic

strain at the crack top, which benefits the metal

hydrogenation, is required as the obligatory condition (Ref 135, 138). In this case, as seen from

the direct measurements (Ref 94), the zone of

the increased hydrogen concentration extends

for approximately 10 mm (0.4 in.) beneath the

crack surface.

Pipe surface metallurgical defects arising

from rolling may present as nucleous cracks.

These defects, for example, arise from surface

microcracks generating at the stage of continuous steel casting. Cracks initiate on the continuously casted ingots over definite temperature

intervals, mainly over the interval from 700 to

1000 C (12921832 F) (webbed and cross

cracks), and at temperatures close to the solidus

one (spider-shaped cracks). In the given temperature intervals, low-alloy steels are subject to

the strong embrittlement connected with the

change of the grain boundary chemical composition. The present state of the problem concerning high-temperature steel brittleness and the

mechanism of the surface crack initiation in continuous casting has been analyzed thoroughly

(Ref 139, 140).

Practically all the steels applied in the presentday gas-pipe lines, whatever their chemical composition, structure, and strength level, are prone

to SCC to some extent. That is why the prediction of a steels resistance to SCC takes on great

significance. The results of testing a number of

pipe steels for SCC have been correlated with

their mechanical properties determined in the air

(Ref 141). In addition to the standard mechanical

properties, the mechanical properties of samples

under bending with the preliminary-created fatigue crack have been determined. Any correlation between the resistance to SCC and the standard mechanical characteristics has not been

detected. At the same time, according to the

analysis made, a distinct relation is observed between the threshold stress, rth, in the corrosion

properties of samples with a crack while performing bending tests in the air. The resistance

to SCC in the studied aggressive media enhances

with an increasing of such characteristics as failure work, yield strength, and ultimate strength.

On this basis, the conclusion has been drawn regarding the necessity of introducing an additional testing of pipe steels in manufacturing

plants, namely, tests of samples with a crack to

predict the resistance of different types of steels

to SCC.

Elements on Hydrogen

Embrittlement Resistance of Steel

Hydrogen-resistant engineering steels should

have a number of the following basic characteristics:

High resistance to a brittle fracture for pre-

To ensure the first requirement, the structure and

alloying of steel should provide:

Hardenability of steel to gain a homogeneous

lower bainite

Decreasing the ductile-brittle transition temperature, T50, and growth of an energy for a

ductile crack propagation, ac

Low-grain-boundaries

concentration of

harmful impurities

Influence of alloying elements and impurities

on these basic characteristics has been set (Ref

51). The corresponding data, also involving the

optimal content of the basic and impurity elements providing a high hydrogen resistance of

steel, are shown in Table 1. In this case, tendency

of steel to hydrogen embrittlement was estimated from the relative reduction of its plasticity

under tension:

FW

W0 WH

100%

W0

plain samples during 5 h, and also from reduction of the average time for failure, st, of the ringnotched ones (h 1 mm, r 0.25 mm) under

approximately 0.6 ry loading.

Carbide-forming elements affect the quantity,

distribution, and morphology of carbonitrides.

The carbides are spheroidized in such a manner

that interfaces of the particles with a matrix are

not collectors for hydrogen. These elements in a

solid solution depress the hydrogen thermodynamic activity; as a result, the occluding ability

of steel is essentially diminished (Ref 142).

Those additions in steel have dissimilar influence on the resistance to different kinds of fracture. If its content does not exceed an optimal

value (Table 1), resistance to all kinds of brittleness increases (Ref 143). At higher contents, the

resistance of steel to hydrogen is increased, but

it is reduced to other kinds of brittle fracture.

Carbide-non-forming elements affect steel resistance to a ductile, brittle, and intergranular

fracture. Addition of 0.1 % Si, Mn, Co or Al in

a solid solution of an improved engineering steel

reduces its fracture resistance; the transition temperature, T50 rising by 5; and the impact strength

at decreasing by 0.07MJ/m2. Nickel is an exception; 0.1% of it makes T50 lower by 4 to 10.

Similarly, these elements influence hydrogen resistance of steel. Thus, manganese, aluminum,

silicon, and cobalt decrease the time for failure,

st, and increase the loss of plasticity, Fw, under

hydrogenation. Nickel increases steel resistance

to hydrogen.

Thus, when these elements are added in a solid

solution, the direct dependence between the hydrogen and brittleness resistance characteristics,

T50, ac, is observed.

Impurities such as sulfur, phosphorus, antimony, tin, and other elements have great influence on the resistance of engineering steels to

hydrogen embrittlement. Their action is manifested via weakening of grain-boundary binding

(temper-embrittlement processes) and via nonmetallic inclusions formation. According to existing opinions (Ref 144), loaded grain junctions

can be zones for the origination of three-axial

stress state. These grain junctions are energetically preferable for the accumulation of hydrogen atoms having a low chemical potential

value. It is assumed that hydrogen decreases the

cohesive forces between iron atoms at grain

boundaries (Ref 145). Segregation of impurities

(phosphorus, antimony, tin, etc.) at boundaries

of original austenite grains results in recurring

temper embrittlement, which further weakens intergranular cohesive forces. These considerations have been proved by direct experiments

(Ref 146). The results indicate that hydrogen

does not interact with impurities or segregations

at grain boundaries (i.e., the grain-boundary embrittlement actions of impurities and of hydrogen

mutually supplement each other).

Published data indicate that steel improving

with impurities and gases cleanliness enables it

to boost its fracture resistance (Ref 147). The use

of metallurgical refining and remelting processes

greatly increases hydrogen-embrittlement resistance due to the removal of nonmetallic inclusions from the steel, especially sulfides and

stringer oxide inclusions (Ref 148).

Improving the quality of heat-hardenable engineering steels (unalloyed with molybdenum

and tungsten) enables them to boost by 2 to 3

times their resistance to brittle and ductile fracture.

Sulfur is present in steel mainly in the form of

manganese sulfides. Nonmetallic inclusions of

this type are very ductile and, hence, during rolling are rolled out in the form of elongated

2:1:1, which comply with the formula (FeMn)S;

their thickness is 10 to 15 nm, and they are

mainly located at the boundaries of original austenite grains (Ref 149). The sulfide-matrix interfaces are sites for the accumulation of diffusionmobile hydrogen atoms, which at subsequent

stages cause hydrogen blistering in low-alloy

ferritic-pearlitic steels due to hydrogen molecules formation from its ions and greater (than

the yield strength) planar pressure. Under loading conditions the hydrogen also will accumulate

at the sulfide-matrix boundaries and other sites

where there is a high stress level and will cause

sulfide stress cracking, which is characteristic

both to mild and alloy high-strength steels (Ref

150, 151). Shape modification of sulfide inclusions from plate-like to globular lowers the sensitivity of steels to hydrogen cracking, since the

globules are less favorable to hydrogen accumulation than plates are. Calcium and rare-earth

metals are suitable for shape modification of inclusions (Ref 152).

Thus, the nonmetallic inclusions, especially

sulfides, enhance accumulation of hydrogen in

steels and increase the susceptibility of steels to

hydrogen embrittlement. Therefore, in order to

lower the sensitivity of steels to hydrogen embrittlement, it is necessary to decrease the sulfur

content in them and to modify the shape and type

of sulfide inclusions.

Sulfur, therefore, is a harmful impurity in

steels, lowering their resistance to fracture and

hydrogen embrittlement. In mild steels with ferritic-pearlitic structure, the sulfur, forming manganese sulfides of plate-like shape (stringers),

enhances hydrogen blistering, which is manifested even at 0.001% S. Sulfide (hydrogen)

stress cracking of such steels is sharply reduced

with the decrease of sulfur content below

0.007%. An effective way for prevention of blistering and hydrogen stress cracking is modification of engineering steels with rare-earth elements (0.10.3%), especially with cerium

spheroidizing sulfide inclusions.

Phosphorus, Antimony, and Tin. The adverse

effect of these elements on steel properties is

well studied. Impurities enriching grain boundaries during tempering cause reversible temper

brittleness. This phenomenon has a considerable

effect on the level of steel resistance to fracture.

Phosphorus in engineering steels greatly lowers brittle and ductile fracture. Phosphorus facilitates steel hydrogen embrittlement by enhancing temper brittleness development and

establishing chemical and structural heterogeneity.

Quantitative relationships have been found for

the embrittling effect of phosphorus, tin, and antimony on steel: Each 0.01% of the impurities

increases brittle fracture temperature, T50, by 20

to 23 C; (3641 F), decreases crack-development energy, Ap, by 10 J/cm2; and increases

plasticity loss value during hydrogenation by

8%.

Reference 153 describes a study on the influence of phosphorus content on the resistance of

Mo, Fe-bal) to hydrogen-sulfide cracking, and it

has been shown that increasing its content from

0.009 to 0.06% lowers the threshold stress from

518 MPa (75 ksi) to 140 MPa (20 ksi). Such an

effect of phosphorus is attributed, first, to the fact

that it activates the hydrogen absorption process

and, second, that it changes the shape and distribution of carbides. Phosphorus or manganese

segregation can be formed at a pearlite band in

low-alloy steels (Ref 154). Such formation

causes structural changes, an increase of steel

occlusive capacity by 2 times, and a decrease of

the time for failure under load in a hydrogenating

medium by 2 times.

The adverse effect of phosphorus on steel resistance to brittle fracture is comparable with the

favorable effect of nickel and molybdenum:

Each 0.01% lowering of phosphorus content in

steel is equivalent, in its influence on the T50

level, to steel alloying with 1% nickel or 0.1%

of molybdenum.

Antimony and Tin. The resistance of engineering steels to brittle fracture depends, to a

large extent, on its cleanliness with regard to antimony and tin impurities. These impurities are

brought into steel in considerable quantities

(0.010.03%) during melting when using scrap.

Both elements increase the cold-shortness

threshold and lower the impact strength level

(Ref 155). The degree of its influence may vary

with the steel alloying system to a great extent.

In low-alloy steels with ferritic-pearlitic structure, the addition of antimony, which is a more

electro-positive noble element than iron, improves the corrosion resistance in hydrochloric

and sulfuric acids, assumed to be due to a film

containing antimony forming on the steel surface. The surface layer inhibits penetration of hydrogen into steel (Ref 156).

To a greater extent, antimony and tin lower

the steel resistance to brittle fracture: Each additional 0.01% Sb or Sn raises the T50 by an average of 18 C (32 F). Their adverse effect in

this respect is comparable with the favorable influence of nickel and molybdenum.

It is possible to assume that at an antimony

content above 0.3%, the effect of lowering steel

resistance to brittle and ductile fracture will prevail over the effect of lowering the quantity of

absorbed hydrogen. Hence, additional alloying

with approximately 0.20 to 0.30% antimony is

advantageous for low-alloy steels to be used in

hydrogen-sulfide media. The associated reduction in the level of toughness properties is insignificant as compared with the improvement in

the resistance to hydrogen embrittlement by 2

times.

With respect to these considerations the following conclusions can be drawn:

Antimony and tin impurities lower the resistance of structural steels to brittle and ductile

fracture: Each 0.01% increase in antimony or

tin raises T50 of heat-hardenable steels by 18

C (32 F), and each 0.01% increase in antimony raises T50 of low-alloy steels by 1 to 5

C (1.89 F).

The embrittling effect of antimony and tin is

brittleness process: Addition of 0.0015% Sb

or Sn increases the T50 of heat-hardenable

steels by 100 C (180 F).

Antimony additions (up to 0.30.5%) in lowalloy ferritic-pearlitic steels retard electrochemical reactions on a steel surface, reducing in the amount of absorbed hydrogen.

Consequently, these additions increase (by

more than 2 times) the resistance of steel to

SCC.

REFERENCES

1. H. Epstein, J. Chipman, and N.J. Grant,

Met. Trans., Vol 9 (No. 4), 1957, p 597

2. N.N. Kolgatin, Zavodskaya Laboratoriya,

Vol 9, 1957, p 1098 (in Russian)

3. T.C. Poulter and L. Uffelman, Physics, Vol

3, 1932, p 147

4. W. Geller and T.H. Sun, Arch. Eisenhuttenwes, Vol 21 (No. 11/12), 1950, p 423

5. E. Martin, Arch. Eisenhuttenwes, Vol 3

(No. 6), 19291930, p 403

6. A. Sieverts et al., Z. Phys. Chem., Vol 183

(No. 1), 1938, p 19

7. L. Luckemeyer-Hasse and H. Schenk,

Arch. Eisenhuttenwes, Vol 5 (No. 6),

19321933, p 209

8. M.J. Armbruster, Am. Chem. Soc., Vol 65

(No. 11), 1943, p 1043

9. W. Eichenbauer, H. Kunzig, and A. Pebler,

Z. Metallkd., Vol 49, 1958, p 220225

10. M.Z. Hill and E.W. Johnson, Trans. AIME,

Vol 221, 1961, p 622629

11. Th. Heumann, D. Primas, and Z. Naturforschung, Vol A-212, 1966, p 260269

12. V.I. Saly, R.A. Riabov, and P.V. Geld, Fiz.

Met. Metalloved., Vol 35 (No. 1), 1973, p

119122 (in Russian)

13. H. Schenk and K.W. Lang, Arch. Eisenhuttenwes, Vol 37, 1966, p 739744

14. J.R.G. Da Silva, S.W. Stafford, and R.B.

McLellan, J. Less-Common Met., Vol 49,

1976, p 407420

15. J.R.G. Da Silva and R.B. McLellan, J.

Less-Common Met., Vol 50, 1976, p 15

16. C.A. Zapffe, Metals Handbook, American

Society for Metals, 1948, p 1208

17. J. Hewitt, Hydrogen in Steel, London

Chapman, Ed., 1962

18. T.N. Stross and F.C. Tomkins, Chem.

Abstr., Vol 50, 1956, p 230

19. W. Eichenauer et al., Z. Metallkd., Vol 49

(No. 5), 1958, p 220

20. E.W. Johnson and M.L. Hill, J. Met., Vol

11 (No. 9), 1959, p 572

21. A.E. Schuetz and W.D. Robertson, Corrosion, Vol 13 (No.7), 1957, p 437

22. R.C. Frank, J. Appl. Phys., Vol 29 (No. 6),

1958, p 892

23. M. Smialowski, Neue Hutte, Vol 10, 1957,

p 621

24. J. Hobson, J. Iron Steel Inst., Vol 189 (No.

4), 1958, p 315

Met. Finish., Reinhold, N.Y., 1961 p 213

26. F. Kazinczy, Jernkontorets Ann., Vol 1,

1955, p 885

27. C.A. Zapffe, J. Iron Steel Inst., Vol 156,

1947, p 123

28. B.A. Kolachev, Hydrogen Embrittlement

of Metals, Metallurgia, Moscow, 1985, p

217 (in Russian)

29. V.I. Shapovalov, Hydrogen Influence on

the Structure and Properties of Iron-Carbon Alloys, Metallurgia, Moscow, 1982, p

232 (in Russian)

30. E. Lunarska, A. Zielinski, and M. Smialowski, Acta Metall., Vol 25 (No. 1),

1977, p 161171

31. J.P. Hirth and H.H. Johnson, Corrosion,

Vol 32 (No. 1), 1976, p 326

32. U.I. Archakov, Hydrogen Resistance of

Steel, Metallurgia, Moscow, 1979, p 152

(in Russian)

33. G.V. Karpenko and R.I. Kripyakevich, Hydrogen Influence on Structure and Properties of Steels, Metallurgia, Moscow,

1962, p 198 (in Russian)

34. L.S. Moroz and B.B. Chechulin, Hydrogen

Brittleness of Metals, Metallurgia, Moscow, 1967, p 255 (in Russian)

35. Hydrogen in Metals, Proc. 2nd International Congress, 1977 (Paris), Pergamon

Press, Oxford, 1978

36. C.J. McMahon and V. Vitek, Acta Metall.,Vol 27 (No. 4), 1979, p 507513

37. M. Cornet and S. Talbot-Besnard, Metall.

Sci., Vol 7, 1978, p 335

38. J.P. Hirth, Metall. Trans. A, Vol 11 (No.

6), 1980, p 861890

39. G.M. Pressouyre, Metall. Trans. A, Vol 10

(No. 10), 1979, p 15711973

40. A.G. Kumnick and H.H. Johnson, Acta

Metall.,Vol 28 (No. 1), 1980, p 3339

41. C.D. Beachem, Metall. Trans., Vol 3,

1972, p 437

42. A. Troiano, Met. Prog., Vol 71, 1960, p

112

43. R. Oriani and P. Josephic: Acta Metall.,

Vol 22, 1974, p 1065

44. L. Engel and H. Klingel, Rasterelectronenmikroskopishe Untersuchungen von Metallschaden,

Karl

Hanser

Verlag,

Munchen-Wien, 1982

45. J.M. Hodge, M.A. Orehoski, and J.E.

Steiner, Trans. AIME, Vol 230 (No. 5),

1964, p 1182

46. J. Mussati and A. Reggiori, Met. Prog.,

Vol 53 (No. 7), 1936, p 51

47. P. Bardenheuer, Stahl u. Eisen, Vol 45,

1925, p 1782

48. W. Eilender and H.Z. Kiessel, VDI, 1932,

Vol 76, p 729

49. E.A. Klausting, Metallurgist, Vol 5, 1937,

p 21

50. E. Houdremont, Handbuch der Sonderstahlkunde, Verlag Springer, Berlin, 1934

51. V.V. Zabilsky et al., Fiz. Met. Metalloved.,

Vol 83 (No. 2), 1997, p 160175 (in Russian)

Steel Parts, Oborongiz, Moscow, 1955 (in

Russian)

53. K. Oniki, Kinzoku, Vol 46 (No. 4), 1976,

p 1518

54. M.A. Balter, Fractography as Tool for Diagnosis of Fracted Parts, Maschinostroenie, Moscow, 1987 (in Russian)

55. Y. Nakatani, K. Utsubo, and K. Yamada,

Trans. Jpn. Soc. Mech. Eng. A, Vol 64 (No.

626), 1998, p 25612566

56. M.M. Shved, Change of Operating Properties of Iron and Steel under Hydrogen

Action, Naukova dumka, Kiev, 1985 (in

Russian)

57. A.Ye. Andreykiv, V.V. Panacjuk, and V.S.

Charin, Fiz. Met. Metalloved., (No. 3),

1978, p 323 (in Russian)

58. E.A. Steigerwald, Trans. AIME, Vol 215,

1959, p 1048

59. L.M. Bilyi and M.M. Shved, Fiz. Met. Metalloved., (No. 6), 1978, p 106107 (in

Russian)

60. I.S. Yablonsky, Fiz. Khim. Mekh. Mater.,

(No. 6), 1979, p 4756 (in Russian)

61. F.F. Azhogin, A.V. Sacharov, and S.S. Ivanov, Fiz. Khim. Mekh. Mater., (No. 3),

1979, p 3538 (in Russian)

62. K. Soeno, Trans. Iron Steel Inst. Jpn., Vol

21 (No. 7), 1981, p 469476

63. K. Soeno and K. Taguchi, Trans. Iron Steel

Inst. Jpn., Vol 21 (No. 9), 1981, p 618

623

64. K. Soeno and M. Tsuhiya, Trans. Iron

Steel Inst. Jpn., Vol 22 (No. 11), 1982, p

848853

65. V.A. Maritchev, Fiz. Khim. Mekh. Mater.,

(No. 5), 1981, p 2429 (in Russian)

66. S. Yamasaki and T. Takahashi, J. Iron

Steel Inst. Jpn., Vol 83 (No. 7), 1997, p

460465

67. O. Tamura and M. Takemoto, J. Iron Steel

Inst. Jpn., Vol 83 (No. 8), 1997, p 532

537

68. C.L. Briant and S.K. Banerji, Ed., Treatise

on Materials Science and Technology,Vol

25, Embrittlement of Engineering Alloys,

Academic Press, 1983

69. G.G. Hancock and H.H. Johnson, Trans.

Met. Soc. AIME, Vol 236, 1966, p 513

70. H. Liebowitz, Ed., Fracture, Vol 3, Mir,

Moscow, 1971 (in Russian)

71. V.S. Savchenko and K.A. Yutshenko, Fiz.

Khim. Mekh. Mater., (No. 1), 1982, p 54

57 (in Russian)

72. V.N. Zikeev, MiTOM, (No. 5), 1982, p 8

23 (in Russian)

73. G.G. Irzhov, S.A. Golovanenko, and T.K.

Sergeeva, Fiz. Khim. Mekh. Mater., (No.

3), 1982, p 8993 (in Russian)

74. S.A. Golovanenko and T.K. Sergeeva,

Steel, (No. 7), 1984, p 7378 (in Russian)

75. L.M. Utevsky, Ye.E. Glikman, and G.S.

Kark, Reversible Temper Brittleness of

Steel and Iron Alloys, Metallurgia, Moscow, 1987 (in Russian)

76. A.I. Kovalev, G.V. Tsherbedinsky, Modern Methods of Investigations of Metal and

Alloy Surface, Metallurgia, Moscow,

1989 (in Russian)

77. V.V. Zabilsky, Metal. Termoobrabotka,

(No. 1), 1987, p 2432 (in Russian)

78. L.I. Gladshtein and N.I. Kelberin, Zavodskaya Laboratoriya, (No. 1), 1995, p 38

41 (in Russian)

79. A.Yu. Kazanskaya, M.A. Smirnov, and

V.V. Zabilsky, Fiz. Met. Metalloved., (No.

8), 1990, p 201203 (in Russian)

80. M.L. Bernshtein and V.A. Zaymovsky,

Mechanical Properties of Metals, Metallurgia, Moscow, 1979 (in Russian)

81. V.V. Zabilsky, O.A. Bartenev, V.V. Velitchko, and S.M. Polonskaja, Fiz. Met.

Metalloved., (No. 4), 1986, p 793800 (in

Russian)

82. V.V. Velitchko, G.M. Micheev, V.V. Zabilsky, and D.I. Maleev, Fiz. Khim. Mekh.

Mater., Vol 27 (No. 1), 1991, p 112114

(in Russian)

83. V.V. Zabilsky and V.V. Velitchko, Metal.

Termoobrabotka, (No. 4), 1993, p 915 (in

Russian)

84. V.V. Velitchko, V.V. Zabilsky, and G.M.

Micheev, Fiz. Met. Metalloved., (No. 1),

1995, p 130140 (in Russian)

85. I.S. Toydorova and V.V. Zabilsky, Fiz.

Met. Metalloved., (No. 7), 1991, p 511

(in Russian)

86. V.V. Zabilsky, V.V. Velitchko, and S.G.

Iliyina, Fiz. Met. Metalloved., Vol 80 (No.

12), 1995, p 108118 (in Russian)

87. V.V. Zabilsky, V.V. Velitchko, F.Z. Gilmutdinov, V.G. Makletzov, and Yu.P. Vorobiov, Fiz. Met. Metalloved., Vol 83 (No.

2), 1997, p 160175 (in Russian)

88. A.V. Vaynman, R.K. Melechov, and O.D.

Smiyan, Hydrogen Embrittlement of Elements of High Pressure Kettles, Naukova

dumka, Kiev, 1990 (in Russian)

89. I.K. Pohodnya, Electron and Ion Spectroscopy of Solids (No. 2), L. Fiermans, J.

Vennik, and W. Dekeyser, Plenum Press,

1988, p 3031 (in Russian)

90. V.V. Zabilsky and V.V. Velitchko, Fiz.

Khim. Mekh. Mater., Vol 25 (No. 3), 1989,

p 1721 (in Russian)

91. V.V. Zabilsky, V.V. Velitchko, and A.Yu.

Kazanskaja, Fiz. Met. Metalloved., (No. 3),

1990, p 143149 (in Russian)

92. V.I. Voronenko, Zashch. Met., Vol 33 (No.

2), 1997, p 132143 (in Russian)

93. O.N. Romaniv and G.N. Nikifortchin, Mechanics of Corrosion Failure of Structural

Alloys, Naukova dumka, Kiev, 1986

94. N.I. Volgina and T.K. Sergeeva, Materialovedenije, (No. 12), 1998, p 2130 (in

Russian)

95. L.N. Petrov, V.N. Tischenko, and A.S.

Tchistov, Fiz. Khim. Mekh. Mater., (No.

