r- I

SSC-345
AD-A231 909 (PART 1)

ELASTIC - PLASTIC
FRACTURE MECHANICS
A CRITICAL REVIEW

DTIC
ELECTE
FEB 12991131

BU

This document has been approved

for public release and sale; its
distribution is unlimited

SHIP STRUCTURE COMMITTEE

1990

91 2 11 163
 SHIP STRUCTURE COMMITTEE

The SHIP STRUCTURE COMMITTEE is cons!Ruted to prosecute a research program to Improve the hull struck, of ships
and other marine structures by an extension of knowledge pertaining to design, materials, and methods of constru,,on.

RADM J. D. Sipes, USCG, (Chairman) Mr. H. T. Hailler

Chief, Office of Marine Safety, Security Associate Administrator for Ship-
and Environmental Protection building and Ship Operations
U. S. Coast Guard Maritime Administration

Mr. Alexander Malakhoff Mr. Thomas W. Allen

Director, Structural Integrity Engineering Officer (N7)
Subgroup (SEA 55Y) Military Sealift Command
Naval Sea Systems Command
Dr. Donald Liu CDR Michael K. Parmelee, USCG,
Senior VI... President Secretary. Ship Structure Committee
American Bureau of Shipping U. S. Coast Guard

CONTRACTING OFFICER TECHNICAL REPRESENTATIVES

Mr. William J. Siekierka Mr. Greg 0. Wc,)ds

SEA 55Y3 SEA 55Y3
Naval Sea Systems Command Naval Sea Systems Command

SHIP STRUCTURE SUBCOMMITTEE

The SHIP STRUCTURE SUBCOMMITTEE acts for the Ship Structure Committee on technical matters by providing technical
coordination for determinating the goals and objectives of the program and by evaluating and interpreting the results in
terms of structural design, construction, and operation.

AMERICAN BUREAU OF SHIPPING NAVAL SEA SYSTEMS COMMAND

Mr. Stephen G. Arntson (Chairman) Mr. Robert A. Sielski
Mr. John F. Conlon Mr. Charles L Null
Mr. William Hanzalek Mr. W. Thomas Packard
Mr. Philip G. Rynn Mr. Allen H. Engle
MILITARY SEALIFT COMMAND U. S. COAST GUARD

Mr. Albert J. Attermeyer CAPT T. E. Thompson

Mr. Michael W. Touma CAPT Donald S. Jensen
Mr. Jeffery E. Beach CDR Mark E. Noll

MARITIME ADMINISTRATION

Mr. Frederick Seibold

Mr. Norman 0. Hammer
Mr. Chao H. Lin
Dr. Walter M. Maclean

SHIP STRUCTURE SUBCOMMITTEE LIAISON MEMBERS

U. S. COAST GUARD ACADEMY NATIONAL ACADEMY OF SCIENCES -

MARINE5BARD
LT Bruce Mustain
Mr. Alexander B. Stavovy
U. S. MERCHANT MARINE ACADEMY
NATIONAL ACADEMY OF SCIENCES -
Dr. C. B. Kim COMMITTEE ON MARINE STRUCTURES

U.S. NAVAL ACADEMY Mr. Stanley G. Stiansen

Dr. Ramswar Bhattacharyya SOCIETY OF NAVAL ARCHITECTS AND

STATE UNIVERSITY OF NEW YORK MICS COMMITTEE

MARITIME COLLEGE
Dr. William Sandberg
Dr. W. R. Porter
AMERICAN IRON AND STEEL INSTITUTE
WELDING RESEARCH COUNCIL
Mr. Alexander D. Wilson
Dr. Martin Prager
 Membwr Agencies: Address Correspondenc, to:

United States Coast Guard % Secretary, Shp Structure Committee

Naval Sea Systems Command U.S. Coast Guard (G-MTH)
k
Mwtlke Adminitation
Amercan Bureau of ShilP'ng
Ship 2100 Second Street S.W.
Washington, D.C. 20593-0001
M9,q SJM Command Structure PH: (202) 267-00 3
Committee FAX: (202)267-0025
An Interagency Advisory Committee
Dedicated to the Improvement of Marine Structures

SSC-345
December 17, 1990 SR-1321

ELASTIC-PLASTIC FRACTURE MECHANICS

A CRITICAL REVIEW

The use of fracture mechanics as a tool for structural design and

analysis has increased significantly in recent years. Fracture
theories provide relationships among fracture toughness, stress,
and flaw size and are used, for example, to establish acceptance
standards for material defects in structures. This first part of
a two part report provides a critical review of the history and
state-of-the-art of elastic-plastic fracture mechanics. Part 2
presents the results of an analytical and experimental study of
fracture in the ductile-brittle transition zone of hull steels.

Rear Admiral, U.S. Coast Guard

Chairman, Ship Structure Committee

Accession For
NTIS GRA&I
DTIC TAB
Unannounced
Justification

By
Distribution/
Availability Codes
Avail and/or
Dist Special
 Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

SSC-345 - Part 1
4. Title and Subtitle 5. Report Date
April 1990
Elastic-Plastic Fracture Mechanics - 6. Performing Organization Code
A Critical Review
8. Performing Organization Report No.
7. Author's)
T. L. Anderson SR-1321
9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)

Texas A & M Research Foundation 11. Contract or Gront No.

P. 0. Box 3578 DTCG23-88-C-20037
College Station, TX 77843 13. Type of Report and Period Covered
16S o* A cy Name and Address
~owmfgaaa "Final Report
U.S. Coast Guard
2100 Second Street, SW 14. Sponsoring Agency Code
Washington, DC 20593 G-M
15. Supplementary Notes

Sponsored by the Ship Structure Committee and its member agencies.

16. Abstruct

This document reviews the history and current state-of-the-art in

elastic-plastic fracture mechanics as applied to welded steel
structures. Fundamental concepts and underlying assumptions are
described. Standardized test methods and recent developments are
reviewed. The results of a parametric study comparing several
elastic-plastic design analyses are presented.

17. Key Words 18. Distribution Statement Available from:

Fracture Mechanics 4at'l Technical Information Service
Elastic-Plastic Deformation 3pringfield, VA 22161 or
Fracture Toughness 4arine Tech. Information Facility
Transition Region lational Maritime Research Center
_ings Point, NY 10024-1699
19. Security Clossif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 160
Form DOT F 1700.7 (8-72) Reproduction of completed page authorized
 i 0m -. .
I. lJ!+
llll ill ii~l . S-