4), 1979, p 2025 (in Russian)

96. V.I. Nikitin, Fiz. Khim. Mekh. Mater., Vol

25 (No. 1), 1989, p 31 (in Russian)

97. F.R. Ford, Mech. Behav. Mater. Proc. 3rd

Int. Conf., 1979 (Cambridge U.K.), Pergamon Press, Toronto, 1979, (No. 2), p 431

A.Z. Student, Fiz. Khim. Mekh. Mater.,

Vol 19 (No. 5), 1983, p 3 (in Russian)

99. V.P. Voronin, A.Yu. Schulte, and Yu.V.

Tomkin, Fiz. Khim. Mekh. Mater., Vol 19

(No. 5), 1983, p 114 (in Russian)

100. A. Mazur and B. Pawlowski, Corros. Sci.,

Vol 26 (No. 1), 1986, p 7

101. T.J. Kaneko, J. Iron Steel Inst. Jpn., Vol

72 (No. 13), 1986 p 1472

102. J. Kuniya, I. Masaoka, and R. Sasaki, J.

Pressure Vessel Technol., Vol 107 (No. 4),

1985, p 430

103. W. Diton and P. Huiying, Proc. Int. Congr.

Met. Corros., 1984 (Toronto), (No. 3),

1984, p 573577

104. L.N. Petrovm, Fiz. Khim. Mekh. Mater.,

(No. 4), 1984, p 1621

105. W.W. Gerberich, R.N. Jones, M.A. Friesel,

and A. Nozue, Mater. Sci. and Eng. A, Vol

103 (No. 1), 1988, p 185

106. A. Nozue, T. Kishi, and R. Horiuchi, J.

Iron Steel Inst. Jpn., Vol 67 (No. 10),

1981, p 1787

107. G.B. Muravin, L.M. Lezvinscaja, N.O.

Makarova, and G.S. Pavlovskaja, Defektoskopiya, (No. 2), 1990, p 1828

108. B.I. Voronenko, Zashch. Met., Vol 33 (No.

5), 1997, p 472488 (in Russian)

109. B.I. Voronenko, Metal. Termoobrabotka,

(No. 9), 1990, p 58 (in Russian)

110. C.S. Carter, Corrosion, Vol 25 (No. 10),

1969, p 423

111. B.I. Voronenko, Zashch. Met., Vol 33 (No.

6), 1997, p 573589 (in Russian)

112. P. McIntyre and T.M. Banks, Eur. Offshore Steels Res. Seminar, VIP 19/1, 1978

(Cambridge), Abington Publishing, 1980

113. D. Huerta and K.E. Heusler, Werkst. Korros., Vol 45 (No. 9), 1994, p 489

114. O.N. Romaniv, G.N. Nikifortchin, A.Z.

Student, Fiz. Khim. Mekh. Mater., (No. 2),

1985, p 2031 (in Russian)

115. B. Pawlowski, Werkst. Korros., Vol 37

(No. 8), 1986, p 448

116. R.O. Ritchie, M.H.C. Cenedo, and V.F.

Zackay, Metall. Trans. A, Vol 9 (No. 1),

1978, p 3540

117. M. Turdy, V. Suchanek, L. Hyspecka, and

K. Mazanec, Adv. Res. Strength and Fract.

Mater./4th Int. Conf. Fract. 1977 (Waterloo, NY), 1978, p 255

118. D. Zhou, W. Cao, and H. Guo, Heat Treat.

Met., (No. 1), 1992, p 8

119. Y. Hirose, K. Tanaka, and K. Okabayashi,

22nd Congr. Mater. Res., 1978 (Kyoto, Japan), Proc. Kyoto, 1979, p 76

120. Y. Hirose and K. Tanaka, Met. Corros./8th

Int. Congr./7th Congr. Eur. Fed. Corros.

(111th event), 1981 (Mainz, Germany),

Frankfurt/M., European Federation of

Corrosion, 1981, (No. 1), p 553

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

A.Z. Student, Fiz. Khim. Mekh. Mater.,

Vol 17 (No. 3), 1981, p 24 (in Russian)

122. W. Cao and X. Lu, Mater. Sci. and Eng.

A, Vol 110 (No. 151), 1989, p 165

123. W. Radeker and B.N. Mishra, Werkst. Korros., Vol 21 (No. 9), 1970, p 692

124. F.W. Fraser and E.A. Metzbower, Fractogr. and Mater. Sci. Symp., 1979 (Williamsburg, VA), Philadelphia, 1981, p 51

125. K. Rippstein and H. Kaesche, Corros. Sci.,

Vol 29 (No. 5), 1989, p 517

126. Z. Szklarska-Smialowska and E. Lunarska,

Werkst. Korros., Vol 32 (No. 11), 1981, p

478

127. H. Hanninen, Corrosion, Vol 41 (No. 7),

1990, p 563

128. A.G. Gareev, I.G. Abdullin, and G.I. Abdullina, Gazovaja Promischlennost, (No.

11), 1993, p 29 (in Russian)

129. A.F. Matvienko, A.V. Baldin, A.P. Grigorjev, V.A. Kanaikin, V.G. Rjibalko,

V.V. Sagaradze, and Yu.I. Philippov, Fiz.

Khim. Mekh. Mater., Vol 86 (No. 2), 1998,

p 139146 (in Russian)

130. R.N. Parkins, Corrosion, (No. 50), 1994,

p 384

131. T.K. Sergeeva, E.P. Turkovskaja, N.P. Michailov, and A.I. Tchistjakov, State of

Problem of Stress-Corrosion in SNG

Countries and Abroad, IRZ RAO, Gazprom, Moscow, 1998 (in Russian)

132. R.N. Parkins, C.S. Dell, and R.R. Fessler,

Corros. Sci., Vol 24 (No. 4), 1984, p 343

374

133. Public Inquiry Concerning Stress Corrosion Cracking on Canadian Oil and Gas

Pipelines, Report NEB, 1996, No. 11, MH295

134. M. Urednicek, S. Lambert, and O. Vosikovsky, Monitoring and Control Proc. Int.

Conf. on Pipeline Reliability (Calgary,

Canada), 1992, N VII-2

135. T.K. Sergeeva, Proc. Int. Symp. on Pipe

Line Safety and Aging, CNIICHERMET,

1995 (Moscow), Vol 1, p 139164

136. O.M. Ivantzov, Stroitelstvo Truboprovodov, No. 6, 1993, p 1016 (in Russian)

137. B.G. Mazel, Stroitelstvo Truboprovodov,

No. 7, 1993, p 3639 (in Russian)

138. D. Rudy, Prev. of Pipeline Corrosion

Conf., 1994 (Houston, TX), p 1721

139. V.V. Zabilsky and V.I. Lebedev, Tchernaja Metallurgija. Bjulleten NTI, No. 2,

1991, p 1329 (in Russian)

140. R.M. Nikonova and V.V. Zabilsky, Fiz.

Metal. Metalloved., Vol 76 (No. 8), 1993,

p 99106 (in Russian)

141. A.F. Matvienko, A.V. Baldin, A.P. Grigorjev, V.A. Kanaikin, V.G. Rjibalko,

V.V. Sagaradze, and Yu.I. Philippov, Fiz.

Khim. Mekh. Mater., Vol 86 (No. 2), 1998,

p 147155 (in Russian)

142. G. Libowitz, The Solid-State Chemistry of

Binary Metal Hydrides, Vol 6, 1965, p 11

143. V.N. Zikeev, Metal. Termoobrabotka, Vol

5, 1982, p 1823 (in Russian)

144. V.I. Shapovalov, Hydrogen Influence on

the Structure and Properties of Iron-Carbon Alloys, Metallurgia, Moscow, 1982, p

232 (in Russian)

145. R. Oriani, Acta Metall., Vol 24, 1977, p

979988

146. M. Vornet, Mem. Sci. Rev. Met., Vol 74

(No. 5), 1977, p 307316

147. A.P. Guliaev, Clean Steel, Metallurgia,

Moscow, 1975, p 184 (in Russian)

148. V.G. Starchak, A.B. Kuslitsky, U.A.

Kliachko, and I.A. Tamarina, Metal. Termoobrabotka, No. 4, 1977, p 1113 (in

Russian)

149. A. Joshi, Corrosion, Vol 34 (No. 2), 1978,

p 4752

p 3437

151. Z. Szklarska-Smilowaska, Werkst. Korros., Vol 32 (No. 11), 1982, p 478485

152. G.M. Itskovich, Deoxidation of Steel and

Modification of Nonmetallic Inclusions,

Metallurgia, Moscow, 1981, p 296 (in

Russian)

153. E. Snape, Corrosion, Vol 24 (No. 9), 1968,

p 261282

154. H. Gondo and M. Iino, Nippon Steel Corporation Studies on Materials for Steel

Pipes for Service in Wet Sour Gas Medium, Technical Report B2 (No. 11), Nippon Steel Corporation, 1979, p 71

155. E. Stephenson, Metall. Trans. A, Vol 11

(No. 3), 1980, p 517524

156. S. Shichi, M. Takeshi, and D. Utaka, Japanese Application kl 12A41 (C23Fg/02),

No. 54, 1979

G. Totten, M. Howes, T. Inoue, editors, p89-98

DOI: 10.1361/hrsd2002p089

All rights reserved.

www.asminternational.org

Residual Stress

H.W. Walton, Consultant

readily available for the determination of residual stresses. Nor do they have the time or inclination to use these procedures for checking production parts. Mathematical derivations such as

the one for plate by the Treuting and Read

method are also very time consuming (the

method is described at the end of this article).

To determine if residual stresses from prior

processing are the cause of uncontrollable distortions during processing of thin section parts,

it is a simple matter to slit an as-received piece

and note the amount of deflection resulting from

relaxation of residual stresses, for example, the

net opening of the saw cut. If no deflection occurs, then it can be safely assumed that the problem lies with the processing parameters and not

with the material (e.g., high feeds and speeds or

high chucking pressures during machining, or

surface overheating during grinding resulting

from insufficient wheel dressing).

These simple techniques, sometimes known

as dissection, are old but still very useful. However, because slitting a component is a destructive method, there are limitations, particularly if

the component in question is large.

This article provides a low-cost, easy method

of determining if residual stresses are the cause

of component distortion during manufacture.

Residual Stress

To understand the type of stress that is being

measured, a distinction first needs to be made

between microstresses and macrostresses extending over large volumes. Microstresses resulting from changes occurring at the atomic

level are equilibrated over very small volumes

and cannot be measured by use of deflection

methods. Macrostresses, on the other hand, are

primarily the result of forming operations or

thermal imbalances during casting, welding, heat

treating, and so on.

If the stress induced by nonuniform cooling

exceeds the yield strength of the material, plastic

deformation occurs. On cooling to a uniform am-

restrained by the surrounding material is now

placed under significant elastic stress. This

would be considered macrostress and is the

stress most likely to cause deformation during

subsequent processing.

Consider the simple example of bending a bar

to form a permanent set. The surface material at

the elbow or concave surface undergoes permanent plastic deformation while the inside diameter is undergoing compressive plastic deformation. When the applied force is removed, the

underlying elastically stretched material (at the

elbow) is now prevented from relaxing completely. The result is that the plastically deformed surface layers of the elbow are placed

under elastic compression by the underlying

elastically stretched material. The opposite is

true for the inside diameter. It follows that if the

bar is placed in service in this condition, corrosion and development of cracks are more likely

to occur at the surface of the inside diameter.

In general terms, if a surface undergoes permanent plastic deformation by cold rolling, shot

peening, machining, thermal processing, and so

on, which in turn results in the underlying material being elastically stretched or compressed,

the material plastically deformed by tensile

forces is left in elastic compression. Material

plastically deformed by compressive forces is in

elastic tension.

The influence of nonuniform residual stress is

sometimes dramatically illustrated by fishtailing of large hot-rolled steel plate. As the plate

cools after rolling, transformational stresses are

developed and, if an extensive central lamination

is present due to inadequate discard or casting

problems, these stresses can cause the plate to

separate into two halves (Fig. 1).

The stress distributions present before and after the occurrence of a fishtail are shown in the

sketch. In the as-rolled plate, tensile stresses

(tv stress) on the surface are trying to contract

the surface while compressive stresses (tv

stress) are trying to expand the plate core. How

this stress distribution develops is illustrated in

Fig. 2. After fishtailing, the two halves of the

plate attain a new equilibrium stress distribution.

surface will be cooling faster than the core material, and thermal stresses developed at this

stage will be relaxed by the plastic deformation

of the low strength austenite. At lower temperatures, transformation to ferrite and pearlite begins at the surface (Fig. 3), causing complex interactions and a variety of stress conditions.

Because of the relatively low yield strength of

the various phases and the role of transformation

induced plasticity, the final stresses are low.

However, over a large surface area even low residual stresses can have a profound effect, particularly in the presence of large defects such as

a lamination.

If not fixtured in a vertical position, severe

distortion of quenched flat steel sections may occur. This is due to differences in the cooling rate

and transformation behavior between the top and

bottom surface.

In a carburized product, because of the compositional changes causing transformation to

progress from the inside out and the considerably

higher-strength case, the final residual stress profile is reversed and a beneficial high-compressive

surface stress is developed (Ref 1). Figure 4

shows the development of these residual stresses

during the heat treatment of a carburized component.

It is very important to avoid heating small,

isolated areas on the surface of hardened steel.

Fig. 1

stress

steels. tv, tensile stress; tv, compressive

Examples of where this may occur are:

Components ground with inadequate cooling

heads)

Using a gas flame to enlarge a hardened component for assembly on a shaft

Electrical arcing from inadequately earthed

drive motors

Accidental arcing during magnetization for

magnetic particle inspection

The thermally damaged area is restrained from

expanding by the surrounding material and becomes plastically deformed under compression.

Consequently, on cooling to ambient temperature, the surface is left in a tensile condition, and

cracking ensues either immediately or after the

component is placed in service (more worrying).

Heating at temperature above the original tempering temperature may also lead to a contraction.

Shot peening appears to be an anomaly. At

first sight, the impact of the shot particles seems

to cause plastic deformation in compression. It

should follow that the surface is in tension following the peening treatment. However, the ad-

plastic extension of the surface must be considered. The overall surface undergoes a plastic extension, thus putting the underlying layers in tension. Consequently, on completion of the

shot-peening operation, the surface is restrained

in a beneficial compressive state.

The maximum level of residual stress that can

be generated is approximately equal to the elastic

limit or yield strength of the material measured

under the same stress condition and for the same

cross section. The existence of triaxial stresses

in heavy section components has a major influence on the ease of plastic flow (plane-strain conditions). If surrounding material inhibits plastic

flow by significantly limiting shear stresses, the

levels of residual tensile stress may become

high. Catastrophic fracture occurs when the tensile stress level exceeds the cohesive strength of

the steel.

Manufacturing Implications of

Residual Stresses

There are many examples that show how relaxation or redistribution of residual stresses

contributes to high processing costs. One example is the need to use multiple grinding passes

to avoid distortion in thin section components.

A frequent problem often wrongly attributed to

the heat treatment department is heat treatment

distortion from prior residual stresses in the raw

material.

Premature failure due to harmful residual surface tensile stress in the manufactured product is

to be avoided at all costs.

On the other hand, the presence of residual

stresses may be advantageous, as in the example

of compressive stresses from case hardening and

shot peening.

Using residual stresses to intentionally deform

a component can be illustrated by the use of shot

peening to shape the aircraft fuselage, by using

a peen hammer to straighten a long shaft, or by

simply hot spotting to bend a simple bar of steel.

For example, a small area on one side of a

steel bar is heated rapidly to a temperature, T, of

approximately 600 C (1100 F) (Ref 2). If free

to expand, the heated disc-shaped region increases in length, l, by:

l T

14 600

Cooler outside

surface

Hotter core

Transformation

Consider a slice through half the thickness of a plate.

At the onset of the cooling, both the surface and

the core are austenitic and stress free.

Transformed

Not transformed

The surface will transform first, acccompanied by

an expansion. Restraint by the underlying material

will place the transformed element in compression "C".

Although the adjoining austenitic element "stretches"

and plastically yields, it will be in tension "T".

CT

in lm/mm K. However, the surrounding cooler

material restrains expansion. Because of the relatively low yield strength at 600 C (YS600) of

approximately 30 MPa (4.5 ksi), the heated steel

deforms plastically under compression, increasing the thickness of the heated disc. On cooling,

the disc contracts the same amount and, assuming no stress relaxation occurred, a tensile stress,

r, is now developed in the disc:

r (E l) YS600

(200 103 8.4 103) 30

1650 MPa (240 ksi)

expansion in the absence of any restraining effect

by the surrounding material

undergoes transformation and expands further. This

induces tensile stresses in the material on either side,

but due to the restraining influence of surrounding

material, compressive stresses are induced in the

newly transformed steel.

TCT

Fig. 2

TCT

section, leaving behind a surface that is in tensile.

The magnitude of stress is determined by the relative

yield strength of the phases. Because of the TRIP

effect (transformation induced plasticity), the plasticity

of the transforming steel is considerably higher.

stress in the disc cannot be greater than the yield

strength of about 500 MPa (73 ksi). Therefore,

either the disc stretches plastically, or the bar

bends elastically to balance the stresses (or

both).

The manufacture of seamless high-carbonbearing steel tube involves hot rolling, sizing,

straightening, and spheroidize annealing. Each

stage of manufacture introduces various degrees

of residual stress. If residual stress levels are

higher than normal, difficulties may be experienced during subsequent machining and heat

treatment of the rings. As seen from the previous

calculation, the higher the yield strength of the

material, the greater the potential for high levels

of residual stress. As thermal and mechanical

treatment of steel increases the strength of a material, potential problems with residual stress

also increase. For example, a small area of thermal damage on the surface of a hardened high-

carbon steel (e.g., electric arc discharge) may result in a level of triaxial residual stress close to

the cohesive strength of the material, possibly

leading to immediate or delayed cracking.

Although the stress may be multidirectional,

hoop stress that arises at several stages of manufacture is the primary cause of many of the

problems. When a length of tube is parted off

and slit in a longitudinal direction, any hoop

stress present tends to open the slit (Fig. 5).

Opening of the slit indicates compressive stress

in the inside diameter (ID) of the tube and tensile

stress on the outside diameter (OD). On rare occasions, the slit may close, indicating the reverse

condition. The stress is not always uniform along

the length of the tube. Stress measurements have,

on occasions, indicated a cyclic variation corresponding to differences in cooling of areas in

contact with the cooling bed cross ties.

Shot peening intensity is monitored by using

a method developed by J.O. Almen of General

Motors Company. In this method, 75 mm long

by 18.75 mm wide (3.00 by 0.75 in.) strips of

1070 spring steel are quenched and tempered to

a deep-blue oxide finish (blue tempered) for a

hardness of 45 to 50 Rc and exposed on one side

to the same shot intensity as the component undergoing treatment (Ref 3). Three different

thicknesses are used: N, 0.79 mm (0.031 in.);

A, 1.30 mm (0.051in.); and C, 2.38 mm

(0.0938 in.) to allow for differences in degree

and shot peening required and materials being

treated. For example, peen forming of an aluminum alloy requires considerably less peen intensity than a carburized gear tooth. During treatment, the Almen strip assumes a concave shape.

The lift height is proportional to the level of

compressive stress developed in the upper sur-

the material has reached saturation; that is, when

the lift height increases by no significant amount

and the compressive stress in the surface layers

corresponds closely to the elastic limit of the

steel strip (Fig. 6).

Consider the residual stress distribution

through the Almen strip. Deflection of the strip

occurs in order to relieve some of the high surface compressive stress. The peened surface area

endeavors to expand by bowing in order to nullify the compressive stress. The magnitude of

bowing is limited by the restraint of the remainder of the material, and equilibrium is reached

when the remaining compressive stress is in balance with the elastic compressive stress developed in the lower surface. The depth and intensity of the compressive layer is proportional to

the lift height at saturation. The Almen strips are

primarily used to ensure that the process is in

control (shot condition, uniformity in application, impeller operation, etc.). A photograph of

three typical strips is shown in Fig. 7.

Metal Improvement Company, Inc., in collaboration with ENSAM, a French advanced engineering school, has developed a software program called PeenStress (Metal Improvement

Co., Inc., Paramus, NJ) that is used to assist in

shot peening callouts (Ref 4). The user selects a

material from a library of about 80 materials,

then selects a shot size and shot intensity and

inputs some basic geometry considerations. Figure 8 is a curve generated on a chromium-silicon

spring wire shot peened with a hardened, 0.023

in. shot to a 10 A intensity. A 10 A intensity

equates to a 0.25 mm (0.010 in.) arc height on

the A-strip.

One of the older standard tests for evaluating

Ar3

Ar1

Austenite

7). The amount of distortion (deflection) of a

quenched test piece is measured by the change

in the gap width (Fig. 9).

Stresses from Deflection Data

Sectioning to allow relaxation of residual

stress in actual components where stresses are

thought to be present can be performed by several methods, from simple saw cutting to the

more sophisticated compliance method (Ref 8

10). In the latter method, the residual stress profile is calculated from the strains caused by introducing a cut of progressively increasing depth

into the component. The strains are measured by

using suitably positioned strain gages cemented

to the surface adjacent to the cut. The cutting is

performed by using various techniques, however, electrical discharge machining is the preferred method. Usually, two computer-based approaches are used to analyze the data. These are

the forward and inverse solution. The forward

solution derives the measurable strains (compliance functions) that develop from introducing a

successively deeper slot into a part containing an

arbitrary known stress distribution. The inverse

solution develops the original residual stress distribution that best matches the actually measured

strains.

Similar methods based on drilling small holes

in the stressed material have been around for

some years. Strain relaxation is measured using

strain gages or photoelastic coatings. Strain

gages have been used to measure strain relaxation in bevel gears following successive removal

of layers by electrochemical machining (Ref 11).

Saw cutting or slitting is an easy quality control test that gives a global overview of the state

of residual bulk hoop stress in rings. Interestingly, such a technique was used recently to validate the measurements of residual stress levels

in railway wheels using electromagneticacoustic transducers (Ref 12).

Ferrite

Pearlite

Temperature

Core

Bainite

Surface

Martensite

Time, T

Schematic representation of the relative transformation at the surface and in the core of the mild steel plate. At

time, T, the core is still austenitic while the surface has already transformed to a ferrite-pearlite structure. Ar3

and Ar1, upper and lower transformation temperatures, respectively, on heating (refroidissant) a hypoeutectoid steel

Fig. 3

Simple Cases Based on the Saw-Cut

Methods

Definitions of the symbols used in the following derivations are given in Table 1.

To develop the mathematical algorithms for

interpreting deflection measurements after slitting a simple shape (plate, round bar, or tube),

simple beam theory is used (Fig. 10).

The basic formula for the state of affairs at

any point (x) along a beam is (Ref 13):

M

r

E

I

c

R

(Eq 1)

beam is subjected at x. Bending moment M is

%C

decreases, the transformation temperature during

quenching increases (Ms). Although the cooling rate

is less below the surface, transformation occurs from

the inside out.

the onset of quench, both the surface and the core

are austenitic and stress free.

Start of transformation

Even though the cooling rate is lower than at the

surface, the low carbon core will begin to transform,

starting first below the case and progressing inward.

The adjoining austenitic case element will plastically

expand outward to accommodate the resulting tensile

stress.

kgf m). If M is not obvious from the first principles for a given case, it may be obtained from

the data given subsequently. I is the moment of

inertia of the section of the beam at x, usually in

inches. Moments of inertia are treated in textbooks of elementary mechanics. The moment of

inertia in question is that about the neutral axis

which, in most beam problems, will be a line

through the center of gravity of the section. If

the I of the section is not already known, it is

usually easily calculated by applying one of the

following formulae.

For a rectangle:

bt3

12

expansion in the absence of any restraining effect

by the surrounding material

C CC T T

+tv

the innermost element of the case transforms. The

resulting expansion of the already extended case

element creates tensile stresses in both remaining

austenitic case and adjoining core.

tubes, H-beams, and channels, can be calculated

by subtracting the I of the empty areas.

For a circular cross section:

higher carbon case and the resulting expansion of the

already plastically stretched elements result in high

compressive stresses. Consequently, the underlying

transformed core is pulled into tension.

near the surface of the case can reach levels of

300/500 N/mm2 and is one of the primary reasons for

carburizing.

tv

Fig. 4

(Eq 2)

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

pr 4

4

(Eq 3)

be calculated by subtraction, where: r is stress

in the material. c is distance from the neutral

axis. In symmetrical sections this is the distance

from the midpoint or center of gravity of the section. E is Youngs modulus. (In thick cross sections and/or high-strength steels, this should be

corrected for Poisson contraction where m is

Poissons ratio [0.3 for steel].)

E

E

1 m2

(Eq 4)

bends under load.

To calculate the stress in a simple beam:

r

Mc

I

(Eq 5)

stress is greatest:

Fig. 5

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

diam; OD, outside diam

bt3

12

(Eq 6)

t

2

(Eq 7)

so

rmax 6

M

bt2

(Eq 8)

EI

R

Units

is four times as strong.

The distribution of residual stress in actual

components is unlikely to be linear. However,

for the subsequent approximate analysis, the

beam stress is assumed to vary linearly through

the section.

In sheet and bar, as the sheet fishtails or is cut

in a central planar direction, the bending moment

created by the residual stress is released (Fig.

11). The bending moment may be expressed as:

M

Table 1

(Eq 9)

Symbol

Description

M

I

Bending moment

Moment of inertia

r

c

Stress in material

Distance from neutral axis

E

R

L

r

t

d

c

Youngs modulus

Radius of curvature of beam or displaced section

Length of curved beam

Radius of round bar cross section

Thickness of beam/plate/tube wall

Measured deflection

Poissons ratio

SI

English

Notes

NM

m4

lbf in.

in.4

Pa

m

psi

in.

Pa

m

m

m

m

m

Dimensionless

psi

in.

in.

in.

in.

in.

...

...

I bt3/12 for rectangle (where b and

t are the dimensions of the

cross section) I pr 4/4

for circle

...

In a symmetrical section this

is the distance from the midpoint or center of gravity

...

...

...

...

...

...

0.3 for steel

given that:

c

t

4

(Eq 10)

30 0.015 in.

where t is plate thickness, and assuming the distribution of the residual stress that resulted in the

fishtail varies linearly over the half-thickness of

the sheet, the maximum longitudinal stress at the

surface is given by:

r

Mc

I

(Eq 11)

0.0310.001 in.

Measuring dial

0.0510.001 in.

N strip

Peening nozzle

A strip

C strip

Almen strips

0.09380.001 in.

0.7450.750 in.

Shot stream

4-6 in.

r

Et

4R

Hardened

ball supports

10/32

screws

(Eq 12)

Arc height

3.0 in.

If the deflection, d, is small compared to the radius of curvature, R, R may be expressed in terms

of the deflection, d, and the length of the curved

surface, L, by:

R

L2

2d

3.0 in.

Holding

fixture

0.75 in.

Strip removed. Residual

stresses induced arching

Fig. 6

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

Fig. 7

(Eq 13)

Referring to Fig. 12, the two radii O1 and O2

are drawn as shown, and the tangents 1,3 and 2,4

are drawn to these radii. A chord is drawn between the points of tangency 1 and 2 and the line

2,3 is drawn perpendicular to the chord at 2. The

two isosceles triangles RLk and mdn are similar

and therefore:

R

m

L

d

(Eq 14)

or

R

Lm

d

(Eq 15)

arc 1, 2, or

Arc height

Strip mounted for

height measurement

1

L

2

(Eq 16)

R

L2

2d

(Eq 17)

According to Eq 12:

Et

r

4R

(Eq 18)

it follows:

r

slit the tube longitudinally with a carborundum

cutoff wheel (using plenty of coolant to avoid

heating of the surfaces) and measuring the

change in diameter of the tube (Fig. 13b):

Etd

2L2

r Et

1

0

1

D1

parted from the tube or machined, is cut axially through the section.

If unknown, the saw curf width (width of saw

(Eq 22)

0.600

0

after slitting.

Alternatively, the net opening displacement at

the saw cut is measured by subtracting the width

of the saw blade, and the residual hoop stress is

measured using the formula:

(Eq 19)

bar (Fig. 11b):

possible cause:

1

1

r Et

x

D0

D0

p

1.000

2.900

B

1.900

(Eq 23)

E

5.000

(a)

1.65 Erd

L2

(Eq 20)

stress is given by:

r

Etd

L2

(Eq 21)

t/2 is used in the former case because only half

the sheet thickness deflects.

The method of sectioning a tube to obtain the

previously mentioned approximate measurement

is shown in Fig. 13(a).

A summary of the derivations is given in Table 2.

A Practical Example. The formulas in Table

2 are used extensively for measuring residual

hoop stresses in seamless tube and machined

bearing rings. These stresses may arise from tube

straightening without a stress-relief anneal or

nonuniform rapid cooling following the spheroidized anneal. In the past, contact with the support bars on cooling beds has been recognized

as being a potential cause of cyclic nonuniform

stress along the length of a tube.

Where 52100 bearing steel rings are being machined on multispindle machines and the operator is having difficulty in maintaining size, the

0.25 in.

0.50 in.

2.50 in.

0.5 in.

1.45 in.

(b)

2 in.

1/4 in.

1 1/4 in.

3/4 in.

2000:5:9

Peen stress

ENSAMMIC

8

Depth, mils

5/8 in.

d = 5.198 mils

h = 0.294 mils

10 A

S 230 shot

V = 124 ft/s

HS/4320 CB/SP

100

(c)

Examples of C-ring test specimens used for

quench distortion studies. (a) Source: Ref 5. (b)

Source: Ref 6. (c) Source: Ref 7

Fig. 9

Dir. Z

x

s (ksi)

102

m (ksi)

203

Pm (in.)

0.001

P0 (in.)

0.008

B

d

A

l

200

W

M = Wx

Max. M = Wl at B

Fig. 8

Calculated residual stress distribution in a shot peened chromium-silicon spring wire. d, depth; h, height. Source:

Peen Stress, Metal Improvement Company, Inc.

Fig. 10

moment. W, weight in lbs or kg

cut) is determined by measuring a shallow cut

adjacent to the final cut.

The gap width is measured using internal calipers. The net opening (x) equals the gap

width minus the saw curf (width).

On occasions the gap may decrease, indicating a

compressive hoop stress. In extreme situations,

the blade may become nipped and may even disintegrate. Ensure proper protection when using

this method.

For greater accuracy in large cross sections,

the value for E can be corrected for Poissons

ratio by:

E

E

28.57 106 psi

1 m2

stress. During subsequent grinding operations,

the stress distribution undergoes a redistribution

and may, in extreme cases, lead to the rings distorting. Although heat treating receives the

blame, many similar problems may, in fact, be

attributed to the incoming raw material.

Calculation

Considerable work has been accomplished in

recent years in the area of heat treatment modeling and, in particular, of software capable of

given in Table 3 (Ref 16). (These must only be

used as a guide. For more rigorous treatment of

elastic and plastic distortion, actual values

should be obtained for each material and condition under consideration.)

From experience in dealing with distortion

problems in manufacturing that are attributable

to residual stress in tube stock, it was determined

that the following criteria could be used as a

guide to the acceptability of the calculated stress

levels.

Residual stress

Acceptability

5000 psi

500010,000 psi

10,000 psi

Acceptable

Borderline

Unacceptable

stress is acceptable must also be based on the

cross section of the ring: the more rigid the design, the higher the stress level tolerated. However, this only applies to holding tolerance during machining. During subsequent heat

treatment, the magnitude of residual stress may

still be sufficient to cause distortion during heating. As the temperature increases, the yield

strength of the steel decreases, allowing relaxation by plastic deformation. This is manifested

in the parts being out-of-round exiting the furnace.

In very thin sections, some of the stress may

be from the machining operation (e.g., feeds and

speeds too high).

In order to lessen the effects of residual

stresses during heat treatment, pre-heating at a

subcritical temperature has been found to be useful. A portion of the stress is allowed to relax

without causing a gross shape change.

Examples of extreme levels of residual stress

in tube material are usually seen when it was

impossible to machine an acceptable part. Stress

relieving one particular batch of tubes (new supplier) resulted in them moving so far out-ofround that they no longer fit in the multispindle

machine collets.

In an attempt to avoid out-of-round rings, die

quenching techniques are often used. However,

example is DANTE (Deformation Control Technology, Inc., Cleveland, OH), developed by a

collaborative team comprising Ford Motor Company, General Motors Company, The Torrington

Company, Eaton Corporation, IITRI, Colorado

School of Mines, The Department of Energy National Labs, Deformation Control Technology,

Inc., and under the auspices of the National Center for Manufacturing Science. In order to accurately predict the thermal and allotropic stressinduced size changes and distortion in

heat-treated products (including carburized), it is

necessary to incorporate many factors into the

finite element-based model, including the transformational characteristics of the steel, the ele-

2d

Residual stress

distribution

(a)

2d + 2 r

(b)

Fig. 11

Determination of longitudinal residual stress by the deflection method. (a) Rolled sheet. (b) Drawn bar. Source:

Ref 14

t

d

L

(a)

D1

D0

x

(b)

Determination of residual stresses in thinwalled tube by deflection methods. (a) Longitudinal stress. (b) Circumferential stress. t, thickness; d, deflection; D0, initial diam; D1, diam after slitting; x, net

opening displacement. Source: Ref 14

Fig. 13

Fig. 12

formula derivation. Source: Ref 15

Using DANTE, it has been possible to subject

the net opening deflection method to a more rigorous analysis.

An example of the simulation of the opening

of a slit ring in the presence of an imposed hoop

stress is shown in Fig. 14 and 15.

Two-dimensional modeling was used, and

plane-strain conditions were assumed. The rings

were assumed to be 152 mm (6 in.) OD and man-

individual and mixed phases, geometry of the

part, quenching factors, heat-transfer coefficients, and so on. Several papers have been published on the development and application of the

model (Ref 1721). Commercialization of the

software package is being handled by Deformation Control Technology, Inc., Cleveland,

Ohio.

Table 2 List of formulas used for calculating approximate levels of residual stress in simple

geometries by deflection methods

Method

Formulas

Etd

r 2

2L

Longitudinal stress in solid bar (Fig. 11b) where r is the radius of the bar

1.65 Erd

L2

Etd

L2

are the initial and final diameters, respectively (Fig. 13b)

r Et

r Et

Table 3

1

0

1

D1

1

1

x

D0

D0

p

Residual-Stress State

Modulus elasticity

Shear modulus

Materials

GPa

106 psi

GPa

106 psi

Poissons ratio

Aluminum alloys

Copper

Steel (plain carbon and low alloy)

Stainless steel

Titanium

Tungsten

72

110

200

193

117

400

10.5

16.0

29.0

28.0

17.0

58.0

28

41

76

66

45

157

4.0

6.0

11.0

9.5

6.5

22.8

0.31

0.33

0.33

0.28

0.31

0.27

Source: Ref 16

Fig. 14

with two different cross sections. A residual

stress profile was superimposed on the rings by

creating a 93 C (200 F) linear temperature gradient across the cross section from ID to OD.

For the ring in Fig. 14, simulating axial slitting

of the ring resulted in a net opening of 2.9 mm

(0.114 in.) with an accompanying redistribution

of the residual stress.

Using Eq 23, the average hoop stress in the

ring prior to slitting was calculated to be 172

MPa (25 ksi). In thinner cross section rings, the

net opening will be larger for the same level of

residual stress, as shown in Fig. 15. For a cross

section of 9.5 mm (0.375 in.) and a similar residual stress condition, the net opening is calculated to be 6.3 mm (0.247 in.). Equation 23

gives a calculated hoop stress of 179 MPa (26

ksi). These calculations used a Youngs modulus

value of 29 106 psi and Poissons ratio of

0.33.

It is interesting to note that one of the many

techniques used to exercise DANTE was based

on the carburizing of a modified Almen strip

(Ref 22, 23). Excellent agreement between the

predicted and measured transverse deflections

were obtained.

Simulation of the effects of residual stress on the net opening displacement of a cut 0.75 in. thick ring

Treuting and Read developed a method for determining the biaxial residual-stress state on the

surface of a thin sheet (Ref 24). The method assumes the metal behaves in an elastically homogeneous manner and that the stress varies, not

in the plane of the sheet, but only through the

thickness. To apply the method, the sheet specimen is cemented into a flat parallel surface, and

Fig. 15

Simulation of the effects of residual stress on the net opening displacement of a cut 0.75 in. thick ring

careful polishing and etching. The sheet specimen is then released from the surface and measurements are made of the longitudinal radius of

curvature Rx., the transverse radius of curvature

Ry, and the thickness, t. Figure 16 illustrates the

orientation of the principle stresses and the curvature of the sheet. The measure values of radius

of curvature are expressed in terms of two parameters, Px and Py.

Px

1

m

Rx

Ry

Py

1

m

Ry

Rx

rx

(t

t)2

ry

(t

E

6(1 m2)

dPx

4(t0 t)Px 2

dt

t0

Px dt

E

6(1 m2)

t)2

dPy

4(t0 t)Py 2

dt

t0

Py dt

the curves of P versus t, and the integrals are

evaluated by determining the area under the P

versus t curve over the appropriate limits.

REFERENCES

Measurements of Rx and Ry are made for different amounts of metal removal, and Px and Py are

plotted against the sheet thickness, t. The residual stresses in the x and y directions of the sheet

are determined for any value of t by the following equations.

Carburizing, Pergamon Press, 1985

2. F.W. Jones, JISI, May 1969, p 556562

3. Shot Peening Applications, 7th Ed., Metal

Improvement Company, Inc., 1995

z

z

y

y

x

x

t0

(a)

Fig. 16

Rx

Ry

(b)

(a) Coordinate system for measuring biaxial stress in thin sheet. (b) Curvature produced by removing material

from top surface

private communication, 1999

5. Heat Treating, Cleaning and Finishing, Vol

2, Metals Handbook, 8th ed., ASM International, 1964, p 41

6. Tenaxal, Ucon Quench AThe Fast Safe

Way to Quench Steel Alloys, Product Information Bulletin, Tenaxal, Inc., Milwaukee, WI, revised 1972

7. H.J. French, The Quenching of Steels,

American Society for Steel Treating, 1930,

p 133

8. B. Prime, Residual Stress Measurement by

Successive Extension of a Slot: The Crack

Compliance Method, Los Alamos National

Laboratory, Publication LA-UR- 983857,

1998

9. W. Gremaud, I. Cheng, M. Finnie, and B.

Prime, The Compliance Method for Measurement of Near Surface Residual

StressesAnalytical Background, J. Eng.

Mater. Technol., Vol 116, p 550555

10. I. Cheng, M. Finnie, M. Gremaud, and B.

Prime, Measurement of Near Surface Residual Stresses Using Electric Wire Machining,

J. Eng. Mater. Technol., Vol 116, p 17

11. Kovac, Residual Stress Measurements in

Bevel Gear after Different Production

Phases, J. Mater. Eng. Perform., Vol 3 (No.

1), Feb 1994, p 6164

12. E. Shramm, J. Szelazek, and A.V. Clark,

DynamometerInduced Residual Stress

in Railroad Wheels: Ultrasonic and Saw Cut

Measurements, National Institute of Standards and Technology Publication NISTIR

5043, Report Number 30, March 1995

13. J.E. Gordon, The New Science of Strong

Materials, Penguin Books, 1968, p 258

261

14. G.E. Dieter, Mechanical Metallurgy, 1st

ed., McGraw-Hill Book Company, New

York, 1961

15. R.L. Anderson and E.G. Fahlman, A

Method for Measuring Internal Stress in

Brass Tubes, J. Inst. Met., Vol 32, 1924, p

367383

16. G.E. Totten and M.H. Howes, Steel Heat

Treatment Handbook, Marcel Dekker Inc.,

New York, 1997, p 256261

17. D. Shick et al., Development of a Carburizing and Quenching Simulation Tool: Determination of Heat Transfer Boundary Conditions in Salt, Second International Conf.

on Quenching and Control of Distortion

Proc., ASM International, 1996

18. D. Brammann et al., Development of a Carburizing and Quenching Simulation Tool: A

Material Model for Low Carbon Steels Undergoing Phase Transformations, Second

Control of Distortion Proceedings, ASM

International, 1996

19. C. Anderson et al., Development of a Carburizing and Quenching Simulation Tool:

Numerical Simulation of Rings and Gears,

Second International Conference on

Quenching and Control of Distortion Proceedings, ASM International, 1996

20. B.L. Ferguson, A.M. Freborg, and G.J. Petrus, A Software Tool to Simulate Quenching of Alloy Steels, Heat Treating Progress,

to be published

21. M.T. Lusk, Y.K. Lee, H.J. Jou, W.E. Elliott,

and G.M. Ludtka, An Internal State Variable

Model for the Low Temperature Tempering

Conference on Thermal Process Modeling

and Computer Simulation, Jiatong University, Shanghai, March 2000

22. V.C. Prantil, M.L. Callabresi, G.S. Ramaswamy, and J.F. Lathrop, Simulating Distortion and Residual Stresses in Carburized

Thin Strips, ASME J. Eng. Mater. Technol.,

to be published

23. M. Henriksen, D.B. Larson, and C.J. Van

Tyne, On the Analysis of Distortion and Residual Stress in Carburized Steels, ASME J.

Eng. Mater. Technol., Vol 114, 1992, p

362367

24. R.G. Treuting and W.T. Read, J. Appl.

Phys., Vol 22, 1951, p 130134

G. Totten, M. Howes, T. Inoue, editors, p99-117

DOI: 10.1361/hrsd2002p099

All rights reserved.

www.asminternational.org

C. Ruud, Pennsylvania State University

the principles, practices, and limitations of residual-stress measurement procedures for steel. It is

not meant to provide sufficient detail for the performance of the various methods described, but

references are cited where such procedural details may be found (e.g., Ref 15 or for general

inquiries, www.residualstress.com on the Internet).

There have been many methods and techniques proposed for the measurement of residual

stress, but only a few may be applied in practice

on components ranging from small to very large,

such as bridges and aircraft. These few include

x-ray diffraction (XRD) and blind hole drilling

with electrical resistance strain gages. For some

of the methods described, the component in

which residual stresses are to be measured must

be brought to the measuring instrument, but for

others the measurement devices are portable and

may be brought to the component (Fig. 1). However, in some cases it may be feasible to remove

a section from the component and bring that section to the residual-stress measuring device.

Great caution must be observed in this sectioning

because it will change the stress field by relieving and/or inducing stresses.

stress relaxation type methods discussed in the

section Introduction to Destructive Procedures

and the XRD methods discussed in the section

Nondestructive Techniques in this article. The

disagreement is nearly always because the volume of the component in which the stress is measured is not the same in the two methods, and

thus a different portion of the stress gradient is

measured by each. Ruud et al. (Ref 6) showed

that the hole-drilling method resulted in measured stresses approaching that of the stresses

measured at the surface by XRD when the holedrilling results were extrapolated to the surface.

Furthermore, because of the steep gradients, unless the stress field caused by the process is well

understood, stress measurements must be performed at many locations in the manufactured

solid in order to establish the magnitude and distribution of the stress field of interest. Many researchers in residual-stress techniques have fo-

shape of a solid, or where severe temperature

gradients exist during the process, causes residual stress. By their very nature, processes that

change the shape of a solid cause nonuniform

plastic deformation in the solid, which leads to

residual stress. These processes include forging,

rolling, drawing, machining, and so forth. Also,

processes that produce high thermal gradients in

a solid often lead to residual stress. These processes include quenching, casting, welding, and

so forth. Furthermore, processes that induce localized phase changes produce residual stress.

These processes include martensitic hardening.

The residual stresses caused by manufacturing

processes usually show very steep residual stress

to distance gradients as shown in Fig. 2. Many

of the other articles in this Handbook discuss and

describe in detail the causes and mechanisms of

residual-stresses development in manufacturing

processes. The steep gradients typical of residual

stresses induced by manufacturing processes

y

650

MPa

Welding

d y

~ 200

dx

3 mm

650

MPa

Machining

d y

~ 3000

dx

y

0

0.1 mm

y

650

platform component weldments using a portable

x-ray diffraction instrument

MPa

Drawing

d y

~ 600

dx

Drawn

cup

0

Fig. 1

Fig. 2

3 mm

x

x

Residual-stress magnitudes and distributions typical of a 650 MPa yield strength steel

cused on enhancing the accuracy of

residual-stress measurement and have ignored

the fact that from a practical standpoint, tens to

hundreds of stress measurements are needed to

define the stress field of interest. Thus, many of

the residual-stress measurement techniques require too much time to perform and are thus impractical. This includes some of the techniques

developed for XRD, strain gaging, and other

methods. Measurement times on the order of a

second are available with XRD and some other

methods, and automated stress mapping has been

performed with such techniques (Ref 7). Another

concern in the measurement of residual stresses

in manufactured components is that the area or

volume over which the stresses are resolved

must often be on the order of 1 mm (0.04 in.) or

less. This is because the residual-stress gradients

are usually quite steep, and measurement resolution larger than this tends to average the

stresses to such an extent that high stresses are

not detected.

Some characteristics of steel that can cause

error in residual-stress measurement by the various methods described in this article include

phase composition, plastic strain, grain size,

crystallographic texture, and others. For example, in a mixed ferrite/austenitic structure the residual stresses in the ferrite are invariably different than those in the austenite (see Table 1).

Also, Wimpory et al. (Ref 8) described the influence of varying amounts of cementite in a ferrite

matrix. The possibility of one or more of these

microstructural characteristics causing error in

the residual-stress measurements performed by

the methods described in this article should be

assessed by an expert in the method selected, and

details of the errors and their causes are not discussed in detail in this article.

The subsequent sections of this article discuss

and describe:

The need for measurementwhat problem is

obtaining information about the residualstress field?

The nature of the residual-stress fields in

steelsexamples of the magnitudes and distributions.

The strain basics for residual-stress measurementselastic strain measured, not stresses.

Table 1 Sample of residual-stress readings

from a 316 stainless steel pipe weldment

Distance from

the weld

fusion line

Stress

(a) in

austenite

Stress

(a) in

ferrite

in.

mm

ksi

MPa

ksi

MPa

0.04

0.07

0.11

0.15

1

1.8

2.8

3.8

21

19

16

17

145

131

110

17

46

67

62

65

317

462

427

448

Note: The ferrite places tensile stresses on the lattice of the austenite,

while the austenite tends to compress the ferrite; therefore, the more

compressive the stress in the ferrite, the less compressive the stress in

the austenite. (a) The precision of these measurements was 3.0 ksi

(21 MPa).

based on sectioning or removal of material to

cause a redistribution of the residual stress,

which is measured as a strain change.

The semidestructive methods of residualstress measurementthese methods are

based on the same principle as the destructive

methods or on the perturbation of the residual-stress field by other means.

The nondestructive methods of residual-stress

measurementthese methods do not permanently disturb the residual-stress field, but

directly measure the atomic lattice strain

caused by the stress or measure some physical

property perturbed by the lattice strain.

Measurements

Before the engineer or metallurgist commits

to measuring residual stresses in some component or workpiece, he or she must be sure that

the reason for the measurement is clearly understood. The major reasons that residual stresses

are of concern are:

Failures that are suspected as being caused by

hydrogen embrittlement

Assessment for the continued serviceability

of a component, for example, life assessment;

this is usually focused on a concern for inservice failure

Distortion occurring during processing of a

component

Distortion of components during storage or in

service

It is extremely important that the investigator

understand the mechanism for the inducement of

the residual-stress field of concern. As implied

by the other articles in this Handbook, most

cases of suspected harmful residual-stress fields

are induced by manufacturing processing or repair procedures, although sometimes abusive

service conditions or an accident may have

caused them (Ref 9). When manufacturing processes or sometimes repair procedures are

judged the most likely source of the residual

stresses, it is often possible to predict the magnitude and distribution of the residual stresses.

Such information can be obtained through consulting the literature or the application of computer modeling (Ref 1018 or the Internet at

www.residualstress.com). A preconceived

model of the residual-stress field will aid the investigation in determining the best method to

measure residual stress and the location and

number of measurements that need to be made.

Nevertheless, sometimes the cause of the residual-stress field is not evident, and the investigator is compelled to perform measurements as

a means to determine the cause. In such cases,

measurement methods and location must be selected without the aid of a priori knowledge of

the stress field, and it is prudent to consult the

literature and experts in the field of residualstress measurement and manufacturing processes.

Residual stresses are the inevitable consequence of thermomechanical processing of steel.

The resulting stress fields usually are nonuniform and show high stress gradients. For example, Fig. 2 illustrates the residual-stress magnitudes and distributions typical of a steel with

a 650 MPa yield strength. Because of the high

stress gradients, tens to hundreds of residualstress measurements with resolution on the order

of 1 mm may be required to identify precisely

the maximum stress and its location.

The characteristically nonuniform, high stress

gradient nature of residual stresses require that

either the induced stress field is well understood

and predictable, or many residual-stress measurements must be performed on one or more

components in order to reveal the nature of the

stress fields. Often, scientists attempt to gain an

understanding of a residual-stress field by making a few measurements on one or two components, and from this often arise erroneous conclusions regarding the nature of the stress field.

A few measurements may be useful if the scientist or engineer knows the distribution of the

stresses a priori. However, this is seldom the

case, and tens to hundreds of measurements are

required on a single component or many samples

to really understand the residual stresses induced

by a given manufacturing process. This means

that the measurement method must be as rapid

and labor-efficient as possible. Some of the new

semidestructive hole-drilling procedures and

XRD and ultrasonic (for special cases) instrumentation meet this criteria. Stress mapping is

offered with some XRD instruments to map

stresses over the surface of a component (Ref

19). Also, because the stress gradients are often

very high, the measurement method must be able

to resolve the stresses in dimensions on the order

of 1 mm or less. Here, some of the hole-drilling

procedures and recently developed XRD instruments can offer the best resolution.

Finally, the component in which residual

stresses are to be determined is often too large

to be brought to a laboratory, and removing sections is not a rational solution. Note that sectioning often disturbs the existing stress field to the

extent that it is not possible to reconstruct the

original stress field from the residual stress measured in pieces removed from the original whole.

Hole-drilling, XRD, and ultrasonic instrumentation are available as portable devices that may

be brought to the component in the field (Ref 7,

9, 20, 21).

Stress Measurement

A number of procedures and methods have

been applied to determine the residual stresses

extant in a metallic component, usually as a result of manufacturing processing. However,

stress is never the quantity measured because

stress is a quantity that is applied to a metal and

can only be measured in the process of its application. What is invariably measured to determine residual stress is elastic straineither the

elastic strain resulting directly from the existing

residual stress in the metal or the elastic strain

change resulting from relief of some portion, or

all, of the existing residual stress. The stress that

is causing, or has caused, the strain is then calculated using the applicable elastic constants for

the metal.

The methods described in the section Nondestructive Procedures in this article measure,

directly or indirectly, the strain response of the

metal to the residual stress in situ, while the

methods described in the sections Destructive

Measurement Procedures and Semidestructive

Procedures measure the strain change caused in

relieving some or all of the residual stress in the

metal.

Destructive Measurement

Procedures

The first concern in selecting a destructive residual-stress measurement procedure is whether

it is reasonable to destroy one or more components or samples in order to determine the residual stresses. Usually this implies that one or a

few of the components are a small portion of the

total number produced. Also coupled to this decision is whether the one or more components in

which the residual stresses are to be measured

are representative of all the others. In other

words, how great is the expected variation of the

residual-stress field from part to part?

As mentioned at the beginning of this article,

the need for the residual-stress information about

a component must be clearly understood. This

includes whether the triaxial stress field must be

established or if the uniaxial or biaxial condition

of the stress field along specific directions is sufficient. Examples of these three situations are

discussed in this section. However, in any case,

the fact that residual stresses are usually not uniform in any direction and show high stress gradients must be kept in mind when stress-measurement criteria are selected.

Destructive methods of residual-stress measurement are fundamentally stress-relaxation

procedures; that is, the information is obtained

by relaxing the residual stress in some finite-volume element of the component and measuring

the resulting strain change. The strain change is

then used, along with applicable assumptions

about the nature of the stress field, to reconstruct

the original stress field. Assumptions about the

nature of the stress field include the magnitudes

and gradients in the stress field and whether it is

sufficient to assume that the gradients are one-,

two-, or three-dimensional. In particular, the gradients that exist will dictate the size of the element that is to be isolated and made stress-free;

smaller the finite element must be in the direction of that gradient. It must be emphasized that

the larger the element and the higher the stress

gradient, the less quantitative and more qualitative are the measurement results. Electrical-resistance strain-gage technologies are emphasized

as the dominant method of strain measurement

due to their economic, procedural, and precision

advantages over other methods. However, modern XRD equipment when available has all of

these advantages as well and can be used to measure the stresses existing before and after sectioning.

A generic destructive stress-relief procedure is

described first, along with the issues generally

involved in each procedural step. This section on

destructive methods concludes with some discussion of qualitative chemical methods of residual-stress measurement.

Once the decision is made to measure the residual stresses destructively, the following steps

are usually applied in a typical stress-relief technique for residual-stress measurement.

Stress-Field Conditions. The engineering

problem for which the residual-stress information is needed must be analyzed. This need is

often generated by failures of the component in

service or by anticipated failures due to problems

with similar components. Distortion of a product

in storage or during manufacturing can also be a

concern. The shape of the componentthat is,

cylinder, disk, plate, and so forth, or some irregular shapemust be considered. This consideration, along with the process or processes by

which residual stresses were introduced, must be

analyzed. The justification for assumptions regarding the condition of the residual-stress field

can be established from these considerations.