" I = "I
Ii I ,I o N-
0

41C 0
2 I .@U I I. -

0
U
a

- 9 7I o
6c,.q II 4, 3_ '21 /eI' inch- s
z 0
0

~~~~
•~~ II ! A1 ,/=~-
- e'.. Sq"" + 0' +++ . ."++ -

I- a .i . R
*5 us. en E-- - u

em-
:: is i
•g 1 3 I 9
a1 9 l S
- I 0

HHI1"Hi
a"
I; I c;
,IOVR 1 1 O OIM
. ! ,.,.
Mai li l -""' '
0- .
U I ~ ' I '' '1 *I** I
 EXECUTIVE SUMMARY
This document reviews the history and current state of the art in elastic-
plastic fracture mechanics, as applied to welded steel structures. First, the
fundamental concepts and underlying assumptions of fracture mechanics are
described. A review of fracture mechanics test methods follows, including
standardized test methods as well as recent developments that are not yet
standard practice. Next, the various procedures for applying fracture
mechanics concepts to structures are outlined. The results of a parametric
study which compares several elastic-plastic design analyses are presented.
This review has led to a number of conclusions and recommendations.

The ductile-brittle transition region is the critical area of concern for welded
steel structures.. If the material is on the lower shelf of toughness, it is usually
too brittle for structural application; if it is on the upper shelf, fracture will
not be a significant problem in most structural steels. One difficulty with
applying fracture mechanics to welded structures in the transition region is
the lack of standardized fracture toughness test procedure for weldments. An
additional problem is that the size dependence of fracture toughness in the
transition region has not been quantified. Scatter of toughness data in the
lower transition region is reasonably well understood, but the upper
transition region introduces complexities that require further study.

The comparison of elastic-plastic fracture analyses revealed that predicted

failure stress and critical crack size are insensitive to the analysis equation.
Under linear elastic conditions, all analyses were identical. In the other
extreme of fully plastic conditions, the analyses approached similar collapse
limits. Since most elastic-plastic fracture analysis equations predict similar
results, the simplest equation seems to be most appropriate.

Current elastic-plastic fracture analyses tend to be conservative when applied

to complex welded structures. In order to improve their accuracy, a number
of issues need to be addressed: the driving force in weldments, residual stress
measurements, three-dimensional effects, gross-section yielding and crack tip
constraint.

iI
 TABLE OF CONTENTS

I. INTRODUCTFION .............................. . ....... . ..... ......................... ... . ... ................

2. FUNDAMENTAL CONCEPI............................... ........

2.1 ENERGY APPROACH TO FRACTURE MECHANICS .....................4

2.2 STRESS IN'TENSITY APPROACH TO FRACTURE MECHANICS ........... 6

2.3 ELASTIC-PLASTIC FRACTURE MECHANICS ................... 10

2.3.1 Irwin Plastic Zone Correction ....................................................... 10
2.3.2 Strip Yield Plastic Zone Correction .............................................. 13
23.3 Plastic Zone Shape .............................. . ..... 14
2.3.4 Crack Tip Opening Displacement ..................................................... 16
2.3.5 The J Contour Integral ....................... 17
2.3.6 Relationship Between J and CTOD ........................ 20
2.3.7 The Effect of Yielding on Crack Tip Stress Fields ........................ 21
2.3.8 Effect of Thickness on Crack Tip Stress Fields ................... 25

2.4 MICROMECHANISMS OF FRACTURE IN FERRITIC STEEL ........ .. 27

2.4.1 Cleavage ........................................................................................... 29
2.4.2 Microvoid Coalescence........................... .35
2.4.3 The Ductile-Brittle Transition ................... 37

3. FRACTURE TOUGHNESS TESTING ............................. 39

3.1 KIC TESTIN G ......................................................... ........................................ 41

3.2 JIc AND J-R CURVE TESTING ....................... ....44

3.3 CTOD TESTING ......................................................................................... 47

3.4 WELDM(EN-JT TESTING ...................................................................... 52

3.4.1 Specimen Design and Fabrication ....................................... 53
3.4.2 Notch Location and Orientation......................... 55
3.4.3 Fatigue Precracking....... ............................. J57
3.4.4 Post-Test Analysis ........................................................................... 59

3.5 SCATTER AND SIZE EFFECTS IN THE TRANSITION REGION............ 62

3.5.1 Scatter in the Lower Transition Region ....................................... 63
3.5.2 Large Scale Yielding ................................... 72

iii
 3.6 DYNAMIC FRACTURE TOUGHNESS AND CRACK ARREST .............. 78

3.7 COMPARISONS BETWEEN J AND CTOD AS TOUGHNESS ................... 82

3.8 CHARPY-FRACTLJRE TOUGHNESS RELATIONSHIPS ......................... 84

3.8.1 Empirical Correlations ...................... 85
3.8.2 Theoretical CVN-Fracture Toughness Relationships ............. 87

4L APPICATION TO STR T ... ................................

4.1 HISTORICAL BACKGROUND ....................... . .. .............. . ......... 90

4.1.1 Linear mastic Fracture Mechanics (LEFM) ................................ 90
4.12 The CTOD Design Curve ............................................................... 91
4.1.3 The R-6 Failure Assessment Diagram ............................................... 94
4.1.4 The EPRI J Estimation Procedure ................................................. 97
4.1.5 Recent Advances in Elastic-Plastic Analysis .................................... 100

4.2 THE REFERENCE STRESS APPROACH ......................................................... 100

4.3 THE THREE TIER APPROACH (REVISED PD6493) .................................... 102

4.3.1 Level I ..................................................................................................... 102
4.3.2 Level 2 ..................................................................................................... 104
4.3.3 Level 3 ....................................................................................................... 106
4.4 COMPARISON OF DRIVING FORCE EQUATIONS:
A PARAMETRIC STUDY ............................................ ................................. 107

4.5 PROBABILISTIC FRACTURE MECHANICS ........................ 118

4.6 TEARING INSTABILITY ANALYSIS .............................................................. 119

4.7 SIMPLIFIED ELASTIC-PLASTIC ANALYSES ................................................ 123

4.7.1 The Yield-Before-Break Criterion ...................................................... 123
4.72 Critical Tearing Modulus ..................................................................... 124

4.8 UNRESOLVED ISSUES IN ELASTIC-PLASTIC FRACTURE

AN ALYSIS ..................... ......................................................................................... .... 125
4.8.1 Driving Force in Wedments .............................................................. 126
4.8.2 Residual Stresses ...... ........................ 126
4.8.3 Three-Dimensional Effects ........................... ......... 127
4.8.4 Crack Tip Constraint ....................................................................... 127
4.8.5 Gross-Section Yielding .......................................................................... 128

5. DISCUSSIO N ....................................................................................................................... 129

5.1 FRACTURE TOUGHNESS TESTING ........................................

129

iv
 52 APPLICATION TO STRUCTURES ................................................................... 131

6. CON CLU SION S .................................................................................................................. 133

6.1 FRACTURE TOUGHNESS TESTING .................................

133

6.2 APPLICATION TO STRUCTURES ................................................................... 133

7. C .....
...............
................................................................. I....................
135
 1. INTRODUCTION

Fracture mechanics is an extremely powerful tool for structural analysis. It quantifies

the critical relationship between stress, flaw size and fracture toughness, as illustrated
in Fig. 1.1. The stress is usually fixed by the design, while the fracture toughness is a
property of the material. All materials contain flaws on some scale; nondestructive
evaluation techniques determine the size and location of such flaws. Fracture
mechanics provides the mathematical relationship between these three variables at
the moment of failure. A critical value of any one of these quantities can be
determined if the other two are known. For example, if the design stress in the
structure and the fracture toughness of the material are specified, one can determine
the critical flaw size by means of a fracture mechanics analysis. This is a departure
from the traditional mode of design that relates applied stresses to the yield strength
of the material.

jMANICS

FIG. 1.1 Fracture mechanics provides a mathematical relationship between critical

combinations of stress, flaw size and fracture toughness.

The field of fracture mechanics has developed rapidly since World War I, with
many important advances made by American researchers, such as Professors G. R.
Irwin, J. R. Rice, and P.C. Paris. Ironically, most U.S. industries have been slow to