This may lead to simplifying assumptions about

the stress-field condition such as axial symmetry

for a cylinder in which the dominant residualstress field is caused by quenching during heat

treat processing or about stress uniformity in the

surface plane of a plate where stress gradient

with depth is the major concern.

These assumptions and considerations lead to

the methodology, that is, equations, to be used

for computational reconstruction of the stress

fields from the measured strains.

Strain-Measurement Technique. With the

stress-reconstruction approach established, the

method of strain measurement and consequently

the number and/or spatial frequency of measurements can be determined. The strain-measurement technique selected will greatly affect

the resolution of the stress measurement because

of the spatial precision inherent in the technique.

There are a number of techniques that have

been used to measure the strain induced by the

relief of stresses due to sectioning or material

removal in destructive residual-stress measurement. These include mechanical gages, often dial

gages, employed with specially made jigs and

coatings, and electrical-resistance strain gages.

However, since the 1960s, the use of the latter

has become dominant due to the variety, availability, and precision of these gages. They are

available as uniaxial, biaxial, and rosette gages

of many sizes. The section Strain-Measurement

Methods in this article provides some detail regarding these methods. Also, since the late

1970s, extensive use of XRD has been applied

to provide rapid and numerous stress measurements on sectioned components to gain information regarding the internal stress field (Ref 2,

2225).

Preparation for Strain Measurement. With

the strain-measurement technique selected, the

measurement location must be established and

the component and/or element prepared for the

measurement by, for example, attaching strain

gages. A prestress-relief reading must now be

made before stress relaxation and isolation of the

element is initiated.

Isolation of Gaged Element. With the measurement technique in place, material removal to

isolate the gaged volume must be performed.

The technique for material removal, or sectioning, must be carefully considered because mechanical chip-removal processes such as lathe

turning, milling, sawing, grinding, and so forth

introduce surface residual stresses that can be as

great as the yield strength of the strain-hardened

material and several thousandths of an inch (tens

of micrometers) in depth. The section Sectioning and Material-Removal Methods in this article discusses methods to isolate the gaged element.

Post-Stress Relaxation Measurement. After

the residual stresses have been relaxed and thus

the elements isolated, strain measurements are

repeated, the final reading is subtracted from the

initial to obtain the strain change resultant from

the residual-stress relaxation. The resultant

quantities are then used in the residual-stress reconstruction equations to obtain the original

stress state of the component. The stress-reconstruction equations were selected as a result of

the assumptions made about the stress-field conditions described previously and in more detail

in the following sections.

Engineers and research scientists have approached the measurement of residual stresses

using destructive methods with the aid of assumptions about the stress-field conditions, including that the stresses in only one axis are of

interest in order to simplify the measurement and

reconstruction of the stress field (Table 2). These

have included certain uniaxial, biaxial, and triaxial stress-field assumptions.

Uniaxial Conditions. A procedure applicable

only to the measurement of residual stresses in

rods, cylinders, and tubesthat is, components

with axial symmetrywas reported by Heyn

(Ref 26). In this work, it was assumed that the

stresses were axially symmetric and that only the

longitudinal stresses were of interest. Thus, the

change in length of the components (cylinders,

rods, and tubes) was measured after removal of

an axially symmetric layer from the outside radius, or it was bored out of an inside radius. The

length of the component was measured after

each layer-removal process and entered into

various equations described in their paper. This

procedure is called the Bauer-Heyn method (Ref

26) and is applicable only to measuring longitudinal stresses in axially symmetric components. It does not measure radial or circumferential stresses.

A second procedure assuming a uniaxial stress

field, or that only the stresses in one axis are of

interest, was proposed by Stablein (Ref 27). Here

the component was a bar with a rectangular cross

section, and the residual stresses acting along its

length and varying through its thickness (smallest dimension) were measured. The material

from one face of the bar, one of the two faces

with the largest area, was removed by milling.

Equations used to reconstruct the original stress

field are described in Ref 27. The depth of the

removed layer must be significantly greater than

the depth of plastic deformation caused by milling (see the section Sectioning and MaterialRemoval Methods) and sufficient to cause a

measurable bend in the material. The length of

the opposite face of the bar from where the material was removed is measured before and after

layer removal to determine the effect of the removal of the stressed layer. Presently, this is usually done using electrical-resistance strain gages

(see Strain-Measurement Methods), but was

performed in the past by sensitive mechanicalgaging techniques. This measurement can also

be accomplished by measuring the bend in the

bar with suitable mechanical gages and fixtures,

for example, a cantilever-beam approach. This

procedure is applicable only to components of

rectangular cross section where the stresses parallel to the length are to be measured as they vary

through the thickness.

Biaxial Conditions. A procedure applicable

to axially symmetric components is the Mesnager-Sachs boring-out technique (Ref 28, 29). The

Table 2

stresses. Here the change in length and diameter

of the component are measured as material is

removed by axially boring-out material from the

inside to produce a hollow cylinder. Presently,

biaxial electrical-resistance strain gages (see the

section Strain-Measurement Methods) are

usually attached to the outside of the component

to measure the dimensional changes in the axial

and tangential directions. The strain change results are entered into equations described in Ref

28 and 29.

A procedure applicable to measuring biaxial

residual stresses homogeneous over the planar

surface of a flat metal plate or sheet was proposed by Trenting and Read (Ref 30). It was

based on removing uniformly thin layers of the

metal on one side of the sheet or plate and measuring the changes in curvature as the layers

were removed. It was assumed that the stresses

were constant over the plane of the sheet or plate

and varied only through the thickness. Electricalresistance strain gages (see Strain-Measurement Methods) or mechanical gaging may be

used to measure the change in curvature.

Another procedure for measuring biaxial

stresses homogeneous over the planar surface of

a metal plate was developed by Gunnert (Ref

31). This procedure assumes that a biaxial stress

condition was uniform throughout the depth of

a circular groove that was milled around an elemental volume of material to render it stress

free. Thus, the strain change on only one surface

was measured. The mechanical-gaging technique involved measuring the distance between

each of four sets of shallow holes drilled in the

element before the groove was milled into the

surface using a core drill. The distance between

each set of holes was measured before and after

the groove was produced and provided the information necessary to calculate original, assumed biaxial residual-stress condition parallel

to the surface of the plate. Theoretically, only

three sets of holes are required to measure the

biaxial stresses, but Gunnert used a fourth set to

improve the accuracy. A mechanical gage,

distance between each set of holes. It should be

noted that this technique could be used to measure the gradient in the biaxial stress condition

by pausing in the milling operation at selected

depths and measuring the distance between the

holes at each groove depth. Also, as with many

of the older techniques that originally applied

mechanical strain-gage apparatus, electrical-resistance strain gages or modern XRD techniques

could be used. A procedure involving the drilling

of a blind hole and electrical-resistance strain

gage is somewhat similar to Gunnerts original

technique and is described in the section on

semidestructive methods.

A more accurate procedure was later used to

measure residual stresses in pipe weldments.

Here the component (pipe) was divided into a

network (grid) of squares, and biaxial electrical

strain gages were placed in the center of each

grid square on the outside diameter of the pipe.

The pipe was then sectioned as shown in Fig. 3

into elements assumed to be stress free, and the

strain induced by the stress relief was read from

the gages. In placing the gages only on the outside diameter, the biaxial stress field was assumed to be uniform with depth; however, had

gages been placed on the inside and outside diameter a more complete measurement of the

stress field could have been obtained, albeit assuming a linear variation in the residual stress

from the outside to the inside surface.

A variation on the Gunnert procedure described previously was later published (Ref 1,

32). This variation was also used for application

to plates and implied that the triaxial stress field

could be measured by the technique. It assumed

that there was a homogeneous residual biaxial

stress field that varied with depth through the

plate thickness. Four holes were drilled in a

square pattern through the plate thickness. The

distances between all of the holes was then measured at selected hole depths. Next, a circular

groove was milled in steps of, for example, 2

mm around the drilled holes using a core drill as

in Gunnerts original procedure (Ref 31). The

distance between the holes at the various depths

Component shape

Rectangular cross-section bar

Plate, sheet

Cylinder, plate

Plate, weldment

Stress-field-condition assumptions

Method

Biaxial stresses, axial symmetry

Uniaxial stresses varying through thickness

Homogeneous planar, biaxial stresses

varying through thickness

Homogeneous planar, biaxial stresses

uniform through thickness

Homogeneous planar, biaxial stresses

varying through thickness

Planar biaxial stresses varying through

thickness

Planar biaxial stresses varying through

thickness

Triaxial

Longitudinal

Longitudinal, radial

Longitudinal

Biaxial in the plane of the

component

Biaxial in the plane of the

component

Biaxial in the plane of the

component

Biaxial in the plane of the

component

Biaxial in the plane of the

component

All

Bauer-Heyn

Mesnager-Sachs

Stablein

Treuting and Read

Various

Triaxial

All

All

Johanssen

Gunnert

Gunnert

Rosenthal and

Norton

Moore and Evans

Chen

Biaxial Conditions (Ref 28, 29)

Triaxial Conditions (Ref 27)

Strain-Measurement Techniques (Ref

30)

Strain-Measurement Techniques (Ref

31)

Strain-Measurement Techniques (Ref 1,

32)

Strain-Measurement Techniques (Ref

33)

Strain-Measurement Techniques (Ref 2)

Preparation for Strain Measurements

(Ref 34)

Strain-Measurement Techniques (Ref 2)

Strain-Measurement Techniques (Ref

25)

was then measured at each core depth until the

core was milled completely through the plate.

The cored plug was then assumed to be completely free of residual stress. The measured distances were then used to reconstruct the original

biaxial stress condition of the plug at each cored

depth. The mechanical measuring gage used in

this technique was similar to that used in Gunnerts first technique (Ref 31).

Another approach to measuring residual

stresses that has a broader application with respect to the shape of the component and the

stress-field distribution was proposed by Rosenthal and Norton (Ref 33). It is applicable to

plates and plate-shaped weldments.

The procedure involved cutting two narrow

blocks having the full thickness of the plate, each

with its long axis parallel to one of the assumed

biaxial principal, residual-stress directions in the

surface of the plate (see the near side of Fig. 4).

Thus, the long axes of the blocks are perpendicular to each other and parallel to the face, the

largest area surface of the plate. The smallest

dimension of the block should be several times

smaller than the plate thickness and its largest

dimension at least twice the thickness. The block

then can be further sectioned in order to determine the biaxial stress variation through the

thickness of the component. This proceeds by

first cutting the block at the location representing

the midthickness of the component plate, then

removing thin slices parallel to the original surface and from the bisected block as shown in the

upper part of Fig. 4.

The change in strain of the blocks is measured

using shallow holes or dimples in the original

surfaces of the component. These gage points are

located along the long axis of the block on the

original faces of the plate (see Fig. 4). The distance between these gage marks is measured before and after removal of the blocks from the

plate and after each sectioning of the block (Ref

33).

This procedure assumes a constant biaxial

stress field over the length of the blocks, which

is not the case in welded plates in the direction

transverse to the weld. The far side of the block

in Fig. 4 shows a sectioning procedure that

would reveal the stresses parallel to the weld

along a gradient transverse to the weld. The far

side of the block in Fig. 4 shows a sectioning

procedure that would reveal the stresses parallel

to the weld along a gradient transverse to the

weld. Also, electrical-resistance strain gages

could be used instead of measuring the distance

between shallow holes or dimples.

Another approach to a constant biaxial stress

field in a flat plate, varying only through thickness, was described by Moore and Evans (Ref

2). They relied on XRD for the measurement of

the strains from which the stress was calculated,

and the procedure consisted of removing layers

0

40

10 20 20 20

50

strains existing at each layer. The process assumed that the stress perpendicular to the surface

was zero. It should be noted that in order for this

procedure to be valid, the areas measured by

XRD would have to be free of plastic deformation caused by layer removal (see Sectioning

and Material-Removal Methods).

Triaxial Conditions. In reality, in most components in which residual stresses have been induced, usually due to manufacturing processes,

the stress field is triaxial and varies from point

to point (element to element) in all three directions. Thus, a number of destructive procedures

and stress-field condition assumptions have been

applied in order to measure the three-dimensional residual-stress-field condition existing in

most components of practical engineering interest. Two of these are described in this section.

Chen (Ref 34) revised Rosenthal and Nortons

approach to deriving the triaxial residual-stress

condition (Ref 33) as follows. The typical

method of residual-stress measurement is by mechanically removing part of a body and measuring the change of stress in the rest of the body.

The method of Rosenthal and Norton instead

deals only with a small element that has been cut

free from a plate. The sectioning procedure consists of removal of a narrow block from a plate

with gages attached, followed by splitting the

block in half with gages attached on the top and

bottom surface of the block, and then successive

slicing of both halves from the midsection to the

outer surface, as shown in Fig. 4.

When the half-block is sliced to a thickness of

0.1 in., gages are removed and stresses are measured (at least two points) on the surface by

XRD. The basic assumptions are:

Partial stress relief occurs in the direction of

90

Initial

sectioning

270

Weld

relief occurs in the direction transverse to the

long axis.

The small amount of stress relaxed in the remainder of the block follows a linear law

through the thickness when a thin slice of

metal is removed.

Variation of transverse stress along the axis

of the weld is small in the middle portion of

the plate weldment.

45

10

Saw cut

180

Fine cutter

(i.e., green abrasive saw 0.5 mm thick)

10

Further

sectioning

5

10

Final

sectioning

Gage

Weld

10

10

10

Rosette gage

Residual-stress measurement of a girth-welded pipe by strain gaging and sectioning. Note that strain gages

shown in the final sectioning should be placed on the pipe prior to initial sectioning; and for a more complete

analysis several of the layer sections detailed in the final sectioning step should be strain gaged and sectioned. Dimensions

given in mm

Fig. 3

the two narrow blocks suggested in Rosenthal

and Nortons procedure (Ref 33). The far side shows several

blocks sectioned to reveal the stresses parallel to the weld

with a gradient transverse to the weld.

Fig. 4

The determination of residual stresses may be

divided into the following steps:

1. Determination of e1 and e,

t which represent

the amount of strain relaxed in longitudinal and

transverse directions by cutting one longitudinal

and one transverse block free from the plate.

This was done by determining the strain relief

between two indentations on the top and bottom

surface of each block using a mechanical gage

(see Fig. 4) and subtracting the strain-gage readings on the blocks from the initial reading from

the plate. Note that this measurement can be performed using electrical-resistance strain gages or

XRD (Ref 7, 22, 23).

2. Determination of strain relaxed on the top

and bottom surfaces by splitting each block in

half and then successively slicing the blocks.

3. Determination of the strain relieved (using

at least two points) on the top and bottom surface

using sensitive mechanical gages to measure the

distance between indentations and using XRD

after the thickness of the top and bottom halves

of the blocks had been reduced to 0.1 in. Note

that this restriction was necessary because they

are using conventional scanning x-ray instrumentation. With modern x-ray instruments, measurements can be made on any size of specimen

(Ref 7, 19, 20, 2224).

4. An enhancement of the Rosenthal and Norton procedure (Ref 33) was suggested by Chen

(Ref 34) where the determination of the residual

stress remaining in the top and bottom slices was

measured using XRD. Here the lattice strain was

measured in the remaining slices in at least two

places near the gage points using XRD, and the

absolute residual stress, S*t and S*1 remaining

was calculated as described in the section Stress

Measurement in this article. These XRD-determined stresses measured for each block were averaged and designated S*

t or S*

1 for that face. The

original stress S present before slicing was determined by:

St S*t Eet

(Eq 1)

S1 S* Ee1

St or S1 by the modulus of elasticity, E.

5. Computation of stress relieved by cutting

the blocks from the plate was accomplished using the values of e and e obtained in steps 1

and 3, the amount of longitudinal and transverse

stress (St and S)

1 relaxed by cutting the blocks

from the plate:

S1

E

mE

(e1 me)

(et me)

t

1

1 m2

1 m2

St

E

mE

(et me)

(e1 me)

1

t

1 m2

1 m2

(Eq 2)

for the case where the length of the block is at

least twice the thickness of the block. In case of

shorter blocks:

S1

E

E

(e1 me)

t

1 m2

1 m2

me

t

m 2 b2

e1

1 b

E

E

St

(et me)

1

1 m2

1 m2

me1

m 2 b2

et

1 b

(Eq 3)

block. When plotted for the top and bottom faces

of the block and joined by a straight line, these

values give the stress S relaxed by cutting the

block from the plate.

6. Computation of stress (S) relieved by splitting and slicing the blocks was done using:

2 d (1 ) 2(e e )

(S)

a tE

1 det

et et

2 dl

0.5 (1 l)

0

3(1 )

(5 4)et (1 )eb0

0

(Eq 4)

eb eb0

2 dl

0.5 (1 l)

a

(5 4)eb0 (1 )et 0

(Eq 5)

fraction of the total thickness removed; e,

t e

b are

the relaxed strain measured on top and bottom

surface for position ; and et 0, eb0 are the relaxed strain on top and bottom surface when

splitting the block in half.

7. The total stresses relaxed across the thickness of the block in the longitudinal and transverse directions were obtained by:

S1 S1 S1

(Eq 6)

St St St

longitudinal and transverse directions were determined by:

sxy syx

Sw S v

2

(Eq 8)

Sy

sxy

syz

Y 0

y

x

z

(Eq 9)

Sz

sxz

syz

Z 0

z

x

y

(Eq 10)

to x, y, z, respectively, yields:

2Sx

2sxy

2sxz

0

x2

xy

xz

(Eq 11)

2Sy

2sxy

2syz

0

y2

xy

yz

(Eq 12)

2Sz

2sxz

2syz

0

z2

xz

yz

(Eq 13)

Eq 13, results in:

2Sz

2Sx

2Sy

2sxy

2

2

2 2

z

x

y

xy

1 deb

(S)

(1 ) 2(eb eb0)

a bE

2 d

3(1 )

Sx

sxy

sxz

X 0

x

y

z

(Eq 7)

stresses measured in the direction w and v making angles 45 and 45, with the longitudinal axis. However, if the biaxial stress condition

is assumed, only the strain relief in one 45 direction need be determined since they would be

equal. This requires that either a block at a 45

angle to the longitudinal block be removed, or

that electric residual-strain-gage rosettes be

used, or XRD measuring at the 45 angle.

9. Determination of stress in the thickness direction was determined by the equilibrium equations:

(Eq 14)

middle portion of the plate; therefore, the first

term in Eq 14 can be neglected, and Eq 14 can

be approximated using:

2Sz

2

2

(Sy1 Sy0)

sx2y2

z2

Dy21

Dx2Dy2

(Eq 15)

value of Sy1 at a distance Dy1 on either side of

the x axis, and sx2y2 is the value of sxy at a distance Dx2 and Dy2 from the y axis and x axis,

respectively.

Therefore, the stress in the thickness direction

Sz can be obtained through double integration

with the boundary condition that Sz vanishes at

both the top and bottom surfaces. In this way,

the entire triaxial residual-stress state was determined. Note that this procedure assumes that the

residual stresses are uniform along the length of

each block.

A variation on Rosenthal and Nortons

method (Ref 33) using electrical-resistance

strain gages or a nondestructive technique such

as XRD is as follows. The largest face of the

blocks described by the thickness of the component (plate) and the longest dimension of the

blocks is divided into a two-dimensional grid of

elements. An electrical-resistant strain gage is

placed on each element, or a nondestructive measurement such as XRD is performed, and the

block is sectioned along the grid lines to produce

elements that are assumed to be stress free. Note

that if a nondestructive technique such as XRD

is used, the plastically deformed surface created

by removing the block from the original component must be removed. This is best done using

electropolishing (see the section Sectioning and

Material-Removal Methods). If electropolishing of these cut faces is done to remove the plastic deformation and resultant residual stress induced by a mechanical cutting procedure, and

XRD is applied, then the blocks need not be sectioned. The measured XRD stress will provide

the absolute residual-stress-field condition in the

block, andcoupled with the strain relieved by

the original removal of the block from the

platethe entire triaxial residual-stress condition of the plate can be obtained.

This variation of Rosenthal and Nortons

method (Ref 33) provides the information necessary to derive the biaxial stress condition of

each block, which can in turn be used to derive

the triaxial condition of the original plate. The

stresses on each of the measured faces of the

blocks must be measured in three directions to

provide the information necessary to obtain the

principal stresses in the block faces in each element. Note that the strain change caused by the

sectioning of the blocks must be added to the

strains measured in each element.

This procedure can be applied to a weldment

with a single weld through its center as described

by Rosenthal and Norton (Ref 33), or to a more

elaborate stress field, for example, where two orthogonal welds existed in the component (plate).

With a single weld, only the block intersecting

weld needs to be sectioned into elements because

the residual-stress field in the block parallel to

the weld is likely to be constant along the direction parallel to the weld.

Moore and Evans (Ref 2) proposed mathematical procedures for the reconstruction of the

original three-dimensional residual-stress fields

in cylindrical and flat plate components, and

Constantinescu and Ballard (Ref 3) proposed a

modification of Moore and Evans work. They

proposed using XRD as the measurement technique and presented stress-reconstruction equations for the following conditions:

Solid cylinder bar, rotationally symmetric

rh(r) and rz(r), as well as the radial stress rr(r).

The theory of elasticity provides nine partial

differential equations: the three equations of

equilibrium and the six equations of compatibility. Unique solutions are possible, depending on

boundary conditions, and for the case considered

they give the following working formulas:

rr(r1)

r1

rhm(r)

dr

r

metric stresses

Hollow cylinder bar, rotationally symmetric

stresses

Flat plate, biaxial stresses

The Moore and Evans procedures were summarized in the Society of Automotive Engineers

Handbook (Ref 4) as described in the following

paragraphs.

Solid Cylinder Bar, Rotationally Symmetric

Stresses. It is presumed that the residual-stress

distribution has both rotational and longitudinal

symmetry, except near the ends where measurements are avoided. Stresses are therefore functions of the radius, r, and do not depend on the

angle, h, measured around the cylinder, nor on

the distance z taken parallel to the axis. By repeatedly removing thin concentric shells, the

stresses on the exposed surface in depth can be

obtained. The circumferential and longitudinal

measured stresses rhm(r) and rzm(r), respectively,

rzm(r)

dr

r

r1

rr(r1, h)

(Eq 17)

r1

g0(r)

2r1

2 g1(r)cos h dr

r

r

rh(r1, h) rhm(r1, h)

g0(r)

2r1

2 Re[g1(r)eih] dr

r

r

rh(r1, h) rhm(r1, h)

(Eq 19)

r1

g0(r)

6r1

2 Re[g1(r)eih] dr

r

r

(Eq 20)

(Eq 25)

2h0(r)

8r1h1

2 cos h dr

r

r

(Eq 26)

2r1

g1(r) sin hdr

r2

(Eq 27)

srh(r1, h)

r1

2r1

Im [g1(r)eih]dr

r2

(Eq 21)

(Eq 22)

respectively, of whatever follows.

The measured stresses rhm (r1, h) and rhz (r1,

h) are related to the g and h functions by a Fourier series relation:

rhm (r1, h)

gn(r)enih

rzm (r1, h)

hn(r)enih

r1

srh (r1, h)

r1

the cylinder 90 apart, and, regardless of symmetry, the g and/or h functions from Eq 23 become:

h 0

h 90

(Eq 28)

values of n 1, 0, and 1. This is a reasonable approximation, providing the stresses do

not drastically vary with h.

Adding the equations cancels the g1 and g1

terms, giving:

g0(r) 14[rhm (r) rhm (r)

r1

2h0(r)

8r1

2 Re[h1(r)eih] dr

r

r

R

r1

g0(r)

6r1

2 g1(r) cos h dr

r

r

r1

(Eq 24)

(Eq 18)

rz(r1, h) rzm(r1, h)

rz(r1, h) rzm(r1, h)

at depth of interest.

Solid Cylinder Bar, without Rotationally Symmetric Stresses. Stresses are again assumed to be

independent of z but allowed to vary in the circumferential h direction. Complex variable

methods give the following general solutions:

stresses

(Eq 16)

rr(r1, h)

2

(Eq 29)

Subtracting the equations in pairs and then adding the results cancels the g0 and g1 terms, giving:

g1(r) 14{[rhm (r) rhm (r)]

0

1

(Eq 30)

solutions, Eq 19 to 22. However, if the stresses

are symmetric with respect to a plane through

the axis, then rhm (r) rhm (r) and Eq 30 becomes:

1

(Eq 23)

plane through the axis of the cylinder, the general solutions simplify because the g1(r) and

h1(r) functions will be real. This is a reasonable

approximation when the residual stresses are induced by heat treatment or straightening and the

equations become:

0

(Eq 31)

to 27, along with similar formulas for the h function, determined from Eq 23 in the same way. It

is again interesting to note that Eq 16 to 18 are

included in the general solution of Eq 19 to 21,

when the stresses are rotationally symmetric.

The shear stress of Eq 22 or 27 exists only as a

result of asymmetry of the h stresses. When sym-

metry is obtained, the g1(r) terms are zero, as is

srh(r1, h).

Hollow Cylinder Bar, Rotationally Symmetric

Stresses. If R1 is the inside radius of the cylinder,

similar analysis generalizes the previous equations to the following final formulas:

1 r r

R21

2

1

r1

r2

rhm(r)

dr

R12

r

rz(r1) rzm(r1) 2

(Eq 32)

(Eq 33)

r21 R21

2

2 rr(r1)

1 R1

(Eq 34)

cylinder when R1 is zero.

Flat Plate, Biaxial Stresses. It is assumed that

the residual stresses in a flat plate of uniform

thickness depend only on the distance from one

of the flat surfaces of the plate, except, of course,

near the edges. It is also assumed that the principal stresses are rx and ry, lying in the plane of

the flat surfaces, and that the stress normal to the

flat surfaces, rz, is zero at all points sufficiently

distant from the edges.

From the authors assumptions and conditions

of equilibrium, the true stresses rx(z1) at depth

z1 can be expressed in terms of the measured

stress rxm (z1) by the relation:

6z1

z1

z1

rxm(z)

dz

z

rxm(z)

dz

z2

(Eq 35)

z1 is the distance from lower surface to uncovered depth of interest.