adopt the concepts of fracture mechanics, while their counterparts in Europe and
Japan have embraced this technology with open arms. Where it has been applied,
fracture mechanics has resulted not only in increased safety, but also in enormous
economic benefits. For example, hundreds of millions of dollars have been saved in
North Sea platform construction by basing weld flaw acceptance standards on
fracture mechanics analyses (1).

U.S. industrial attempts to incorporate fracture mechanics technology have often

applied linear elastic fracture mechanics (LEFM), which is appropriate in some
applications but unsuitable for many others. With local or global plastic deformation
in the structure, LEFM can be extremely nonconservative.

Mastic-plastic fracture mechanics should be applied in situations where LEFM is

invalid. Unfortunately, this technology is not well known to most U.S. industries.
One exception is the nuclear power industry in the United States. Because of their
concern for safety, electric utilities and government regulating bodies have funded
extensive research in elastic-plastic fracture mechanics over the past 20 years. This
work has produced well established procedures in the form of design handbooks,
testing standards, and regulatory guides. In addition, numerous articles have been
published in technical journals and conference proceedings.

Although the elastic-plastic fracture technology in the U.S. is fairly well advanced,
much of it cannot be translated directly to industries other than nuclear power.
Since nuclear reactors operate at several hundred degrees above room temperature,
the steel in these structures is on the upper shelf of toughness. More conventional
welded steel structures, such as ships, bridges, and pipelines, operate at much lower
temperatures, where the material may be in the ductile-brittle transition region. In
this region, failure occurs by rapid, unstable cleavage fracture, but this so-called brittle
fracture is often preceded by significant of plastic deformation and ductile crack
growth. Thus, fracture in the transition region is elastic-plastic in nature, but the
procedures developed by the nuclear power industry are intended to aitalyze ductile
fracture on the upper shelf.

Considerable research in elastic-plastic fracture in the ductile-brittle transition region

in the United Kingdom, driven largely by the development of oil reserves in the
North Sea, has helped oil companies to build platforms both safely and
economically. The design codes and regulatory guides fof North Sea construction
contain requirements for fracture mechanics testing and analysis. Consequently, a
number of American oil companies with platforms in the North Sea are becoming
familar with elastic-plastic fracture technology.

In recent years significant technolgy transfer among fracture mechanics researchers

in the U.S., Europe, and Japan has benefited all countries involved. For example, the
analyses developed in Britain for the transition region have incorporated some of
the advances that been made by the nuclear industry in the U.S. In addition,
 3

researchers in the U.S. have begun to turn their attention to the transition region,
primarily as a result of Interactions with researchers on the other side of the Atlantic.

Thus much of the fracture mechanics technology needed by U.S. industries that
construct and use welded steel structures is in place. The problem is the availability
of relevant information to engineers in these industries. The details of fracture
mechanics testing, analysis, and application are scattered throughout the published
literature.

This review attempts to define the state of the art in elastic-plastic fracture
mechanics, as applied to welded steel structures. The advantages and shortcomings
of existing approaches are outlined, and possible future dlrectic-s are discussed.
Information from a wide variety of resources is.*kcluded. The author hopes that this
review will help to codify elastic-plastic fracture mechanics so that it will gain more
widespread acceptance in industry.

Chapter 2 summarizes some of the fundamental concepts and basic assumptions of

fracture mechanics. This chapter serves as a framework for subsequent topics; later
chapters refer back to the concepts in Chapter 2. Chapter 3 covers fracture toughness.
testing, including standardized tests methods and newer test methods, such as
fracture testing of welds. Recent research on data scatter and crack tip constraint is
also reviewed. Chapter 4 describes the application of fracture toughness data to
design, and critiques the available methods, identifying the shortcomings of existing
approaches and making recommendations for future work. Chapter 5 summarizes