A similar expression holds for the y direction.

Equation 35 holds, even if rx and ry are not prin-

in.

I

II

III

I

4 ft

2 ft

1 in.

showing the layers proposed by Johanssen (Ref

25) where the thickness (T) of the layers are 0.5 in. (12.5

mm) and rI ry, rII rx, and rIII rz.

Fig. 5

z1

z1

sxym(z)

z

sxym(z)

dz

z2

(Eq 36)

r1

rx(z1) rxm(z1) 2

r2

rzm(r)

dr

R12

r

rh(r1) rhm(r1)

1/ 2

sxy(z1) sxym(z1) 2

dz 6z1

rr(r1)

exists, expressed by:

(Eq 37)

stress ram (z1) makes with the x axis. When measurements are taken 45 apart, sxym (z1) becomes:

rx

2

y2

ry

sxym(z1) r45(z1)

12[rxm(z1) rym(z1)]

resulting from the removal of material can be

determined by Dr(x) and m(x). These were to be

measured at a number of positions xi, i 1,,

N, on the lower side of the plate. Drx(x) and m(x)

are the difference between the stress and deformation measured prior to and following the removal of material. Changes in the internal stress

conditions are thus calculated directly and need

not be calculated as accumulated stress changes

resulting from several layers of material being

removed.

By introducing Aireys stress function u(x, y)

that is defined as:

(Eq 38)

required after each layer removed.

Johanssen proposed a procedure (Ref 25) for

the determination of the three-dimensional residual-stress field in thick plate (plate weldments)

components using XRD techniques to measure

the strains on the surfaces of the plate and plate

sections, and upon removal of layers of surfaces.

The procedure includes the measurement of the

biaxial stress field existing on the top surface of

the component (see Fig. 5), assuming that the

stress perpendicular to the surface is zero. Material is removed from this surface by, for example, milling and electropolishing, or electropolishing alone (see Sectioning and

Material-Removal Methods) and the biaxial

stresses remeasured at the new depth. Each time

material is removed, the forces that the removed

layer exerted on the remaining component must

be accounted for, and the subsequent measurements corrected for this change in the stress field.

Johanssen based his method on the following assumptions:

When a layer of material is removed, the re-

linear elastic; that is, Hookes law is applicable.

The residual-stress distribution is constant in

the z-direction, except at the surface, and rz

is a principal stress in the z-direction (see Fig.

5).

Upon material removal, it is assumed that the

strain ez remains unchanged. Together with

the previous assumption, this implies that the

change in stresses can be treated as a plane

problem.

It is assumed that the stresses are symmetrical

with respect to the y-z plane. This assumption

is, however, not necessary. The procedure can

be developed to include asymmetrical stress

states.

Johanssens justification for his procedure to

measure the three-dimensional stress field in the

weldment shown in Fig. 5 is as follows (Ref 25).

2

2

and sxy

x2

xy

(Eq 39)

differential equation:

2 2u 0

(Eq 40)

ry(x, 0) 0

(Eq 41)

y 0

sxy(x, 0) 0

(Eq 42)

sxy(a, y) 0

(Eq 43)

x a

rx(a, y) 0

(Eq 44)

experimental measurements; that is:

rx(x, 0) Drx(x)

(Eq 45)

Pickel (Ref 35) described a method for analytical solution of problems with this type of

boundary condition, using infinite, related series;

however, in this instance an approximate method

is used with trial solutions that are both more

convenient and numerically more stable.

r

The force T is divided into Tx(x) and Ty(x) (see

Fig. 6) and as a solution to Eq 45 to 47, u is

taken as u ux uy where ux is the solution

for the loading case where only Tx(x) differs

from zero (and therefore the boundary condition

ry(x, b) 0 must be applied), and uy is similarly

the loading case where only Ty(x) is not equal to

zero. This ensures that the trial solutions will be

sufficiently general and not involve any limitar

tions on T (x).

The solutions for ux and uy are then adjusted

so that they satisfy Eq 43 and as many of the

boundary conditions as possible. By direct substitution, it is easily verified that the trial solutions in the following satisfy all the boundary

conditions except rx(a, y) 0; however, this

condition can be filled and the measurement can

be fit using an approximate method, such as the

least-squares method. If this can be performed

sufficiently accurately, these trial solutions are

close to the exact solution, since it has been

shown that the solution is uniquely determined

by the stated boundary conditions.

For the loading case Tx(x):

y3

y2

Cnx

C0x

2

6

2

n1 n

(Eq 46)

width and thickness of the plate after removing

material.

Equation 39 gives:

exy

(1 nb coth nb)ny sinh ny]cos nx

sxxy

2x

xy

Cnx[(nb(coth

n1

(sin hny ny cosh ny)]coshnx

(Eq 47)

vx

y

(Eq 52)

one obtains:

(Eq 48)

v(1 v)

y2

Cx

C0xy

E

2

vx(x, y)

Cnx

n1

1v

E

(Eq 49)

n1

cosh ny

v

nby sinh ny

1v

calculated from:

(1 nb coth n y cosh ny

xz m(rx ry)

1

v

sinh ny

n

1v

(Eq 50)

exy can be obtained from:

exy

2x

Cxy C0x

y2

rxx

2x

x2

n1

x(x, y) Cx

rxy

1 m2

m

ry

rx

E

1 m

(Eq 51)

Since the condition for compatibility is satisfied, the displacement in the y-direction is

uniquely determined, excepting rigid body displacements, from the relationship:

cos x

sinh ny y cosh ny

(Eq 53)

r

T (x). In that case, the solution is given by:

y3

y2

C0y

6

2

y(x, y) Cy

Cny

n2

n1

cosh ny) 2n by sinh ny]cos nx

(Eq 54)

yx

2y

Cy C0y

y2

n1

(1 nb coth nb) (sinh ny

Removed layer

ny cosh ny)]cos nx

x

x(x)

Ty (x)

yy

2y

x2

(Eq 55)

Cny

n1

T (x )

Internal forces

xy

2

Tx(x)

syxy

(Eq 56)

n1

x

x(x)

(1 nb coth nb) ny

sinh ny]sin nx

(Eq 57)

x(x)corr

v(1 v)

y2

Ey

C0yy

E

2

vy(x, y)

1 v

E

Cny b sinh ny

n1

nby cosh ny

(x)

x(x)

v

1 v

x(x)

Residual-stress distributions, forces, and distortion of a plate before and after layer removal. Top: residual-stress

distribution in the x-direction in the center of the plate in the x-z plane; center: same as top after removal of a

layer with the forces Ti(j) caused by the residual stresses tending to distort the plate; bottom: same as center with the

distortion displacement shown (Ref 25)

Fig. 6

2

v

cosh ny

n

1 v

y sinh ny cos nx

(Eq 58)

By letting the stresses and displacements be

equal to the measured values, the following relations are obtained:

rxx(xi, 0) ryx(x2, 0) Drx(xi)

i 1, 2,

(Eq 59)

i 1, 2,

(Eq 60)

Eq 44 at a suitable number of positions:

rxx(a, yi) rxy(a, yi) 0

r(z1) rm(z1) 2

x

using the exact equations of the previous sections. The method is summarized for two of the

previous cases as follows.

Flat Plate (see Flat Plate, Biaxial Stress). A

generalized solution is written:

i 1, 2,

(Eq 61)

Equations 53 to 61 form an overdetermined system, from which the unknown coefficients Cnx

and Cny can be determined using, for example,

the method of least squares. The desired correction of the stresses measured at the surface after

removing material can finally be expressed as:

rxcorr rxx(x, b) rxy(x, b)

(Eq 62)

z1

dz 6z1

z1

rm(z)

z

rm(z)

dz

z2

(Eq 64)

because the form of Eq 35 and that of Eq 36 are

exactly the same. r(z1) represents the true stress

in any direction at depth z1 before a layer was

removed, and rm(z1) represents the measured

value at that depth.

The correction in stress c(z1) at z1 is the difference between the true and measured value,

given by:

H

and

dz 6z1

z1

z1

m[rxcorr

b)

ryy(x,

b)]

(Eq 63)

Sikarskie (Ref 36) proposed a stress-reconstruction procedure (series method) when thin

layers were removed from the surface of a component. He described procedures applicable to

flat plates and solid cylinders. The procedure

works well for shallow depths (a few percent of

specimen diameter or thickness), or in instances

where the stress gradient over the total depth removed does not change too rapidly and is of essentially one sign. The practicality of this

method depends on the fit of the measured

stresses in depth by a Taylors series referred to

the surface values of stress and successive derivatives at the surface. When the method is applicable, very convenient relations are obtained,

which describe the stress correction in terms of

the influencing factors, for example, layer depth,

stress magnitude, stress gradient, and specimen

size. Judgment is necessary, however, in using

the series approximation that does not arise when

(Eq 65)

z1

Fig. 7

H z1

H

H z1

H

r1

dropped because the form is the same.

When rr(r1) is desired:

r(r1) rr(r1)

rm(r) rhm(r)

(Eq 71)

r(r1) rz(r1)

rm(r1) rzm(r1)

(Eq 72)

(Eq 66)

(Eq 67)

be approximately proportional to the magnitude

of the surface stress and thickness of the removed layer (Fig. 7). It is inversely proportional

to the specimen thickness.

By solving for Dz1, the question of proper

slice is given by:

1 Hc(z1)

4 rm(H)

(Eq 68)

be in error by less than 5%, c(z1)/rm(H)

0.05 and the appropriate slice depth is:

(Eq 73)

Taylors series and integrating term by term, a

final form for the correction is obtained:

R r1

1

R

2

c(r1) k rm(R)

R r1

R

[rm(R) Rrm(R)]

Dz1

H

(Eq 70)

The correction term in stress, c(r1), is written

as before:

successive derivatives with respect to z at the

surface.

For shallow depths, only the first terms of the

series can be used and:

c(z1) 4rm(H)

rm(r)

dr

r

k 2

1

[2rm(H) Hrm(H)

3

2H 2rm(H)]

r(r1) rm(r1) k

rm(r) rzm(r)

[rm(H) 2Hrm(H)]

Dz1

Removed layer

The integrands are then expanded in a Taylors series referred to the surface values, after

which the integration is performed term by term.

The final form for the correction is:

H z1

c(z1) 4rm(H)

H

the slice depth is 0.005 in. (0.13 mm).

If the stress gradient is high, then the next term

in the correction series should be included and a

quadratic in Dz1 solved. This requires an estimate of rm(H) based on experience.

Solid Cylinder (see Solid Cylinder Bar, Rotationally Symmetric Stresses). A generalized

solution from Eq 16 and 17 is written:

k 1

rzcorr

ryx(x,

(Eq 69)

rm(r1) 0

rm(z)

z

rm(z)

dz

z2

1

(0.05)H

4

Dz1

1

6

R r1

(Eq 74)

stress and successive derivatives with respect to

z at the surface.

Insight into the factors that influence the correction holds exactly as previously discussed, as

do the limitations of the method.

Ruud et al. (Ref 2224) applied a modification of Johanssen method to measure the triaxial

stress condition of thick plate 214Cr-1Mo plate

weldments. Ruud et al. actually measured the

strains in all directions and calculated the

stresses, but did not correct for layer removal due

to the complex nature of the stress field. Ruud et

al. also measured the residual-stress condition of

expanded tubing including 304 stainless steel

tubing (Ref 37) but focusing on the residual

stresses on the inside surface of the heat-exchanger tube components.

to the milling direction, ksi

10

20

30

0

Residual stresses at the surface and near the surface due to milling a medium-carbon steel work-

Fig. 8

piece

40

20

0

20

40

60

80

100

120

10

12

14

Methods

As discussed in the previous sections in this

article, many procedures require that the component (sample or part) be sectioned and/or

some material be removed from it to measure the

residual stresses. This is especially true for the

measurement of internal residual-stress fields

where the component nearly always has to be

sectioned to reveal the internal stress field. The

two exceptions to the necessity of sectioning and

material removal, neutron diffraction and ultrasonic methods, are described in the section

Nondestructive Procedures in this article.

Sectioning or material removal is required by a

particular residual-stress measurement procedure

(see the section Stress-Field Condition Assumptions) or method (see the section StrainMeasurement Methods). Mechanical chipremoval processes are usually applied because

of their economy and speed. All chip-removal

processes, including lathe turning, drilling, milling, sawing, grinding, and so forth, introduce

surface residual stresses that can be as high as

the yield strength of the strain-hardened metal

and several thousandths of an inch (tens of microns) in depth (Ref 3843). Figures 8 to 10

show the residual stresses in steels caused by

various machining processes. Furthermore, some

steels are especially prone to strain hardening

for example, austenitic stainless steelsand extra care must be used with these materials when

selecting a material-removal technique. Figure

11 shows plots of the residual stresses in 304

austenitic stainless steel caused by various grinding methods. Note that these plots are only a

sample and may not be typical. If the size of the

element in which the strain change is measured

is smaller, or thinner in the case of surface-depth

gradients, than about 0.1 in. (23 mm), then a

chemical or electrochemical material-removal

technique must be used to remove the surface

residual stresses caused by mechanical chip re-

steel workpiece induced by facing

Fig. 9

100

80

20

Residual stresses, ksi

0

20

40

60

80

Strain-Measurement Methods

Abusively

machine

ground

Handheld grinder

60

measure elastic strain, not stress, and the residual

stress is calculated from the strain values. Several methods for the measurement of strain have

been applied in residual-stress studies and have

been mentioned previously. These methods include:

40

Handheld

flapper

wheel

20

0

100

120

20

140

10

12

14

Fig. 10

induced by turning

common grinding procedures

Fig. 11

in conjunction with, and after, the chip-removal

method. It should be noted that material-removal

techniques such as electrodischarge milling

(EDM) induce residual stresses (Ref 44) as do

chip-removal methods. Other methods, such as

laser, flame, or plasma cutting that cause heating

of the element, must be applied with caution because they may reduce the stress field by annealing before it can be measured.

Thus, the only methods for material removal

from a component surface that do not induce residual stresses are electrolytic or chemical polishing. Electropolishing is described in some detail in Ref 45, and guidelines are provided for

application to various ferrous alloys. In electropolishing, the electrolyte and operating conditions depend on the alloy being polished, as

shown in Ref 4 and 45. Electropolishing combined with XRD is used extensively to reveal

residual-stress gradients on machined, ground,

and hardened surfaces. However, application of

these techniques requires that the subsurface

stresses be corrected for the removal of prior surface layers (Ref 4).

Another concern when reducing components

to a more convenient smaller size to place them

on or in a measurement device is that the stresses

of interest are not changed by the sectioning.

Generally, plates should be cut to a length and

width of at least three times the thickness to

avoid end effects. Cylinders, both thin-walled

and solid, should be a minimum of three diameters in length.

Where the manufacturing process affects the

entire thickness of a component, such as heat

treatment or forging, it might not be advisable

to section without means of measuring the stress

change extensively over the entire component

before sectioning. In other words, selection of

the stress-measurement procedure and methods

should avoid sectioning unless techniques to

measure the effects of sectioning are applied before sectioning is initiated. On the other hand,

when the processes that have induced the residual stress produce only shallow stress fields, then

the three-times rule suggested previously is applicable.

Mechanical gages

Electrical-resistance gages

Optical gages

Birefringent methods

Diffraction methods (x-ray and neutron)

Ultrasonic methods

Magnetic methods

Diffraction, ultrasonic, and magnetic methods

are discussed in the section Nondestructive Procedures in this article.

Mechanical Gages. The application of mechanical gages such as those described by Heyn

(Ref 26), Stablein (Ref 27), Gunnert (Ref 31),

and others generally preceded the availability of

electrical-resistance strain gages and are not discussed due to their general lack of precision,

poor spatial resolution, and inefficiency.

Electrical-Resistance Strain Gages. Most

bonded electrical-resistance strain gages are

made from either metallic wire or foil materials.

There are also the recently developed semiconductor gages. A variety of sizes, shapes, and configurations are available including single-element gages and rosettes with two, three, or four

elements. Electrical-resistance strain gages are

available in sizes as small as about 1 mm and

thus provide a resolution of strain measurement

on that order. Information on electrical strain

gages is available in numerous sources including

Ref 46 and in reviews by Crites (Ref 47) and

Masubuchi (Ref 14) as well as suppliers (e.g.,

Measurements Group Inc., Raleigh, NC).

Changes in temperature tend to cause an apparent strain; therefore, some type of temperature compensation is needed. Frequently, a

dummy gage, which is not subjected to the

strain, is exposed to the same temperature as the

actual gage to provide a basis for comparison. A

temperature-compensated gage can also be used.

Gages must be bonded securely to the specimen. Various types of cements have been developed. Sometimes gages must be protected

from metal chips produced during machining as

well as from the oil or water. A number of systems have been devised for protecting gages under various conditions.

Brittle coatings are a simple inexpensive

strain gage that will only provide qualitative indications of residual stress. A brittle lacquer is

applied to the area where the stresses are to be

measured by material removal. After the lacquer

has cured (dried), a change in the stress field is

induced, and if the change is sufficient strain will

be produced in the lacquer, causing it to crack.

If the material removal is in the form of a hole

drilled in the lacquer, radial cracks indicate a tensile residual stress in the plane of the component

surface and circular cracks indicate compressive

stress.

Optical Gages. Light reflective methods can

be used to magnify the movement of a reflective

surface in a well-fixtured component that is held

securely in place during material removal to

change the stress field and therefore induce strain

change. This method can also be used if the component can be removed and replaced precisely in

a fixture and the position of the reflected light

measured before and after removal and replacement during which a change in the residualstress field was induced.

Laser Methods. Other techniques applying laser light have been proposed as well. These have

included shearography (Ref 48), interferometry

(Ref 49), speckle-correlation interferometry (Ref

50), and others. Vikram and others (Ref 50) suggested that a small volume of the material could

be stress relieved by heating via a laser to induce

a change in the stress field and the strain change

measured by an optical technique to reveal the

residual stresses existing in the volume before

heating. However, it must be recognized that

heating a volume of steel sufficiently to change

the residual-stress field will result in tensile residual stresses in the heated volume (Ref 51),

and this would likely be detrimental to the component in which stresses were being measured.

Birefringent Methods. Under the action of

stresses, transparent materials become doubly refracting (birefringent), and if a beam of polarized

light is passed through a model (under stress)

made of such a material, a colored picture is obtained from which the stress distribution can be

determined. This technique is called the photoelastic technique (Ref 46). The actual component

in which the stresses are to be measured is coated

with a photoelastic polymer. When residualstress changes are induced in the component,

strain changes are caused and transmitted to the

polymer coating, which then becomes birefringent. This can be observed and measured using a reflection polariscope (Ref 14). Instructions

for analyzing fringe patterns in this application

(nearly the same as those obtained in ordinary

photoelasticity) are provided by the manufacturer.

The photoelastic coating may be applied by

brushing a liquid polymer on the surface of the

specimen and polymerizing it by applying heat.

Alternatively, a prefabricated flat or contoured

sheet of polymer may be bonded to the part at

room temperature (Ref 52). The maximum strain

that can be measured ranges between 3 and 50%,

depending on the type of polymer used; the

strain sensitivity usually decreases with the increase in the maximum measurable strain.

Chemical Methods

A number of qualitative methods to detect residual stresses that may lead to stress-corrosion

hydrogen-induced cracking in steel have been

applied to specimens representing components

to be manufactured. Magnesium chloride solutions have been applied extensively to the study

of stress corrosion in austenitic stainless steels,

including some recent work by Bouzina et al.

(Ref 53).

Masabuchi and Martin (Ref 54) studied the

susceptibility of SAE 4340 steel weldments to

hydrogen-induced stress cracking. The test procedure was to immerse the weldment specimens

in a 4% H2SO4 aqueous solution charged with

H2 and to which two drops of a 5% P solution

of CS2 was dissolved. A direct current (dc) was

applied between a specimen and a lead anode to

provide a current density of 0.35 to 0.8 A/in.2

The crack patterns that developed were related

to the surface tensile residual-stress distribution

in each specimen.

Stress-corrosion cracking induced by residual

stresses in carbon and low-alloy high-strength

steels has been investigated by several researchers (Ref 5456). One procedure consisted of immersing the specimens in a boiling aqueous solution of 60% Ca(NO3)2 and 4% NH4NO3 for 31

h. The crack patterns that developed were related

to the surface tensile residual-stress distributions

in the specimen.

A number of standard practices for testing the

susceptibility of metals to stress-corrosion cracking have been published by ASTM, including:

ASTM G 38 Standard Recommended Prac

tice for Making and Using C-Ray Stress Corrosion Test Specimens, 1979

ASTM G 58 Standard Practice for the Preparation of Stress Corrosion Test Specimens

for Weldments, 1978

ASTM G 39 Standard Practice for Preparation and Use of Bent-Beam Stress Corrosion

Test Specimens, 1979

ASTM G 30 Standard Recommended Practice for Making and Using U-Bend Stress

Corrosion Test Specimens, 1979

ASTM STP 425 Stress Corrosion Testing,

1967

reveal the residual stress, but rather the susceptibility of the metal to cracking under known

stresses in the specified corrosion medium.

Semidestructive Procedures

Nondestructive methods of residual-stress

measurement are characterized as methods that

in no way affect the serviceability or reduce the

mechanical strength or other properties of the

component in which stresses are measured. Between the nondestructive and destructive methods, which have a severe effect on the serviceability, strength, and properties, are the

semidestructive methods. These are methods

that have a small to negligible effect on the components in which stresses are measured, or methods in which the component may be repaired after the measurement.

The methods that are considered semidestructive are those that require small holes to be

drilled or rings to be trepanned in the component

or indentations to be made in the surface. The

first two methods provide quantitative data and

the third only qualitative data.

The hole-drilling method was proposed in the

1930s (Ref 57) and is based on measurement of

the change in surface strain caused by stresses

relieved by machining a shallow hole in the testpiece. The principle is that removal of stressed

material results in the surrounding material readjusting its stress state to attain equilibrium.

The method has been standardized in ASTM E

837 (Ref 58). The ring-core method (Ref 59) is

also based on the strain caused by disturbing the

stress field, but in this case a relatively stressfree island of material is isolated by making a

shallow ring around a strain gage. This method

is also called trepanning. These two methods are

the least destructive mechanical stress-relief

techniques and are relatively simple and economical. They, as do nearly all stress-relief techniques, rely on electrical-resistance strain gages

to measure the strain change due to metal removal. Rosettes of strain gages are available especially for hole drilling. The size of the rosettes

has been progressively reduced over the last few

decades and are now available in sizes less than

about 10 mm from a number of manufacturers.

As with most residual-stress techniques, hole

drilling and ring coring have been applied

mostly to steels. Most applications have been

done on flat plate or cylindrically round parts

(Ref 5, 60, 61). Stresses can be determined at

various depths into the surface of the material,

down to a depth equal to the diameter of the hole

or core (Ref 62, 63). Kelsey (Ref 64), however,

observed that stresses with depth cannot be measured accurately to greater than half the hole diameter. The thickness of the layers in which

measurements are resolved is about 10 to 20%

of the hole or core diameter.

The equipment necessary to perform the measurement is reasonably inexpensive, portable,

and can be used in a manufacturing shop environment. However, experienced technologists

are needed to take the readingsfrom selecting

the area in which stresses are to be measured to

preparing the surface, applying the strain gages,

and reading and interpreting the data. Due to the

possibility of residual stresses being induced by

the hole-drilling or coring technique, prior calibration of the application is recommended in all

cases, with the possible exception of certain applications where holes are produced by abrasive

jet machining (AJM) (Ref 65).

Rendler and Vigness (Ref 66) developed calibration constants for cold-rolled steel that they

proposed as generally applicable to all metals

provided that the elastic constants were known.

However, they seem to have overlooked variations in the strain-hardening coefficients and the

accompanying residual stress, which exist between alloys and even between tempers of the

same alloy. Dini et al. (Ref 67) showed that direct experimental determination of the necessary

constants for any isotropic material with known

elastic constants can be eliminated by using data

available for cold-rolled steel and calculating

these constants using a formula presented. Despite the success that some researchers have

claimed in circumventing the development of

calibration constants, experimental calibration is

strongly recommended. This is best done by applying strain gages and drilling the holes in testpieces prior to stressing them known amounts

(Ref 65, 66). The AJM technique should be applied to any material with high propensity to

work-hardening during machining, for example,

austenitic stainless steel (Ref 68).

The following are general limitations and/or

concerns of hole drilling and ring coring:

Areas of high stress gradients should be

avoided because the stress gradient must be

diameter.

Areas where stresses are greater than onethird the yield strength of the material are

likely to produce erroneous results due to local plastic yielding during metal removal.

The thickness of the part or specimen must

be at least four times the hole or core diameter.

Strain hardening of the steel in the vicinity of

the hole may result during metal removal,

which can cause tens of ksi (10 ksi 69

MPa) error.

Heating may result during the metal removal.

Holes or cores must be spaced at least eight

times their diameter apart.

The area in which stresses are to be measured

must be accessible to a rather bulky drilling

or coring alignment device.

Preparation of the surface for strain-gage adherence may induce residual stresses that introduce substantial error to the subsequent

measurement (Ref 69).

methods are nearly nondestructive variations of

the destructive mechanical stress-relief techniques and require only rather simple equipment

and instrumentation. The state-of-the-art is relatively well developed compared to many nondestructive methods, some of which require considerable research and development work before

they will ever be suitable to general application

in terms of alloys and stress-field conditions.

Technological advancements in hole drilling and

ring coring have largely been due to advancements in the more general area of mechanical

stress-relief methods and research in new metalremoval techniques for metal fabrication.

Indentation Methods

Engineers and scientists have proposed the

use of indentors, such as those used to perform

hardness measurements, as a means to measure

or detect surface-residual stresses since the

1930s. Kokubo in 1932 reported that stresses applied under bending load changed the apparent

Vickers hardness values in carbon-steel rolled

sheets, both as-rolled and annealed (Ref 70). He

showed that tensile stresses tended to decrease

the apparent hardness and compressive stresses

tended to increase the hardness. The stresses applied in tension and compression were sufficient

to cause 0.3% strain.