the major points in the two previous chapters and gives the author's perspective on
the state of the art in elastic-plastic fracture mechanics.
4

2. FUNDAMENTAL CONCEPTS

Modern fracture mechanics traces its beginnings back Griffith (2), who in 1920 used a
simple energy balance to predict the onset of fracture in brittle materials. The
Griffith model, with some modifications, is still applied today. An alternative but
equivalent view of fracture considers the stresses and strains near the tip of the crack.
Both of these approaches are outlined below for the case of linear elastic material
behavior. This is followed by an introduction to elastic-plastic fracture mechanics
and a brief review of the micromechanisms of fracture in steels and weldments.

2.1 ENERGY APPROACH TO FRACTURE MECHANICS

Consider a plate with a crack that is subjected to an external force, as

illustrated in Fig. 2.1. The crack will grow when tLe energy available for crack
extension is greater than or equal to the work required for crack growth.
Stated another way, crack extension occurs when

driving force for fracture > material resistance.

This is essentially a restatement of the first law of thermodynamics. If the plate in

Fig. 2.1 is held at a fixed displacement, the conditions for crack advance are given by

"a > a [21]

where U is the elastic energy (per unit thickness) stored in the plate and W is the
work required to grow the crack. Irwin (3) defined the term on the left of side of this
inequality as the energy release rate, G, and the term on the right as the material
resistance, R. Figure 2.1 illustrates the energy release rate concept. If the crack
extends an increment da under fixed grip conditions, the stored energy decreases by
dU. For this incremental crack extension to occur, dU must be at least as large as dW,
the work required to fracture the material and create new surface.

If the driving force, G, is greater than the material resistance, R, the crack extension is
unstable. If G a , the crack growth may be-stable or unstable, depending on the
material and configuration. When G and R are equal, stability depends on the
second derivative of work, as discussed below.
 5

LOAD (P)

DISPLACEMENT (A)

Fig. 2.1 Definition of energy release rate for fixed grip conditions.

Figure 2.2 illustrates stable and unstable behavior for a structure with an initial crack
of length ao. The driving force is represented by a series of G versus crack length
curves for 'various load levels. Figure 2.2(a) corresponds to a material in which R
remains constant with crack growth. When the load - P1, the crack does not grow
from its initial value because G < 1. When the load is increased to P2, G = R, and
the structure is unstable because any crack growth will cause G to be greater than .
The critical value of energy release rate, Gc, is a measure of the fracture toughness of
the material. Some materials have a rising R curve, as illustrated in Fig. 2.2(b). In
this case, when the load reaches P2, the crack grows a small amount, but further
crack growth at this load is impossible, the driving force would be less than the
material resistance. The crack grows an additional increment when the load
increases to P 3. When the load reaches P4, the structure is unstable because the rate
of increase in G with crack extension exceeds the material's resistance. The
instability occurs at the point of tangency between the driving force and the R curve.
Thus the conditions for stable crack growth are as follows:

G=R [2.2a]
and
dG dR (2.2 b]
M -bda
6

G,R G,R
P4

Z
CRACK sMACK SIZE
a) FlatRcurve. b) RisingR curve.

FIG. 2.2 Schematic driving force and resistance curves that Illustrate stable and
unstable behavior.

2.2 STRESS INTENSITY APPROACH TO FRACTURE MECHANICS

In 1939, Westergaard (4) performed an elastic stress analysis of a continuum with a
sharp crack. Irwin (5) and Williams (6) later applied this analysis to engineering
problems. The Westergaard-Irwin-Williams solution for the stresses near the crack
can be written as
K
Oij = 7 fi1() + other terms [23]

where aij is the stress tensor, K is the stress intensity factor, r is the radial distance
from the crack tip, and fij is a dimensionless function of e, the angle from the crack
plane. Both r and 0 are defined in Fig. 2.3. The higher order terms, which depend on
the configuration of the cracked body, are negligible near the crack tip.
Figure 2.4 shows the variation of stress ahead of the crack tip. The Westergaard
solution implies that stresses approach infinity as r approaches zero. Real materials,
however, deform plastically at finite stress levels (see Section 2.3). The stress
intensity factor, K, is a proportionality constant that measures the severity of the
stresses at the crack tip. If K doubles, for example, the stresses at a given r and e also
double.
 7

Crack0