In the 1950s, Sines and Carlson (Ref 71) proposed a method that required various amounts of

external loads to be applied to the component in

which residual stresses were to be measured

while hardness measurements were made. The

loads were made to cause both tensile and compressive applied stresses. The quality, that is,

whether the residual stress was compressive or

tensile, was then revealed by comparing the effect of the applied stress, and whether it was tensile or compressive, on the hardness measurement. At about the same time Pomey et al. (Ref

72) proposed that residual stresses could be measured by pressing a ball-shaped penetrator into

the component in which residual stresses were

to be measured and establish the relationship between the pressing load while it was progressively increased, and the electrical resistance at

the interface between the penetration and the

component. He maintained that a smaller decrease in electrical resistance indicated that portions of material under the ball were plastically

yielding and that the corresponding load on the

ball could be related to the existing residual

stress.

Later, Chiang et al. (Ref 73) provided a critique of several existing indentation analyses and

proposed an interpretation of indentations exhibiting hemispherical plasticity. Nevertheless, the

applications illustrated in this paper focused on

brittle materials and not metals.

There have been numerous papers published

proposing various approaches to interpreting the

indentation loads and shapes so as to estimate

the residual-stress field on the surface and nearsurface regions of materials. However, indentation methods have not earned the degree of confidence of XRD or hole-drilling methods for

general applications and thus are rarely applied.

Spot Annealing

Another method that has been proposed to

measure residual stresses in metal surfaces is to

reduce the residual stresses in a small volume by

annealing the metal in the volume. It has been

proposed that this annealing be performed by intense laser light (Ref 50). This technique was

envisioned to be similar to residual-stress relief

by removal of the material as accomplished in

the hole-drilling techniques. However, as Cullity

discussed in Ref 51, such localized heating

would induce high surface residual tensile

stresses in the heat-affected region.

Nondestructive Procedures

The methods described in the section StrainMeasurement Methods all measure the change

in some dimension (strain) of the component

produced by the removal of a finite volume of

stressed metal from that component. Thus, they

measure the strain induced by removing material

so as to perturb the residual-stress field. On the

other hand, nondestructive procedures measure

a dimension in the crystal lattice of the metal or

some physical parameter affected by the crystal

lattice dimension. Whenever a mechanical force,

resulting in stress that is less than the yield

strength, is placed on a solid metal component,

that component distorts and strains elastically.

That elastic strain results in a change in the

atomic lattice dimension, and this dimension, or

change, is measured by the nondestructive

stress-measurement procedures. For example,

the diffraction methodsx-ray and neutron

measure an actual crystal dimension, and this

dimension can be related to the magnitude and

direction of the stress that the metal is subject to,

whether that stress is residual or applied. Subsequently, in this section the following methods

of stress measurement are described:

X-ray diffraction

Neutron diffraction

Ultrasonic velocity

Magnetic Barkhausen noise

X-Ray Diffraction

X-ray diffraction techniques exploit the fact

that when a metal is under stress, applied or residual, the resulting elastic strains cause the

atomic planes in the metallic crystal structure to

change their spacings. X-ray diffraction can directly measure this interplanar atomic spacing;

Ns

X-ray source

S1

Film

Np1

Np2

S2

Specimen

surface

O

2

that the incident beam makes with Ns, Np1, and Np2 the

normals to the diffracting planes 1 and 2, respectively; w1

and w2, angles between Ns, Np1, and Np2, respectively; g,

the angles between the incident beam and the diffracting

plane normals; Ro, camera radius; O, point of incidence of

the x-ray beam of the specimen; 1 and 2, diffracting planes

at various attitudes to the specimen surface; S1 and S2, measured parameters representing the distance from a reference point of known distance from the incident beam and

the diffracted x-ray beam position. S1 and S2, directly related to the Bragg angles, h1 and h2.

Fig. 12

Miniature x-ray diffractometer for the one-angle technique arrangement of XRD stress measurement. This device incorporates a Ruud-Barrett Position-Sensitive Scintillation Detector and is capable of being

inserted in a 4 in. (100 mm) inside diameter and measuring

residual stress (Ref 22).

Fig. 13

can then be obtained.

Since metals are composed of atoms arranged

in a regular three-dimensional array to form a

crystal, most metal components of practical concern consist of many tiny crystallites (grains),

randomly oriented with respect to their crystalline arrangement and fused together to make a

bulk solid. When such a polycrystalline metal is

placed under stress, elastic strains are produced

in the crystal lattice of the individual crystallites.

In other words, an externally applied stress or

one residual within the material, when below the

yield strength of the material, is taken up by interatomic strain. X-ray diffraction techniques

can actually measure the interatomic spacings

that are indicative of the elastic strain in the specimen. Stress values are obtained from these elastic strains in the crystals by knowing the elastic

constants of the material and assuming that stress

is proportional to strain, a reasonable assumption

for most metals and alloys of practical concern.

Reference 74 describes XRD and instrumentation in detail.

There are three basic techniques for measuring

stresses, based on XRD. They are the doubleexposure or two-angle technique (DET), the single-exposure or one-angle technique (SET), and

the sin-square-psi or multiangle technique. The

angle of exposure referred to is that between the

incident x-ray beam and the specimen surface

normal. It should be noted that in any XRD

stress-measurement technique, x-ray peaks in the

far back-reflection range, that is, peaks with

Bragg angles (h) of near 90, are much preferred

because they show the greatest effect with a

given amount of applied or residual stress. This

is illustrated by:

r (2h1 2h2)

cot h1

E

1

p

2

1 m sin2w 180

(Eq 75)

where h1 is the Bragg angle of the planes diffracting at w1 and h2 is the Bragg angle of the

planes diffracting at w2. As h1 increases, its cotangent decreases; therefore, a larger difference

(2h1 2hw) would result from a given ru to

maintain an equality.

For a residual-stress measurement, the diffracting angle, h, of interatomic planes of at least

two different psi (w) angles with respect to the

surface normal must be measured (Fig. 12).

These planes are crystallographically equivalent

(same Miller indices, hkl ) and in the unstressed

state of the metal would have the same interatomic (d) spacing for the planes labeled 1 and

2 in Fig. 12 (Ref 51, 75, 76). In a stressed material, however, the two or more orientations of

diffracting planes are selected so that they are at

different angles to the surface; thus their normals

are at different psi (w) angles to the surface normal. Then, depending on the angle of these

planes to the stress vector, their interplanar

atomic spacing is increased or decreased by

varying amounts.

The most common sources of errors and misapplications in stress measurements by x-rays

are related to stress constant selection, focusing

geometry, diffracted peak location, cold-working crystallography, texture, grain size, microstructure, and surface condition. The source, significance, and correction techniques for these

errors will not be elaborated upon here; details

may be found in Ref 77.

A point of interest in the error sources concerns cold working and microstresses. As stated

previously, microstresses are usually considered

to be those manifested by strain variation across

single metallic grains. This strain variation is detected in XRD as broadening of the x-ray peak

a distinctly different phenomenon from the peak

shift caused by residual stresses. However, microstrain variation can be measured simultaneously with stress. This microstrain phenomenon has been proposed as a means of judging

cold work, dislocation density, and fatigue damage (Ref 78).

Despite the facts that x-rays provide stress

readings only to a depth of less than 0.001 in.

(0.025 mm) and that the error sources listed previously must be considered, the noncontact XRD

method is presently the only time-proven, generally applicable, truly nondestructive method

for measuring residual stresses. Its reliability has

been proved and documented by thousands of

engineers and scientists since the 1960s, beginning with Bolstad et al.s classic work at Boeing

using x-ray film cameras (Ref 79). This documentation includes measurement of stresses in

the Brooklyn Bridge (Ref 20) and tempering

evaluation of carburized steels. The Society of

Automotive Engineers considers the method of

sufficient practical importance to have printed

three handbook supplements on the subject (Ref

4), and another is under revision.

Even so, this nondestructive technology has

been largely restricted to the laboratory because

of the general lack of knowledge regarding the

state-of-the-art instruments and the limitations of

the more widely known and available conventional scanning XRD equipment. Instrumentation for bringing this technology into the field

and manufacturing area has advanced rapidly

since the 1980s, especially toward increased

portability, compactness, and speed of operation.

As shown in Fig. 13 and 14, instrumentation

has been developed and is commercially available for stress measurement in situ on the inside

diameter of 10 mm (4 in.) diam pipe, and soon

for 5 mm (2 in.) diam steel tube (Ref 22, 23).

Position sensitive x-ray detectors have been

largely responsible for these improvements to

both laboratory-based and field-deployable residual-stress-measuring instruments (Ref 9, 19,

2123, 8082). Also, with the speed of data collection being less than 0.1 s with conventional

x-ray tube sources, XRD stress measurement can

be performed on moving components (Ref 11).

Nevertheless, many engineers have been frustrated in applying XRD to residual-stress measurement. This has been largely due to crystallographers inexperienced in residual-stress

measurement, attempting to apply conventional

scanning x-ray diffractometers and techniques to

residual-stress measurement. For example, in

conventional XRD analysis and crystallography,

sharp resolution of the diffracted spectra is very

beneficial. However, in XRD stress measurement the need to measure psi (w) angles that are

not zero defocuses the beam, and attempts to refocus leads to significant error in the stresses

read (Ref 83). In XRD stress measurement, the

ability to repeatedly measure the position of a

defocused diffracted x-ray peak is more important than sharp resolution (Ref 84). Thus, it is

recommended in most cases that XRD residualstress measurement be performed by trained

technologists using x-ray instrumentation specifically designed and built for stress measurement, not conventional scanning diffractometers.

Software packages specifically for residualstress measurement used with conventional

scanning diffractometers do not in most cases

eliminate the mechanical and focusing problems

of applying these instruments to residual-stress

measurement. It is necessary to mount the component (or specimen) in which stresses are to be

measured on the conventional scanning diffractometer, which usually requires sectioning of the

component and which complicates and adds error to the measurement procedure.

Neutron Diffraction

Neutron diffraction (ND) is capable of measuring the elastic strains induced by residual

stresses throughout the volume of relatively

thick steel components with a spatial resolution

as small as 1 mm3. Such capabilities provide for

residual-stress measurement inside components

without sectioning or layer removal. Principal

ND methods, as with XRD methods, measure the

spacing between crystallographic planes in a

component, and this spacing is affected by residual and applied stress. The spacing between a

selected set of crystallographic planes (u) is related to the angle of incidence and diffraction of

the wavelength of the monochromatic radiation

(k) by Braggs Law:

k 2d sin h

(Eq 76)

stress perpendicular to the diffracting crystallographic plane then is related to d by:

e

d do

sin h0

do

sin h sin h0

(Eq 77)

crystallographic planes. If the orientation of

principal stresses are known in the component,

the stress in any principal direction may be calculated by:

rA

E

(1 m)

m

(eA eB eC)

(1 2m)

(Eq 78)

eA, eB, and eC are the strains measured in the

corresponding principal directions, and E and m

are the Youngs modulus and Poissons ratio,

respectively. If the principal stress directions are

unknown, strains in at least six directions must

be measured to determine the residual stresses

acting on the volume of material in which strains

are being measured.

Limitations and Applications. For residualstress measurements in most alloys, especially

steels and cast irons, the unstressed spacing (do)

between crystallographic planes at the exact

point of strain measurement is not known and

not easily measured. This means that do or ho in

Eq 77 cannot be precisely established, and this

leads to various degrees of error in the accuracy

and precision of ND residual-stress measurements. This condition is aggravated by the fact

that the elemental composition, and thus do, vary

considerably within a component and markedly

41/ 8 in. (10.5 cm)

within the phase (e.g., martensite, austenite, ferrite) of the alloy at various locations. Additional

limitations are that the component must be

brought to a nuclear reactor, each strain measurement requires over an hour, a single stress

determination in one small volume of the component requires at least three strain measurements, and the measurements are very costly.

Nevertheless, the ND methods have been applied

to residual-stress measurements in steel weldments (Ref 85), cylindrical steel forgings, plastically deformed steel plate (Ref 86), steel rocket

case forgings (Ref 87), and many other types of

components.

Ultrasonic Velocity

The principle underlying the measurement of

stress and thus elastic strain by ultrasonic (acoustic) techniques is the phenomenon of an approximately linear change in ultrasound velocity with

applied stress. In addition, it has been shown that

under certain narrow conditions residual stress

can be measured by exploiting this phenomenon.

Stress is measured by inducing a sound wave in

the frequency of several megahertz into the

metal specimen and measuring the time of flight

or some other velocity-related parameter. Since

many other characteristics of metals besides

stress-induced elastic strain affect velocity, these

must be sorted out; however, neither the technology nor the fundamental knowledge for such

sorting is usually available. The great interest in

ultrasonic techniques for residual-stress measurement stems from their promise for three-dimensional nondestructive measurements within

the material.

Principle. A number of velocity-related phenomena have been used in various methods to

measure stress effects by ultrasound. All utilize

the deviation of the reaction of the metal from

the linearity of Hookes law of elasticity, r

Me, where r is stress, e is strain, and M is elastic

modulus. This has been referred to as the anharmonic property of the solid and may be represented by a power series r Me Ce2 De3

, where C is the third-order anharmonic

constant, D is the fourth, and so on. Most research done for stress measurement has used expressions in which terms past the third-order

constant, C, are dropped. Of the several anharmonic property effects that may be used to measure stress, the following are probably the most

exploited:

Velocity dependence on the elastic modulus

Dispersion of frequency amplitudes in sur-

face waves

11/ 4 in.

(3.3 cm)

waves

1.6 in. (12.9 cm)

Fig. 14

Engineering drawing of the miniature XRD stress measurement head showing major dimensions

stress strain law has been written as r Me

Ce2 and rewritten as r e(M Ce). The term

in parentheses is approximately related to the velocity of sound as qV 2 M Ce, where q is

the density of the medium and V is the velocity

of sound. This may be approximately rewritten

in terms of velocity dependence on strain as:

V

q 1 2M e

M

(Eq 79)

2M)/C (Ref 88).

A simple view of the dependency of ultrasonic

velocity on the elastic modulus and density may

be shown by rewriting the equation pV 2 M

Ce in terms of V, differentiating, and dividing

by V to yield an expression for DV/V. The result

will readily show that a fractional change in elastic modulus or density would affect the velocity.

The density of metal, for which the Poisson ratio

is near 0.3, obviously changes as a compressive

or tensile stress is placed on the specimen, and

it is reasonable that the speed of sound would

then change.

Limitations and Applications. Ultrasonic

technology offers a number of types of wave

modes in which to probe metals; these include

bulk waves, such as longitudinal and shear, and

surface waves usually confined to Rayleigh type.

Each mode offers many unique parameters for

extracting information. As has been discussed,

the primary effect of stress-induced strain on ultrasonic propagation in metals is on velocity.

This may be detected in a number of ways, including measurements of wave velocity, shear

wave birefringence, and dispersion. However,

there are other characteristics of metals that affect the ultrasonic velocity to the same degree as

stress. These include crystallographic texture,

microstresses, multiple phases, coherent precipitates, composition gradients, and dislocation

density and distribution.

Crecraft (Ref 89) discussed velocity effects,

manifested as texture-induced birefringence, and

the marked change seen with ultrasonic frequency. He also reported birefringence due to

cold-work in nickel-steel specimens, but did not

attempt to separate the cold-work effects in

terms of texture, dislocation density, and so

forth. In the early 1950s, Bradfield and Pursey

(Ref 90) and Pursey and Cox (Ref 91) reported

showing the influence of small degrees of texture

on ultrasonically measured elasticity in polycrystalline bars. They showed how the true isotropic elastic constants can be determined using

measurements of both longitudinal and shear

wave speeds along several directions. They presented stereographic charts that illustrated the relationship between elastic behavior of cubic

crystals and results of x-ray texture determinations.

McGonagle and Yun (Ref 92) noted the coldwork effects in a paper comparing XRD results

with Rayleigh wave velocity measurements. Boland et al. (Ref 93) also recognized that other

material properties can affect ultrasonic velocity

and recommended that methods be developed to

distinguish stress-induced velocity changes from

those from other sources.

James and Buck (Ref 94) pointed out that

since the third-order elastic constants for most

from the literature, ultrasonic stress measurement must be calibrated relative to the particular

material being investigated. In the same paper

they discounted the possible effect of mobile dislocations on the sound velocity in structural engineering metals with high yield strengths due

to the short dislocation loop lengths prevalent.

However, they did mention that crystallographic

preferred orientation (texture) during deformation or fatigue is capable of severely modifying

the elastic constants on which the sound velocity

depends.

Papadakis (Ref 95) noted marked velocity

changes for ultrasonic waves in various steel microstructures, and Moro et al. (Ref 96) measured

the effect of microstructural changes caused by

tempering on the ultrasonic velocity in low-alloy

steel. Tittman and Thompson (Ref 97) evaluated

the near-surface hardness of case-hardened steel

with Rayleigh waves; because hardness in this

case is a combination of composition, microstress, and macrostress, the velocity change was

due to a combined effect.

The temperature sensitivity of ultrasonic

stress measurements has also been cited as an

important source of error. Salama et al. (Ref 98,

99) proposed that this dependence be used as a

means to measure stress, but also noted the

marked effect of dislocations and did not address

a methodology of separating the stress from the

dislocation effect.

Much of the work cited above is concerned

with attempts to measure the effects of a variety

of material properties on the changes in ultrasonic velocity. However, there apparently is no

comprehensive study that demonstrates the capability of quantitatively separating stress effects

on ultrasonic propagation from other variables

found in structural metals, such as dislocation

density or crystallographic texture. Furthermore,

most of the studies cited observed velocity

changes in bulk waves. Velocity measurements

on these waves must be measured through the

thickness of a component, and, as most metallurgists recognize, obtaining uniform properties

through thicknesses greater than a few millimeters, especially in steels, is difficult. The subtle

property variations to which ultrasound velocity

is sensitive and the inherent lack of homogeneity

in engineering metals present additional serious

problems for through-thickness stress measurements.

In spite of the microstructural variations in

manufactured steel products, success in the application of ultrasonic methods to residual-stress

measurement has been achieved in specific

cases. One is in the measurement of hoop

stresses in railroad wheels (Ref 100). Here

changes or variations of the residual stress in the

hoop direction is of concern while that in the

radial or axial direction can often be assumed to

be constant or negligible. Some techniques then

for the measurement of the residual hoop stresses

has relied on normalizing the hoop velocity

against the axial (Ref 101). Also, European railroads have monitored ultrasonic velocity along

to residual-stress changes (Ref 102). Schramm

in his paper mentioned a number of approaches

for the application of ultrasound to the measurement of residual stresses in railroad wheels, and

these examples may find application in the measurement of residual stress in other axially symmetric shapes (Ref 100). Ultrasonic residualstress measurements have also been applied to

rails as reported by Egle and Bray (Ref 103) and

Bray and Leon-Salamanca (Ref 104).

The Barkhausen noise analysis technique

(BNA) is concerned with measuring the number

and magnitude of abrupt magnetic reorientations

made by expansion and contraction of the magnetic domains in a ferromagnetic metal. These

reorientations are observed as pulses somewhat

random in amplitude, duration, and temporal

separation and therefore are roughly described as

noise.

Applications. A few applications of BNA to

ferromagnetic metallic components have been

made, Gardner (Ref 105) mentions a number of

applications that include helicopter rotor blade

spans, autofrettaged gun tubes, gas turbine engine components, and rolling-element antifriction bearing components. In these examples, the

change in residual stresses caused by known service histories was measured.

Chait (Ref 106) qualitatively measured the residual-stress condition of a high-hardness laminar composite steel weldment and compared

some of the BNA data with XRD stress readings.

Sundstrom and Torronen (Ref 107) applied their

BNA method to a number of microstructural

measurements, including evaluation of grainsize measurement for low-carbon ferritic and ferritic-pearlitic steels, evaluation of anisotropy in

deep drawing and textured steels for electrical

applications, measurement of the degree of aging

in rimmed carbon steels, and pearlite morphology in steel wires. They have also measured iron

loss in magnetic material used for transformers

and have proposed using BNA for residual-stress

measurements, pointing out that quantitative results can be obtained if the material and its fabrication history are known and calibration is possible.

Most studies and applications of BNA to

stress measurement have focused on the uniaxial

stress state. However, Sundstrom and Torronen

(Ref 107) implied that the instrumentation they

used could simultaneously measure stress in two

directions to give biaxial stress conditions for

magnetic inspection of roller-bearing components, including BNA for monitoring residualstress change.

Summary. The BNA method certainly has

been demonstrated to be sensitive to the stress

condition in ferromagnetic materials (Ref 108).

Nevertheless, its possibilities for application are

limited by the condition that the material must

be ferromagnetic, the narrow total range of stress

sensitivity (i.e., 40 ksi, or 5.6 MPa), and the

shallow depth of measurement. The latter condition might be relieved by using magnetomechanical mechanical acoustic emission (MAE)

(Ref 109), an ultrasound analog to BNA. However, the sensitivity of either of these techniques

to other characteristics of metallic components

and the consequent need for calibration with a

nearly identical specimen severely restrict the

general applicability of BNA and MAE. Many

misapplications have been made that have severely damaged the reputation of the BNA methods (Ref 107, 110). Such restrictions can be removed only if the basic phenomena responsible

for the effect of microstructural properties on

BNA and MAE are understood and quantified in

terms of the signal.

Barkhausen noise analysis is not recommended where variations in elemental composition, phase composition, grain size, strain hardening, crystallographic texture grain shape, grain

orientation, carbide size and distribution, and

other microstructural characteristics accompany

variations in residual stress. A recent evaluation

of BNA by Allison and Hendricks (Ref 111)

confirms the uncertainty of BNA residual-stress

measurements.

5.

6.

7.

8.

9.

Only the destructive stress-relief, semidestructive hole-drilling, or nondestructive XRD

methods of residual-stress measurement are generally reliable over a broad range of steel alloys

displaying residual-stress fields induced by the

various manufacturing processes. Measurement

of residual stresses can be very expensive and

time consuming, and it is often worthwhile to

consult experts in the field before deciding on a

measurement method. Before an engineer or scientist who is inexperienced in residual-stress

measurement selects a method and attempts to

measure stresses he/she should consult someone

experienced in residual-stress measurement and

analysis.

10.

11.

12.

13.

REFERENCES

1. R. Gunnert, Method for Measuring TriAxial Residual Stresses, Document No.

X-184057-OE, Commission X of the International Institute of Welding, 1957;

Weld. Res. Abroad, Vol 4 (No. 10), 1958,

p 1725

2. M.G. Moore and W.P. Evans, Mathematical Corrections in Removal Layers in XRay Diffraction Residual Stress Analysis,

SAE Trans., Vol 66, 1958, p 340345

3. A. Constantinescu and P. Ballard, On the

Reconstruction Formulae of Subsurface

Residual Stresses after Matter Removal,

Fifth Int. Conf. Residual Stresses, Vol 2,

Institute of Technology, Linkopings University, Sweden, 1997, p 703708

4. Residual Stress Measurement by X-Ray

Diffraction, SAE J784a, Handbook Sup-

14.

15.

16.

A.J. Bush and F.J. Kromer, Residual

Stresses in a Shaft after Weld Repair and

Subsequent Stress Relief, SESA Spring

Meeting Paper No. A-16, Society for Exp.

Stress Analysis, 1979

C.O. Ruud, P.S. DiMascio, and J.J. Yavelak, Comparison of Three Residual Stress

Measurement Methods on a Mild Steel

Bar, Exp. Mech., Vol 25 (No. 4), Sept

1985, p 338343

J.A. Pineault and M.E. Brauss, In-Situ

Measurement of Residual and Applied

Stresses in Pressure Vessels and Pipeline

Using X-Ray Diffraction Techniques, Determining Material Characterization, Residual Stress and Integrity with NDE,

PUP-Vol 276, NDE-Vol 12, American Society of Mechanical Engineers, 1994

R. Wimpory, G.M. Swallow, and P. Lukas,

Neutron Diffraction Residual Stress Measurements in Carbon Steels, Fifth Int.

Conf. Residual Stresses, Vol 2, Institute of

Technology, Linkopings University, Sweden, 1997, p 676681

M.E. Brauss and J.A. Pineault, Residual

Strain Measurement of Steel Structures,

NDE for the Energy Industry, NDE-Vol

13, D.E. Bray, Ed., American Society of

Mechanical Engineers, 1995

C.O. Ruud and P.S. DiMascio, A Prediction of Residual Stress in Heavy Plate Butt

Weldments, J. Mater. Eng. Sys., Vol 31

(No. 1), Jan 1981, p 6265

C.O. Ruud and M.E. Jacobs, Residual

Stresses Induced by Slitting Copper Alloy

Strip, NDC of Materials VI, Plenum Press,

1994, p 413424

E.F. Rybicki and R.B. Stonesifer, Computation of Residual Stresses Due to Multipass Welds in Piping Systems, ASME

78-PVP-104, American Society of Mechanical Engineers, 1978

M. Ehlers, H. Muller, and D. Loke, Simulation of Stresses and Residual Stresses

Due to Immersion Cooling of Tempering

Steel, Fifth Int. Conf. Residual Stresses,

Vol 2, Institute of Technology, Linkopings

University, Sweden, 1997, p 400405

K. Masubuchi, Analysis of Welded Structures: Residual Stresses, Distortion, and

Their Consequences, 1st ed., Pergamon

Press, 1980

M. Bijak-Zochowski, P. Marek, and M.

Tracz, On Subsurface Distributions of Residual Stress Created by Elasto-Plastic

Rolling Contact, Fifth Int. Conf. Residual

Stresses, Vol 2, Institute of Technology,

Linkopings University, Sweden, 1997, p

430445

D. Green and S. Bate, Calculation of Residual Stresses Using Simplified Weld

Bead Modelling Technique, Fifth Int.

Conf. Residual Stresses, Vol 2, Institute of

Technology, Linkopings University, Sweden, 1997, p 508513

Three Dimensional Representation of the

Residual Stress Distribution in Steel Wires

or Bars, The Fifth Int. Conf. Residual

Stresses, Vol 2, Institute of Technology,

Linkopings University, Sweden, 1997, p

534538

18. K.W. Mahin, W.S. Winters, T. Holden,

and J. Root, Measurement and Prediction

of Residual Elastic Strain Distributions in

Stationary and Traveling Gas Tungsten

Arc Welds, Practical Applications of Residual Stress Technology, ASM International, 1991, p 103109

19. J. Pineault and M. Brauss, Automated

Stress MappingA New Tool for the

Characterization of Residual Stress and

Stress Gradients, Proc. Fourth Int. Conf.

Residual Stresses, Society for Experimental Mechanics, 1994, p 4044

20. M. Brauss, J. Pineault, S. Teodoropol, M.

Belassel, R. Mayrbaurl, and C. Sheridan,

Deadload Stress Measurement on Brooklyn Bridge Wrought Iron Eye Bars and

Truss Sections Using X-Ray Diffraction

Techniques, ICB-9751, Proc. 14th Annual International Bridge Conf. and Exhib., Engineering Society of Western PA,

1997, p 457464

21. M.G. Carfaguo, F.S. Noorai, M.E. Brauss,

and J.A. Pineault, X-Ray Diffraction

Measurement of Stresses in Post-Tensioning Tendons, IABSE Symposium (San

Francisco, CA) 1995, Extending the Life

Span of Structures, Vol 71/1, International

Association for Bridge and Struct. Eng.,

ETH Honggerberg, CH-8093, Zurich,

Switzerland, 1995, p 201206

22. C.O. Ruud, P.S. DiMascio, and D.J.

Snoha, A Miniature Instrument for Residual Stress Measurement, Adv. X-Ray Anal.,

Vol 27, Plenum Press, 1984, p 273283

23. C.O. Ruud, R.N. Pangborn, P.S. DiMascio,

and P.J. Snoha, Residual Stress Measurement on Thick Plate Low-Alloy Steel Narrow Gap Weldments by X-Ray Diffraction, ASME 84-PVP-128, American

Society of Mechanical Engineers, 1984

24. C.O. Ruud, J.A. Josef, and D.J. Snoha, Residual Stress Characterization of ThickPlate Weldments Using X-Ray Diffraction,

Weld. Res. J., Supplement, March, 1993, p

875915

25. P. Johanssen, Determination of Residual

Stresses in Thick Welded Plated Utilizing

X-Rays and Layer Removal Techniques,

Part 1: Theory for Reconstruction of Original Stress State, Internal Report, Dept. of

Materials and Solid Mechanics, The Royal

Institute of Technology, Stockholm, Sweden, April 1978

26. E. Heyn, Internal Strains in Cold Wrought

Metals, and Some Troubled Caused

Thereby, J. Inst. Met., Vol 12, 1914, p 1

37

27. E. Stablein, Stress Measurements on Billets Quenched from One Side, Stahl Eisen,

Vol 52, 1932, p 1517

28. M. Mesnager, Methods de Determination

dees Tensions Existant dans un Cylindre

Circulative, C.R. Hebed. Seances Acad.

Sci., 1926, p 169

29. G. Sacks, Evidence of Residual Stresses in

Rods and Tubes, Feitsch. Metallkde., Vol

19, 1927, p 352357

30. R.G. Treuting and W.T. Read, Jr., J. Appl.

Phys., Vol 22 (No. 2), Feb 1951, p 130

134

31. R. Gunnert, Residual Welding Stresses,

Method of Measuring Residual Stresses

and Its Application to a Study of Residual

Welding Stresses, Alquist E. Wicksell,

Stockholm, 1955

32. R. Gunnert, Method for Measuring Residual Stresses in the Interior of a Material,

Document No. X-16257, Commission X

of the International Institute of Welding,

1957; Weld. Res. Abroad, Vol 6, 1960, p

1024

33. D. Rosenthal and J.T. Norton, A Method

for Measuring Triaxial Residual Stresses

in Plates, Weld. J., Vol 24 (No. 5), Research Supplement, 1945, p 295s307s

34. R. J.-S. Chen, Determination of Residual

Stresses in a Thick Weldment, M.S. Thesis, Department of Engineering Science

and Mechanics, The Pennsylvania State

University, Nov 1983

35. G. Pickel, Application of the Fourier

Method to the Solution of Certain Boundary Problems in the Theory of Elasticity,

J. Appl. Mech., 1944, p 176182

36. D.L. Sikarskie, On a Series Form of Correction to Stresses Measured Using X-Ray

Diffraction, AIME Trans., Vol 39, 1967, p

577580

37. A.R. McIlree, C.O. Ruud, and M.E. Jacobs, The Residual Stresses in Stress Corrosion Performance of Roller Expanded Inconel 600 Steam Generator Tubing, Proc.

Int. Conf. Expanded and Rolled Joint

Technology, Canadian Nucl. Society,

1993, p 139148

38. Y.W. Park, P.H. Cohen, and C.O. Ruud,

The Development of a Model for Plastic

Deformation in Machined Surface, Mater.

Manuf. Proc., Vol 8 (No. 5), 1994

39. Y.W. Park, P.H. Cohen, and C.O. Ruud,

Sensitivity of Shear Process on Metal Cutting to the Development of Residual

Stress, Rev. Prog. NDE, Vol 14, 1995, p

11831188

40. Y.C. Shin, S.J. Oh, and C.O. Ruud, Investigation of Residual Stresses of Machined

Surfaces by an X-Ray Diffraction Technique, NDC of Materials IV, Plenum Press,

1992, p 408418

41. J.E. Hoffman, D. Lohe, and E. Macherauch, Influence of Machining Residual

Stresses on the Bending Fatigue Behaviors

of Notched Specimens of Ck45 in Different Heat Treating States, Residual Stresses

in Science and Technology, Vol 2, DGM

Informationsgesellschaft Verlag, Germany, 1987, p 801814

Metal Cutting, Residual Stresses in Science and Technology, Vol 2, DGM Informationsgesellschaft Verlag, Germany,

1987, p 839846

43. E. Schreiber and H. Schlicht, Residual

Stresses after Turning of Hardened Components, Residual Stresses in Science and

Technology, Vol 2, 853860, DGM Informationsgesellschaft Verlag, Germany,

1987, p 853860

44. F. Ghanem, H. Dishom, C. Braham, R. Fatallak, and J. Leider, An Engineering Approach to the Residual Stresses Due to

Electric-Discharge Machining Process,

Proc. Fifth Int. Conf. on Residual Stresses,

Vol 1, Ericson, Oden, and Anderson, Ed.,

ICRS-5, Institute of Technology, Linkopings University, Sweden, 1977

45. Electropolishing, Surface Cleaning, Finishing and Coating, Vol 5, 9th ed., Metals

Handbook, American Society for Metals,

1982, p 303309

46. Hetenyi, Ed., Handbook of Experimental

Stress Analysis, John Wiley & Sons, 1950

47. N.A. Crites, Your Guide to Todays Strain

Gages, Prod. Eng., Vol 33 (No. 4), 17 Feb

1962, p 6981; Equipment and ApplicationTodays Strain Gages, Prod. Eng.,

Vol 33 (No. 6), 19 March 1962, p 8593

48. Y.Y. Hung, Shearography: A New Optical

Method for Strain Measurement and Nondestructive Testing, Opt. Eng., Vol 21 (No.

3), 1982, p 391394

49. K. Li, Interferometric Strain/Slope Rosette

Technique for Measuring Displacements/

Slopes/Strains/ Residual Stresses, Proc.

of the 1997 NSF Design and Manufacturing Grantees Conf., University of Washington (Seattle, WA), National Science

Foundation, Arlington, VA, 1997, p 571

572

50. C.S. Vikram, M.J. Pedensky, C. Feng, and

D. Englehaupt, Residual Stress Analysis

by Local Laser Heating and Speckle-Correlation Interferometry, Exp. Tech., Nov/

Dec 1996, p 2730

51. B.D. Cullity, Elements of X-Ray Diffraction, 2nd ed., Addison-Wesley, 1978, p

469472

52. C.O. Ruud, Residual and Applied Stress

Analysis of an Alloy 600 Row 1 U-Bend,

NP-5282, R.P 53033, Electric Power Research Institute, 1987, p 211

53. A. Bouzina, C. Braham, and J. Ledion,

Evaluation and Prediction of Real Stress

State Stress Corrosion Cracking Specimens, Fifth Int. Conf. Residual Stresses,

Vol 2, Institute of Technology, Linkopings

University, Sweden, 1997, p 10601065

54. K. Masubuchi and D.C. Martin, Investigation of Residual Stresses by Use of Hydrogen Cracking, Weld. J., Vol 40 (No.

12), Research Supplement, 1961, p 553s

556s

55. C.E. McKinsey, Effect of Low-Temperature Stress Relieving on Stress Corrosion

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

Research Supplement, p 161s166s

W. Radeker, A New Method for Proving

the Existence of Internal Stress Caused by

Welding, Schneissen Schneider, Vol 10

(No. 9), 1958, p 351358 (see also abstract

by Claussen, Weld. J., Vol 38 (No. 4)

1959, p 300

J. Mathar, Determination of Initial Stresses

by Measuring Deformation around Drilled

Holes, Arch. Eisenhuttenwes., Vol 6, 1934,

p 277281; Trans. ASME, Vol 56 (No. 4),

1934, p 249254

Determining Residual Stresses by the

Hole Drilling Strain-Gage Method, E

837, Annual Book of ASTM Standards,

ASTM, 1983

H. Wolf and D.C. Sauer, New Experimental Technique to Determine Residual

Stresses in Large Turbine-Generator Components, presented at the Amer. Power

Conf. (Chicago, IL), 1 May 1974

H. Wolf and W. Bohn, Origin, Measurement and Assessment of Residual Stresses

in Large Forging for Turbines and Generators, Arch. Eisenhuttenwes., Vol 42 (No.

7), 1971, p 509511 (in German)

H. Wolf, E. Stucker, and H. Nowack, Investigations of Residual Stresses in the

Turbine and Generator Industry, Arch. Eisenhuttenwes., Vol 48 (No. 3), March

1977, p 173178

J. Lu and J.F. Flavenot, Applications of the

Incremental Hole-Drilling Method for

Measurement of Residual Stress Distributions, Experimental Techniques, Presented

at the SEM Spring Conference (Portland,

OR), Society of Exp. Mechanics, 510

June 1988

M.T. Flaman and B.H. Manning, Determination of Residual Stress Variation with

Depth by the Hole-Drilling Method, Exp.

Mech., Vol 25, 1985, p 205207

R.A. Kelsey, Measuring Non-Uniform Residual Stresses by the Hole Drilling

Method, Proc. Soc. Exp. Stress Anal., Vol

14 (No. 1), 1956, p 181184

A.J. Bush and F.J. Kromer, Simplification

of the Hole-Drilling Method of Residual

Stress Measurement, ISA Trans., Vol 12

(No. 3), 1973, p 249259

N.J. Rendler and I. Vigness, Hole-Drilling

Strain-Gage Method of Measuring Residual Stress, Exp. Mech., Vol 6 (No. 12) Dec

1966, p 577586

J.W. Dini, G.A. Beneditti, and H.R. Johnson, Residual Stresses in Thick ElectroDeposits of a Nickel-Cobalt Alloy, Exp.

Mech., Feb 1976, p 5660

F. Witt, F. Lee, and W. Rider, A Comparison of Residual Stress Measurements Using Blind-Hole, Abrasive-Jet and Trepanning Methods, Exp. Tech., Vol 7, Feb

1983, p 4145

P.S. Prevey, Residual Stress Distribution

Produced by Strain Gage Surface Prepa-

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

S. Kokubo, Changes in Hardness of a Plate

Caused by Bending, Sci. Rep. Tohoku Imperial Univ., Jpn., Series 1, Vol 21, 1932,

p 256267

G. Sines and R. Carlson, Hardness Measurements for the Determination of Residual Stresses, ASTM Bull., No. 180, Feb

1952, p 3537

J. Pomey, F. Goratel, and L. Abel, Determination des Cartraintes Residuelles dans

les Pieces Einentees, Publication Scientifiques et Techniques du Ministere de lair,

1950, p 263

S.S. Chiang, D.B. Marshall, and A.G.

Evans, The Response of Solids to Elastic/

Plastic Indentations I: Stresses and Residual Stresses, J. Appl. Phys., Vol 53 (No.

1), 1982, p 291311

C.O. Ruud, X-Ray Analysis and Advances

in Portable Field Instrumentation, J. Met.,

Vol 31 (No. 6), June 1979, p 1015

H.P. Klug and L.E. Alexander, X-Ray Diffraction Procedures, 2nd ed., John Wiley

& Sons, 1974, p 768770

C.S. Barrett and T.B. Massalski, Structure

of Metals, 3rd ed., McGraw-Hill, 1966, p

474476

C.O. Ruud and G.D. Farmer, Residual

Stress Measurement by X-Rays: Errors,

Limitations, and Applications, Nondestructive Evaluation of Materials, J.J.

Burke and V. Weiss, Ed., Plenum Press,

1979, p 101116

R.N. Pangborn, S. Weissman, and I.R.

Kramer, Dislocation Distribution and Prediction of Fatigue Damage, Metall. Trans.,

Vol 12A (No. 1), 1981, p 109120

D.A. Bolstad, R.A. Davis, W.E. Quist, and

E.C. Roberts, Measuring Stress in Steel

Parts by X-Ray Diffraction, Met. Prog.,

1963, p 8892

D.S. Kurtz, P.R. Moran, K.J. Kozaczek,

and M. Brauss, Apparatus for Rapid Psi

Squared Stress Measurement and Its Applications to Titanium Alloy Jet Engine

Fan Blades, Fifth Int. Conf. Residual

Stresses, Vol 2, Institute of Technology,

Linkopings University, Sweden, 1997, p

744749

B.B. He and K.L. Smith, Strain and Stress

Measurements with Two-Dimensional Detector, Adv. X-Ray Anal., Vol 41, 1977

M.R. James and J.B. Cohen, The Application of a Position-Sensitive X-Ray Detector to the Measurements of Residual

Stresses, Adv. X-Ray Anal., Vol 19, Gould,

Barnett, Newkirk, and Ruud, Ed., 1976, p

695708

H. Zantopulou and C.F. Jatczak, Systematic Errors in Residual Stress Measurement

Due to Specimen Geometry, Adv. X-Ray

Anal., Vol 14, Plenum Press, 1970

C.O. Ruud, D.J. Snoha, and D.P. Ivkovich,

Experimental Methods for Determination

of Precision and Estimation of Accuracy in

85.

86.

87.

88.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

X-Ray Anal., Vol 30, 1987, p 511522

J.H. Root, C.E. Coleman, J.W. Bowden,

and M. Hayashi, Residual Stresses in Steel

and Zirconium Weldments, J. Press. Vessel Technol., Vol 119, 1997, p 137141

M. Hayaski, S. Ohkido, N. Minakawa, and

Y. Murii, Residual Stress Distribution

Measurement in Plastically Bent Carbon

Steel by Neutron Diffraction, Fifth Int.

Conf. Residual Stresses, Vol 2, Institute of

Technology, Linkopings University, Sweden, 1997, p 762769

J.H. Root, R.R. Hosbaus, and T.M.

Holden, Neutron Diffraction Measurements of Residual Stresses Near a Pin Hole

in a Solid-Fuel Booster Rocket Casing,

Practical Applications of Residual Stress

Technology, C.O. Ruud, Ed., ASM International, 1991, p 8393

G.A. Alers, Ultrasonic MethodsOverview, Proc. of a Workshop on Nondestructive Evaluation of Residual Stress,

NTIAC-762, 1975, p 155161

D.I. Crecraft, Ultrasonic Measurement of

Stresses, Ultrasonics, 1968, p 117121

G. Bradfield and H. Pursey, Philos. Mag.,

Vol 44, 1953, p 295

H. Pursey and H.L. Cox, Philos. Mag., Vol

45, 1954, p 295

W.J. McGonagle and S.S. Yun, Measurement of Surface-Residual Stress by Nondestructive Methods, Proc. Fifth Int. Conf.

Nondestructive Testing (Montreal, Canada), May 2125 1967, p 159164

A.J. Boland et al., Development of Ultrasonic Tonography for Residual Stress

Mapping, EPRI RP 5042, Final Report,

Electric Power Research Institute, 1980

M.R. James and O. Buck, Quantitative

Nondestructive Measurement of Residual

Stresses, Crit. Rev. Solid State Mater. Sci.,

Aug 1980, p 61105

E.P. Papadakis, Ultrasonic Attenuation

and Velocity in Three Transformation

Products of Steel, J. Appl. Phys., Vol 35

(No. 5), 1964, p 14741482

A. Moro, C. Farina, and F. Rossi, Measurement of Ultrasonic Wave Velocity of

Steel for Various Structures and Degrees

of Cold-Working, NDT Int., Aug 1980, p

169175

B.R. Tittman and R.B. Thompson, Measurement of Physical Property Gradients

with Elastic Surface Wave Dispersion,

Proc. Ninth Symp. on NDE, Southwest Research Institute, 1980, p 2028

K. Salama and C.K. Ling, The Effect of

Stress on The Temperature Dependence of

Ultrasound Velocity, J. Appl. Phys., Vol

51 (No. 3), March 1980, p 15051509

K. Salama and G.A. Alers, Third-Order

Elastic Constants of Copper at Low Temperature, Phys. Rev., Vol 161 (No. 3), Sept

1967, p 673680

Clark, Jr., Ultrasonic Measurement of Residual Stress in the Rims of Inductively

Heated Railroad Wheels, Mater. Eval.,

Aug 1996, p 929933

101. H. Fukuaka, H. Tada, K. Hirakawa, H.

Sakawoto, and Y. Toyo, Acoustoelastic

Measurements of Residual Stresses in the

Rim of Railroad Wheels, Wave Propagation in Inhomogenous Media and Ultrasonic Nondestructive Evaluation, Vol 6,

G.C. Johnson, Ed., ASME, 1984, p 185

193

102. E.R. Schneider, R. Harzer, D. Bruche, and

H. Frotscher, Reliability Assurance of

Railroad Wheels by Ultrasonic Stress

Analysis, Proc. Third European Conf. on

Residual Stress Analysis (Frankfurt, Germany), 46 Nov 1992

103. D.M. Egle and D.E. Bray, Ultrasonic Measurement of Longitudinal Rail Stresses,

Mater. Eval., Vol 378 (No. 4), March

1979, p 4146

104. D.E. Bray and T. Leon-Salamanca, ZeroForce Travel-Time Parameters of Ultrasonic Head-Waves in Railroad Rail, Mater. Eval., Vol 43 (No. 7), June 1985, p

854858

105. C.G. Gardner, Barkhausen Noise Analysis,

Proc. Workshop on Nondestructive Evaluation of Residual Stress, NTIAC-762,

Nondestructive Testing Information Analysis Center, 1975, p 211217

106. R. Chait, Residual Stress Pattern in a High

Hardness Laminar Composite Steel Weldment, Proc. Workshop on Nondestructive

Evaluation of Residual Stress, NTIAC-76

2, Nondestructive Testing Information

Analysis Center, 1975, p 237245

107. O. Sundstrom and K. Torronen, The Use

of Barkhausen Noise Analysis in Nondestructive Testing, Mater. Eval., Feb 1979,

p 5156

108. J.R. Barton, F.N. Kusenberger, R.E. Beissner, and G.A. Matskanin, Advanced

Quantitative Magnetic Nondestructive

Evaluation Methods, Theory and Experiment, Nondestructive Evaluation of Materials, J.J. Burke and V. Weiss, Ed., Plenum

Press, 1979, p 461463

109. K. Ono, M. Shibata, and M.M. Kwan, Determination of Residual Stress by Magnetomechanical Acoustic Emission, UCLA

Department of Materials, ONR Tech. Rep.

No. 8001, Conference on Residual Stress

for Designers and Engineers, ASM International, April 1980

110. R.R. King and V.D. Smith, Residual

Stress Measurements in Structural Steels,

Final Report SWRI Project 154600, Contract DOT-FH-119133, Federal Highway

Administration, April 1981

111. H.D. Allison and R.W. Hendricks, Correlation of Barkhausen Noise Signal and XRay Residual Stress Determinations in

Grinding-Burned 52100 Steel, Fifth Int.

Conf. Residual Stresses, Vol 2, Institute of

Technology, Linkopings University, Sweden, 1997, p 640645

G. Totten, M. Howes, T. Inoue, editors, p118-124

DOI: 10.1361/hrsd2002p118

All rights reserved.

www.asminternational.org

COATINGS AND THIN FILMS can be produced by a large variety of deposition techniques. Typical processes are physical vapor

deposition (PVD), chemical vapor deposition

(CVD), electroplating, electroless deposition,

anodizing, thermal growth, and thermal spraying. Since the early 1980s, considerable progress

has been made in improving deposition processes for a wide range of high-technology applications. Consequently, many new ceramic

coatings and films have been introduced in various industries. Typical examples are metal-oxide

semiconductors for microelectronics; titanium

nitride, titanium carbide, aluminum oxide, and

silicon nitride for machining tools; and thermal

sprayed tungsten carbide/cobalt, M-chromiumaluminum-yttrium (where M stands for iron, cobalt, or nickel), and yttrium oxide/partially stabilized zirconia coatings for aerospace

applications.

Residual stresses, which are internal and

therefore locked in, are contained in materials

that are produced by nearly every mechanical,

chemical, and thermal process, either alone or in

combination. As a result, most coatings are in a

state of internal stress, including metallics and

ceramics. The stress can be either compressive

or tensile. It is generally recognized that compressive stresses in coatings are more favorable

than tensile stresses because they increase resistance to fatigue failure. However, extremely high

compressive stresses may cause either coating

separation from the base metal or intracoating

spallation. Generally, if a tensile stress causes

strain that exceeds the elastic limit of the coating,

then it will cause cracking in the coating perpendicular to the direction of the stress. Therefore,

understanding the formation of residual stress in

the coating is important to prevent the coating

from peeling or cracking during service. Furthermore, residual stresses have significant influences on the mechanical and physical properties

of the coatings, particularly electrical resistivity,

optical reflectance, fatigue, and corrosion.

There are three types of residual stresses:

Macrostresses, which are nearly homoge-

Microstresses, which are nearly homogeneous over microscopic areas, such as one

grain or subgrain

Residual macrostresses are the ones of most interest in engineering practice because they can

substantially affect component service performance. Both residual and inhomogeneous microstresses are of more interest in materials science.

This article intends to provide a useful guide

for measuring residual macrostress on a coating.

The most commonly used measurement methods

are mechanical deflection, x-ray diffraction, and

hole-drilling strain gage. After a discussion on

the origins of residual stress, the fundamental

principles, as well as examples of practical measurements, are described for each method.

Residual macrostress in a coating combines

the intrinsic stress and the thermal stress acting

in the coating plane parallel to the coating/substrate interface:

rt ri rth

(Eq 1)

are intrinsic stress and thermal stress, respectively. Intrinsic stress results from the growth

processes, depending primarily on deposition

parameters, whereas thermal stress arises from a

mismatch in coefficients of thermal expansion

between the coating and the substrate.

Many phenomenological models have been

proposed to explain the occurrence of intrinsic

stresses by correlating them with a variety of

coating microstructure and process features. To

varying degrees, the intrinsic stress of a coating

is associated with these deposition conditions

and coating features:

Incorporation of residual gas atoms in the

coating

in the coating

a growing coating

Although many studies have described the intrinsic stresses, information on the corresponding structural details is limited. It seems unlikely

that one can formulate a generalized model of

intrinsic stress for various coating materials and

deposition processes.

Any coating that is prepared at elevated temperatures (T2) and then cooled to room temperature (stress measurement temperature, T1) will be

thermally stressed because of the difference in

the coefficients of thermal expansion between

the coating and the substrate. Assuming no deformation of the substrate, the magnitude of the

thermal stress in the coating is:

rth (c s)(T2 T1)Ec /(1 mc)

(Eq 2)

where c and s are the thermal expansion coefficients for the coating and the substrate, respectively, and Ec and mc are the Youngs modulus and Poissons ratio of the coating,

respectively. A coating deposited at an elevated

temperature exhibits compressive stress if s

c, but tensile stress if s c. In the case of s

c, the substrate shrinks more than the coating

does during cooling from the deposition temperature and compresses the coating to maintain

dimensional compatibility.

In some cases, thermal stress is the primary

residual stress of the coating. For example, a titanium nitride coating can be deposited on a cemented carbide substrate (tungsten carbide-1

wt% Ta-10 wt% Co) via a CVD process at 1000

C (1830 F). With the values of s 5

106 /K, c 9.54 106 /K, ETiN 411 GPa

(60 106 psi), and mTiN 0.24, the stress in

the coating, as calculated from Eq 2, is 2.39 GPa

(0.35 106 psi) in tension at 25 C (77 F).

Deflection Method

ing growth

Lattice misfit between the substrate and the

growing coating

technique for determining the residual stress in

*This article is adapted from Stress Determination for Coatings by J. Albert Sue and Gary S. Schajer in Surface Engineering, Volume 5, ASM Handbook, ASM International, 1994, p 647

653.

a coating. In terms of basic principles, it involves

measuring the amount of bending in a strip that

is due to the deposition of the coating. A formula

for calculating the residual stress in an electrodeposited coating was derived first by Stoney

(Ref 1) in 1909. Subsequent researchers have derived more complex formulas to improve the accuracy of the stress evaluation. Comparative

studies on those stress-evaluation formulas are

reviewed and discussed elsewhere (Ref 24).

Nearly all formulas are variants of Stoneys formula.