FIG. 2.3 Definition of xy and polar coordinates at the tip of a crack.

ij
Increasing K

r
FIG. 2.4 Schematic stress distributions ahead of a crack In an elastic material.

Irwin and Williams identified three ways in which a cracked body could be loaded:
opening (Mode I), sliding (Mode H) and out-of-plane shear (Mode HI). In each case,
the resulting stresses ahead of the crack tip can be described by Eq. [2.3] The stress
intensity factor typically has a subscript to identify the mode of loading- KI, KII or
KM. In most metals, the Mode I component of loading is the most important
because it controls fracture. This document will consider only Mode I loading. (See
Fig. 2.1 for an example of Mode I opening of a cracked body.)

The most important ramification of the Westergaard analysis is that all nine
components of the stress tensor, as well as all components of the strain tensor, are
completely defined by a single constant, K. That is, if K is known for a given
situation, all stresses and strains near the crack tip can be computed from Eq. [2.3].
8

Consider a small element of material at the crack tip. It is reasonable to assume that
this material element fails when it experiences a critical combination of stress and
strain. Thus, this material element must fail at a critical K value. This philosophy
led to the definition of a critical stress intensity, KiC, at the onset of crack extension
(7.
Although the stress intensity factor contains much information about the crack tip
environment, it would be of little practical value if it were not possible to relate K to
remote loads and displacements. Fortunately, K is related to global behavior. For
example, if an infinitely wide plate with a crack of length 2a is loaded to a remote
stress a (Fig 2.5), the stress intensity factor is given by

KI = c ' ?(2.4]
The subscript is on K in this case because the configuration in Fig. 2.5 produces pure
Mode I loading. Note that KI, and thus the crack tip stresses, can be increased either
by increasing the remote stress or by increasing the crack length. In addition,

FIG. 2.5 Through thickness crack In a Infinitely wide plate subjected to a remote stress.
 9

by setting KI to the critical value for the material, it is possible to relate stress, fracture
toughness (KIC), and critical crack size:

KIC W ( M2.5]

Equation [2.4] applies only to a through-thickness crack in an infinite plate; i.e., a

plate whose width is >> 2a. For other configurations, Kj can be written in the
following form.

KI MYd [2.]

where Y is a dimensionless geometry correction factor. In this case, a is a

characteristic stress and a is a characteristic crack dimension. Stress intensity
equations can be expressed in a wide variety of ways but can always be reduced to the
form of Eq. [2.6]. A number of handbooks of stress intensity factors for a wide range
of configurations have been published over the last 20 years (8-10).

Given the stress intensity solution of a particular structure, it is possible to predict

the fracture behavior of this structure from a small-scale laboratory specimen, as long
as both the structure and the specimen behave in a linear elastic manner. Although
the specimen and structure may be of vastly different size and shape, the crack tip
conditions are identical as long as both configurations are loaded to the same KI
value. This situation is analogous to design against yielding. That is, a simple
tensile test can predict the onset of yielding in a complex structure, given an
appropriate stress analysis of the structure.

The previous section describes the energy approach to fracture mechanics, where the
energy release rate, G, is compared to the material resistance, . This section outlines
the stress intensity approach, where the applied mode I stress intensity factor, KI, is
compared to the critical value for fracture initiation, KIc. Irwin showed that the
energy and stress intensity approaches are equivalent by demonstrating a unique
relationship between K and G for linear elastic materials:
K2 [27
G W [7-71

where E' = E (Young's modulus) for plane strain and E' - E/(1 - v2 ); v is Poisson's
ratio. The stress intensity approach to linear elastic fracture mechanics is more
common than the energy approach. Fracture toughness data are seldom reported in
terms of critical G values. The energy approach, however, does have merit in some
situations. For example, stability analyses, such as that presented in the previous
section, are more straightforward with an energy parameter such as G. In addition, G
10

is more convenient than K in mixed mode problems because G components are

additive:
Gtotal Gi + Gil + Giii

but
Ktota p Kj + I¢41 + Kill
All of the above analyses are strictly valid only for isotropic materials that behave in
a perfectly linear elastic manner. When there is a small amount of plastic