Consider a coating deposited on one side of a

strip substrate. Both coating and substrate are assumed to be homogeneous. A mismatched force

at the coating/substrate interface results in residual stress in the coating, which bends the strip

either upward (concave) or downward (convex),

depending on whether the stress is tensile or

compressive. For overall force and moment

equilibrium of the coating/substrate composite,

it can be shown that the residual stress in the

coating is (Ref 5):

rc (Es d 2s )/[6(1 ms)Rdc](ds dc)

{(1 CH3)/(1 H)}

(Eq 4)

where

C [Ec(1 ms)]/[Es(1 mc)]

H ds /dc

R L2 /8f L2 /2d

(Eq 5)

d is thickness, with the subscript c denoting coating

and s denoting substrate; R is the radius of curvature of the bent strip, L is the length of the strip, f

is the deflection from the free end of the strip, and

d is the deflection at the center of the strip

(Fig. 1).

(Eq 3)

or, in general:

and represents a generalized Stoneys formula

for a planar state of stress, it tends to overestimate the value of measured stress. However, it

does not require knowledge of the elastic properties of the coating. Equation 4 provides a much

better approximation than Eq 3, but it does require knowledge of elastic properties. Equation

4 differs from Eq 3 in terms of a correction factor

[(1 CH3)/(1 H)]. Table 1 compares

stresses as a function of ds /dc, calculated from

both Eq 3 and 4. In the calculation, it was assumed that Ec 400 GPa (60 106 psi), Es

200 GPa (30 106 psi), and mc ms. It is clear

that both equations are in good agreement when

ds /dc 50. By choosing a large value of ds /dc,

the error can be minimized if Eq 3 is used.

Cantilever Beams. To measure the deflection

of cantilever beams, various techniques have

been developed (Ref 7), including optical, capacitance, mechanical, electromechanical, interferometric, and electromechanical or magnetic

restoration. Figure 1(a) shows the setup for the

deflection measurement when an optical system

is used. In this example, a titanium nitride coating was deposited on a quartz beam via a PVD

process, and the average residual stress in the

coating was determined by measuring the

amount of deflection at the free end, f, of the bent

beam, according to the relation:

rc [4Es d s2 f ]/[3(1 rs)L2 dc]

(Eq 6)

psi), ms 0.16, ds 3 mm (0.12 in.), dc 3

106 m (1.2 104 in.), L 50 mm (2

in.), and f 2 106 m (8.0 105 in.), the

stress in the coating is determined to be 273 MPa

(40 ksi) in compression.

Disks. The average stress in the coating on a

disk substrate can be determined from the

amount of deflection, d, at the center of the disk

caused by the deposition of a coating on one

side. This value can be measured optically by

interferometry or microstylus profilometry. Figure 1(b) shows a typical interferometry apparatus

setup. Either technique measures deflection at

the same position, across a diameter of the disk,

both before and after coating deposition. The

stress in the coating is then calculated:

rc {Es d s2[(Ddx Ddy)/2]}/[3(1 ms)r2dc]

Fig. 1

cantilever beam. (b) Disk deflection. Source:

(Eq 7)

Ref 6

stress calculated using Eq 3 and 4

Calculated stress

Eq 3

Eq 4

ds /dc

MPa

ksi

MPa

ksi

200

100

50

20

10

1143

571

286

114

57

165.7

82.8

41.5

16.5

8.3

1137

565

280

109

52

164.9

81.9

40.6

15.8

7.5

ry. (b) rx ry, where both components are in

compression or tension. (c) rx ry, where one component

is in compression and the other is in tension. Source:

Ref 8

Fig. 2

where Ddx and Ddy are deflection changes measured before and after coating deposition, and r

is the radius of the disk. Whether the stress is

tensile or compressive, it is determined by the

curvature of the disk before and after coating

deposition, as determined by a depth microscope. When the change in curvature is upward

(concave), viewed from the coating side, the

stress in the coating is tensile. The opposite

change in curvature indicates a compressive

stress in the coating.

Typical interference fringe patterns are shown

in Fig. 2. Figures 2(a) and (b) correspond to

equal and unequal principal stresses of the same

sign, whereas Fig. 2(c) corresponds to a biaxial

stress state with principal stresses of opposite

signs. Using Fig. 2(a) as an example, the measured deflection, d, is equal to the wavelength of

monochromatic light multiplied by the number

of light fringes (d 632.8 nm 17).

Figure 3 shows typical microstylus traces on

the titanium nitride coated surface of an AISI

304 stainless steel disk, both before and after

coating deposition. The curvature of the coated

surface was downward (convex), viewed from

the coated side. With the values of Es 193

GPa (28 106 psi), ms 0.28, ds 4.74 mm

(0.186 in.), r 11.94 mm (0.4700 in.), dc 20

lm (800 lin.), and the deflection Dd 6.3

106 m (2.5 104 in.), the stress in the titanium nitride coating is 4.44 GPa (0.644 106

psi) in compression. The measurement errors

from both the optical and the microstylus trace

setups are within one-half light band and 0.1

lm (4 lin.), respectively.

Practical Considerations. First, either strip

or disk-shape substrate specimens should be parallel within 0.02 mm (0.001 in.) over their length

or diameter. Specimen edges should be free from

visible flaws and chips. In addition, the surfaces

on which the deflection measurements (interferometer or profilometer) will be performed should

be ground and lapped to a finish better than 0.08

lm (3 lin.) Ra.

Second, the dimensions of a strip of length L,

width w, and substrate thickness ds should obey

L 10w 10 ds. The substrate thickness of a

strip or disk depends on the coating thickness to

be deposited. The thickness ratio of the substrate

to the coating, ds /dc, should be greater than 50

to ensure the accuracy of the stress calculation

from Eq 3.

Third, substrate specimens should be stress relieved before coating deposition. They should be

placed between two stainless steel surface plates

under at least an 8.8 kPa (1.3 psi) normal load,

at an annealing temperature, in a vacuum furnace

dependent on the substrate material.

Fourth, for coatings with an inherently smooth

surface, such as those produced by PVD and

CVD, the deflection can be determined either by

interference fringe or microstylus profilometer

measurements on the coated surface. For a coating with a relatively rough surface, such as those

produced by thermal spraying and electroplating, the measurements can be made on the surface that is opposite the coated surface.

Significance and Use. The deflection measurement method is recommended for determining the average stress in the cross section of a

coating with a thickness ranging from several

hundred angstroms to several hundred micrometers. Typically, the elastic constants of a thin

coating are much different from those of a bulk

material. Equation 3 provides a means for stress

measurement on a thin coating without any

knowledge of its elastic constants. The measurement normally applies only to a test sample. The

disk deflection method is particularly useful for

the direct inspection of silicon wafers used in

solar cells or integrated circuits.

Basic Principles. Stress measurement, using

the x-ray diffraction method, is based on the

change in the interplanar spacing (strain) close

to the surface of the specimen material. The details of the theory and interpretation of residual

stress measurements are well described in the article X-Ray Diffraction Residual Stress Techniques in Volume 10 of the ASM Handbook, as

well as in Ref 9 to 11. Consider an isotropic

material with a lattice parameter d hkl

for un0

stressed material in the sample plane normal.

The strain in a direction inclined by an angle y

to the surface normal of the coating and the stress

acting in the surface plane of the coating at an

angle u with the principal axis of the specimen

are related by:

hkl

hkl

hkl

ehkl

uw (d uw d 0 )/d 0

(rx cos2 u sxy sin2 u ry sin2 u rz)

sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)

(rx ry rz) [(1 m hkl)/E hkl]

(sxz cos u syz sin u) sin w)

2

Typical microstylus trace on titanium nitridecoated AISI 304 stainless steel disk before and

after coating deposition

Fig. 3

using Eq 8.

Biaxial Stress. At a free plane, the out-ofplane stress components rz, sxz, and syz are all

zero, at a free surface. Because the penetration

depth of x-rays is very small, the resulting measurements refer specifically to near-surface material. Plane stress conditions, therefore, often

apply to x-ray measurements, and Eq 8 is simplified to:

hkl

hkl hkl

hkl

hkl

ehkl

uw (d uw d 0 /d 0 [(1 m )/E ]

where ru rx cos2 u ry sin2 u is the macrostress in the coating parallel to its surface at

an angle u with the principal axis of the sample.

For a biaxial stress state, rx ry ru, and

at w 0:

hkl

hkl

hkl

hkl

(d hkl

u,w0 d 0 )/d 0 2m ru/E

(Eq 10)

hkl

hkl

hkl

hkl

2

d hkl

uw d u,w0 rud 0 [(1 m )/E ] sin w

(Eq 11)

(hkl) reflection are obtained from w to w at

a given angle u. Lorentz polarization, absorption, and background corrections are applied to

the diffraction peak profile. The peak positions

are determined by profile fitting or other methods

and are subsequently converted to interplanar

spacing d hkl

uw for stress analysis.

2

In the linear plot of d hkl

uw versus sin w, the

intercept is I d hkl

,

and

the

slope

is M

u,w0

hkl

hkl

rud hkl

]. The stress of the coating

0 [(1 m )/E

can then be determined by:

ru M/{d 0hkl[(1 m hkl)/E hkl]}

(Eq 12)

a material in a three-dimensional (triaxial) stress

state without shear stress, but with the stress

component rz having a finite value within the xray penetration volume, Eq 8 becomes (Ref 9,

11, 12):

hkl

hkl

hkl

ehkl

uw (d uw d 0 )/d 0

sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)

(Eq 8)

hkl

reflection; ehkl

uw and d uw are the strain and interplanar spacing of (hkl) in the direction of (u, w),

respectively; m hkl and Ehkl are Poissons ratio and

Youngs modulus in (hkl) in the coating; rx, ry,

and rz are normal stresses; sxy, syz, and sxz are

shear stresses; rx is the normal stress acting in

the x direction on a plane perpendicular to the x

axis; and syz is the shear stress on a plane normal

to the y axis (the first subscript) in the z direction

(the second subscript). Residual stresses of the

(rx ry rz)

(Eq 13)

to obtain rx, ry, and rz. The slopes and intercepts

2

of a linear function of d hkl

uw versus sin w at u

0 and u 90 are given by:

hkl

hkl

Mu0 d hkl

0 [(1 m )/E ](rx rz)

hkl

hkl

hkl

hkl

I d hkl

0 {[(1 m )/E ]rz (m /E )

(rx ry rz)}

(Eq 14)

Stresses rx, ry, and rz can be determined from

the sum of the slopes and the intercept in Eq 14.

Triaxial Stress State with Shear Stress. A

coating with a three-dimensional (triaxial) stress

state, including shear stresses, is fully described

by Eq 8. The shear stresses, sxz and syz, have a

2

sin2 w dependence. The d hkl

uw versus sin w distribution is no longer linear and has two branches

of an ellipse for w 0 and w 0. This effect

is termed w splitting, which is an indication of

the presence of shear stress.

To obtain these stress-tensor components,

three data sets (u 0, u 45, and u 90)

are obtained for both w and w. The average

strain a1 and the deviation a2 from the strains of

w splitting are determined to be (Ref 9, 11,

12):

hkl

a1 (euw

ehkl

uw)/2

hkl

hkl

[(d uw

d uw

)/2d 0hkl] 1

(rx cos2 u sxy sin2 u ry sin2 u rz)

sin2 w [(1 m hkl)/E hkl]rz (m hkl/E hkl)

(rx ry rz)

(Eq 15)

hkl

a2 (euw

ehkl

uw)/2

hkl

hkl

(d uw

d uw

)/2d 0hkl

(Eq 16)

hkl

hkl

The stress-tensor components can be calculated from the slopes of linear plots of a1 versus

sin2 w and a2 versus sin2 w. For a1 versus sin2

w, (rx rz) is obtained at u 0, (ry rz)

at u 90, and sxy at u 45, whereas rz is

evaluated from the intercept if d hkl

0 is known.

Similarly, sxz and syz are obtained when u 0

and u 90, respectively, from the slope of a2

versus sin2 w.

Stress Measurement. Modern diffractometers are fully automated and equipped with computer software for performing numerically intensive analyses. Diffractometers are capable of

measuring residual stresses efficiently and economically. Typically, the measurement can be

completed in several hours. The following examples illustrate some stress measurements using Cu K radiation on cathodic arc PVD titanium nitride coatings on a substrate with various

stress states.

Biaxial. An approximately 10 lm (400 lin.)

thick coating of highly (111) oriented titanium

nitride was deposited on AISI 304 stainless steel

at 500 C (930 F). The x-ray diffraction sin2 w

technique was applied to determine the residual

stress in the (333)/(511) reflection of the coating.

Figure 4 shows the linear distribution of d hkl

uw versus sin2 w, indicating typical biaxial stress in the

coating. As shown, the slope M 1.1933

103 nm (4.698 1011 in.) and the intercept

I d hkl

uw0 0.082123 nm (0.003233 lin.).

Youngs modulus and Poissons ratio for titanium nitride in (333)/(511) are 364 GPa (52.8

106 psi) and 0.245, respectively (Ref 13). The

calculated residual stress for titanium coating is

4248 MPa (615 ksi) in compression. For

highly (111) oriented titanium nitride film, (511)

contribution is negligible. This is addressed in

Ref 13. The (422) reflection has 2h 130,

which is not desirable.

Triaxial without Shear Stress. To exemplify a

triaxial stress distribution, without shear stresses,

triaxial stress analysis was applied to a titanium

nitride coating deposited on Inconel 718 sub2

strate at 550 C (1020 F). The d hkl

uw versus sin

w distributions were obtained from (333)/(511)

for u 0 and u 90 (Fig. 5). The slopes

Mu0 and Mu90 are 0.96782 and 1.01415

103 nm (3.8103 and 3.9927 1011

in.), respectively, and the intercept I is 0.082083

nm (0.003231 lin.). The strain-free interplanar

spacing for (333)/(511) is 0.08160 nm

(0.003213 lin.). Based on Eq 14, the stress tensor (rij) from this analysis, in units of MPa, is:

rx sxy sxz

rij syx ry syz

szx szy rz

2653 151

0

0

0

2819 150

0

0

0

815 140

equal biaxial within experimental error and that

the stress perpendicular to the coating surface is

in tension.

Triaxial with Shear Stress. A triaxial stress

distribution, including shear stresses sxy, syz, and

sxz, was studied in a cathodic arc PVD titanium

nitride coating on AM-355 stainless steel. The

2

d hkl

uw versus sin w distributions were obtained

from (333)/(511) at u 0, u 45, and u

90 using Cu K radiation. Figure 6 shows the

2

typical ellipse distribution of d hkl

uw versus sin w

at u 0. Based on analysis discussed at the

beginning of this section, the stress tensor in the

titanium nitride coating, in units of MPa, is:

rij

2699 148

1 181

110 32

1 181

2776 148 34 34

110 32

34 34 833 148

Practical Considerations. First, a W-diffractometer is preferable for conducting stress measurements. The X- and W-diffractometers are defined on the basis of the sample axis for w tilt

perpendicular or parallel to the plane of the incident and detected x-ray beam, respectively. A

W-diffractometer gives symmetric irradiated areas in w and w tilted angles at a given (hkl)

reflection line and a greater range of sin2 w (from

0 to 0.95 for a W-diffractometer and 0 to 0.5 for

an X-diffractometer).

Second, the diffractometer should be mechanically aligned and calibrated using a stress-free

standard sample (NIST SRM 660 lanthanum

hexaboride powder or SRM 640 silicon powder)

to obtain the peak position of the (hkl) reflection

line within 0.01 at Bragg angles 2h in the

range of w 0 and w 0.

Third, an appropriate x-ray wavelength should

be selected to achieve the desired sampling volume (penetration depth) for a particular set of

(hkl) planes. The penetration depth, which is defined as the distance from the surface to the depth

with 63% or 1/e of the intensity of the reflection

line, is calculated for X- and W-diffractometers,

respectively:

sX (sin2 h sin2 w)/[2l(sinh cosw)]

sW (sinh cosw)/2l

Fig. 4

flection

hkl

Linear distribution of d uw

as a function of sin2 w

of titanium nitride coating from (333)/(511) re-

hkl

Linear distribution of d uw

as a function of sin2 w

of titanium nitride coating from (333)/(511) reflection at u 0 and u 90

Fig. 5

which can be obtained from Ref 14 for various

materials and radiations.

Fourth, to achieve high accuracy in residual

stress, the measurement should only be carried

out on an (hkl) reflection line with Bragg angles

2h 130 and with sufficient intensity for peakposition determination.

Fifth, an appropriately sized collimator should

be selected. It should have an irradiated area

large enough to ensure that a statistically relevant number of grains or subgrains in the coating

are included in the measurement. The angular

resolution is essential in stress measurement. If

the spot mode of x-ray beam cannot sample a

sufficient number of grains, one should use another method.

Sixth, appropriate methods (Ref 9, 10, 15

17), such as modified Lorentzian, Gaussian,

Cauchy, Pearson VII, parabola, center of gravity,

gravity line, and cross correlation should be applied to determine the peak positions of the reflection line (hkl).

Seventh, a sufficient number of measurements

in the w and w directions should be conducted to obtain an accurately linear distribution

2

or w-splitting of d hkl

uw versus sin w. The significance of this distribution should be verified by

repeating measurements at different u angles. If

a nonlinear relation other than w-splitting is ob2

tained in the d hkl

uw versus sin w distribution, then

the coating being measured is inhomogeneous

and, therefore, the x-ray diffraction sin2 w

method is no longer applicable.

Eighth, the x-ray elastic constants of the coathkl

hkl

ing for an (hkl) plane, S hkl

and

1 m /E

hkl

hkl

S hkl

/2

(1

l

)/E

,

can

be

calculated

from

2

single-crystal compliance, according to an appropriate model (Ref 11, 12), or measured experimentally in uniaxial tension or bending tests

with a series of loads (Ref 13, 18, 19). X-ray

elastic constants and applied loads (ra) obey

these equations:

spacing measured at w 0, d hkl

uw0, can be substituted for d hkl

0 . In this case, the contribution to

the total error is less than 0.1%. For a triaxial

uw0

and d hkl

0 is included in the calculation. Consequently, a small error in d hkl

0 can lead to a relatively large error in the measured stress. To obtain acceptable stress results, d hkl

0 must be within

0.01% of its true value. The stress-free interplanar spacing d hkl

0 in the strain-free direction w*

is given by (Ref 20):

hkl

sin2 w* [S hkl

1 /(S 2 /2)]

hkl

[3 (S hkl

2 /2S 1 )rz]/(rx rz)}; u 0

{1 [(ry rz)/(rx rz)]}

when |rx ry 2rz| 2|rz|

d hkl

0

hkl

d uw0

/[1 S hkl

1 (rx ry rz)

(S hkl

2 /2)rz]

x-ray diffraction method is its capability for

measuring residual stress directly and nondestructively on a product component. Portable diffractometers are commercially available and can

be used for on-site measurement. The method

can apply to a coating with a thickness ranging

from 0.5 to 350 lm (2 105 to 1.4 102

in.). However, it is difficult to measure residual

stresses in extremely thin coatings and, in some

cases, highly textured coatings. Furthermore, xray diffraction is inapplicable to amorphous

coatings, and a large scatter in stress measurement is often encountered in coatings with large

grain size.

hkl

hkl

S hkl

2 /2 (1 m )/E

hkl

2

(1/d hkl

0 )[/ra(d uw/sin w)]

Hole-Drilling Method

uw0/2ra)

(a) Typical three-element strain-gage rosette. (b)

In-plane strain components caused by release of

residual stress through introduction of a hole. Source:

Ref 21

Fig. 7

0 ,

for a coating may not be readily available in

Fig. 6

for measuring residual stresses involves drilling

a shallow hole in the test specimen to a depth

approximately equal to the hole diameter. Typ-

hkl

Typical ellipse distribution of d uw

as a function of sin2 w of titanium nitride coating from (333)/(511) reflection at varying angles. (a) u 0. (b) u 45. (c) u 90

ical hole diameters range from 0.8 to 5.0 mm

(0.0300.200 in.). The creation of the hole redistributes the stresses in the material surrounding the hole. A specially designed three-element

strain-gage rosette, such as that shown in Fig.

7(a), measures the associated partial strain relief.

The in-plane residual stresses that originally existed at the hole location can then be calculated

from the measured strain reliefs using the

method described in ASTM E 83792 (Ref 21).

The ASTM standard also gives details of practical drilling procedures.

The partial strain relief measured by one of

the three strain gages in the rosette in Fig. 7(b)

is related to the principal in-plane residual

stresses by:

er (rmax rmin)A

(rmax rmin)B cos

2

(Eq 17)

principal residual stresses, and is the angle

from the gage axis to the maximum principal

stress direction. A and B are calibration constants, the values of which depend on the specimen material properties, the rosette geometry,

the hole diameter, and the hole depth. ASTM E

837-92 tabulates the calibration constants for the

standard rosette pattern shown in Fig. 7.

Equation 17 can be inverted to determine the

principal residual stresses from the measured

strain reliefs. The result is:

rmax, rmin [(e3 e1)/4A]

{[(e3 e1)2 (e3 e1 2e2)2]1 / 2 /4B}

b 12 arctan [(e3 e1 2e2)/(e3 e1)]

(Eq 18)

calibration coefficients a and b

D0 /D

0.30

0.31

0.32

0.33

0.34

0.35

0.36

0.37

0.38

0.39

0.40

0.41

0.42

0.43

0.44

0.45

0.46

0.47

0.48

0.49

0.50

Through-thethickness hole(a)

a

b

0.089

0.095

0.101

0.108

0.114

0.121

0.128

0.135

0.143

0.150

0.158

0.166

0.174

0.183

0.191

0.200

0.209

0.218

0.228

0.237

0.247

0.278

0.295

0.312

0.329

0.347

0.364

0.382

0.400

0.418

0.436

0.454

0.472

0.490

0.508

0.526

0.544

0.562

0.579

0.596

0.613

0.629

Blind hole(b),

depth 0.4D

a

0.111

0.118

0.126

0.134

0.142

0.150

0.158

0.166

0.174

0.182

0.190

0.199

0.208

0.217

0.226

0.236

0.246

0.255

0.265

0.275

0.285

0.288

0.305

0.322

0.340

0.358

0.376

0.394

0.412

0.430

0.448

0.466

0.484

0.503

0.521

0.540

0.558

0.576

0.594

0.612

0.630

0.648

the location of gage 1 to the direction of rmax.

The above equations are valid for a homogeneous isotropic material that is wide, when compared with the hole diameter, and thick, when

compared with the hole depth. The equations

also apply to a through-the-thickness hole in a

material in the form of a thin sheet, provided that

the sheet thickness is uniform near the hole. The

associated calibration constants have slightly

different values than those of the thick-material

case. Table 2 (from ASTM E 837-92) lists the

dimensionless calibration coefficients for both

the thin-sheet and thick-material cases. A and B

can be determined from the dimensionless coefficients using:

A (1 m)a/2E

B b/2E

(Eq 19)

the combination of coating and substrate is no

longer a homogeneous material. The calibration

coefficients provided in ASTM E 837-92 no

longer accurately apply. However, the above two

equations are still valid for coated materials, if

the coating has a uniform thickness. If the substrate is less than several times greater than the

hole depth, then it is also necessary for the substrate thickness to be uniform.

The calibration coefficients A and B for hole

drilling in a coated material differ from the standard values given in ASTM E 837-92. The actual

values depend on the elastic properties of the

coating and the substrate, the coating thickness,

and the hole diameter and depth. These coefficients can be determined by either experimental

calibrations (Ref 22) using known externally applied stresses or finite-element calculations (Ref

23).

Approximate values of A and B for thick coatings can be estimated from the values given in

ASTM E 837-92. For this purpose, a thick

coating is one that is at least 0.25 times the mean

radius of the strain-gage rosette. For the smallest

commercially available hole-drilling rosette, the

mean radius is about 1.25 mm (0.050 in.). Therefore, the minimum acceptable coating thickness

is about 0.3 mm (0.012 in.).

An approximate estimation of A and B for a

coated material is based on the observation that

the hole-drilling method is most sensitive to the

stresses closest to the specimen surface. Almost

all of the measured strain relief is due to the

stresses in the material within a depth of about

0.25 times the mean radius of the hole-drilling

rosette. Thus, a specimen coated to at least this

depth is likely to behave similarly to a homogeneous thick specimen consisting only of coating material. Thus, the A and B calibration coefficients for a thick coating are approximately

equal to the ASTM tabulated values for a homogeneous material with the elastic properties

of the coating.

The A and B coefficients for coatings that are

thinner than 0.25 times the mean radius of the

from the ASTM tabulated values. The coefficients must be determined on an individual basis,

either by experimental or computational means.

The use of the hole-drilling method with such

thin coatings is not generally recommended

because the sensitivity of the resulting strain

measurements is rather low. As a result, small

absolute errors in the strain measurements can

cause large relative errors in the computed residual stresses.

Stress Measurement. The following example

illustrates a hole-drilling measurement on a detonation-gun type of tungsten carbide-cobalt

(WC-Co) coating deposited on an AISI 1018

steel substrate. The coating was approximately

0.75 mm (0.03 in.) thick and had a macroscopically homogeneous structure. The elastic properties of the coating were Ec 172 GPa (25

106 psi) and mc 0.3.

A 062-RE hole-drilling strain-gage rosette

(5.13 mm, or 0.202 in., strain-gage mean diameter) was attached to the coated specimen. A 2.44

mm (0.096 in.) diam hole was cut in the WC-Co

coating by abrasive-jet drilling using 27 lm

(1080 lin.) alumina particles. Drilling proceeded

in four approximately equal depth increments,

up to a final depth of 0.356 mm (0.014 in.). The

strain measurements listed in Table 3 were made

after each hole-depth increment. Using the A and

B calibration coefficients from ASTM E 837-92,

adjusted for the elastic properties of the coating

material, the principal residual stresses in the

coating were found to be 260 MPa (38 ksi)

and 286 MPa (41 ksi), respectively. As

might be anticipated, the residual stresses in the

coating are approximately isotropic.

Practical considerations for the use of the

hole-drilling method are:

A high-speed drilling technique using carbide

## Lebih dari sekadar dokumen.

Temukan segala yang ditawarkan Scribd, termasuk buku dan buku audio dari penerbit-penerbit terkemuka.

Batalkan kapan saja.