deformation at the tip of the crack, linear elastic fracture mechanics (LEFM) gives a
good approximation of actual material behavior. Eventually, however, the theory
breaks down. The following section describes the limitations of LEFM and the
existing methods to account for crack tip plasticity.
2.3 ELASTIC-PLASTIC FRACTURE MECHANICS
Fracture mechanics approaches to crack tip plasticity fall into two main categories: 1)
simple corrections to LEFM theory, and 2) fracture parameters which allow for
nonlinear material behavior. Irwin (11) proposed a simple plastic zone correction to
the stress intensity factor. An alternative plastic zone correction was developed by
Dugdale (12) and Barenblatt (13). The first truly elastic-plastic fracture parameter, the
crack tip opening displacement (CTOD), was proposed by Wells (14) in 1961. Several
years later, Rice (15) developed the J contour integral, a parameter that approximates
elastic-plastic deformation with a nonlinear elastic material assumption. The J
integral can be viewed as both an energy parameter and a stress intensity-like
quantity. In addition, J is uniquely related to CTOD under certain conditions.

2.3.1 Irwin Plastic Zone Correction

Equation [2.3J predicts infinite stresses at the crack tip, but no material is capable of
withstanding infinite stress. In metals, a plastic zone forms at the tip of the crack,
thereby limiting the maximum stress in the material.

Consider the stresses normal to the crack plane (0-0) in mode I loading. Equation
[2.31 reduces to
j[2.8]
- nKI

For plane stress conditions (rzz-0), the material yields when ayy = cys, the uniaxial
yield strength of the material. Assuming a nonhardening material, the
 11

Gyy

FIG. 2.6 First order estimate of plastic zone size for plane stress conditions.

stress field is truncated as indicated in Fig. 2.6. Substituting ays into Eq. [2.8] gives a
first order estimate of the plastic zone size:

r* - I-KIN [2.9]
2x n

However, the schematic in Fig. 2.6 is not totally correct. The shaded area represents
load that would be carried by a purely elastic material, but cannot be carried by an
elastic-plastic material. In order for the structure to remain at equilibrium, the load
represented by the shaded area must be redistributed, as shown in Fig. 2.7. Note that
this increases the plastic zone size. Irwin (11) estimated the overall plastic zone to be
2.r.

An important effect of the redistribution in crack tip stresses is that the effective
stress intensity-increases. Outside of the plastic zone, Eq. [2.8] holds only if Ki is larger
than the pure LEFM value. Irwin (11) modeled this increase in the effective KI by
treating the crack as if it were slightly longer than its true physical length:

f MYa 2.o10]
The effective crack length is obtained by adding a plastic zone correction to the
physical crack length.
12

Oayr

r.
rTr

FIG. 2.7 Irwin plastic zone correction for plane stress conditions.

aeff = a + ry [2.11]

The geometry correction factor, Y, in Eq. [2.10] must take account of the longer
effective crack. For plane stress conditions, ry = r*; thus the plastic zone correction is
given by

ry = u-1 (Keff'~
r o!l 21 a
[2.12a]

For plane strain conditions, yielding is restrained and the plastic zone correction is
smaller:

M (Keff)
ry= [2.12b]

Calculation of Keff is an iterative process: ry is computed from Kerf, a new Keff is

computed from the ry, a new ry is then computed from this Kerr, and so on. The
process usually requires only two or three iterations to converge on an accurate
estimate of Kerr. Some simple configurations have a closed form solutions. For
example, Keff for a through crack in an infinite plate (Fig. 2.5) is given by
 13

Keff
E'1(
a c'I
J
F1"2 rc,0)-S
21/2
[2-131

2.3.2 Strip Yield Plastic Zone Correction

In 1960, Dugdale (12) modeled plasticity ahead of a notch in a thin plate as a thin strip
of yielded material. Barenblatt (13) independently developed a similar model. The
strip yield model is illustrated in Fig. 2.8. The assumptions of this model include
plane stress deformation and a nonhardening material. The initial analyses
considered only a through crack in an infinite plate.

As indicated in Fig. 2.8, a plastic zone of length p is produced at each crack tip
according to the model. Since the material cannot support stresses higher than yield,
the model treats the problem as a crack of length 2a + 2p, with closure stresses equal
to yield at each end (Fig. 2.8b). The plastic zone size, relative to the physical crack
size, is given by

IPlastic zone
a) Strip yield plastic zones.

b) Assumed closure stresses In the strip yield zone.

FIG. 2.8 The strip yield model for a through crack of length 2a.
14

a -Cos'0 [2-141

Ifaeff is taken as (a + p), the effective KI is given by

However, this equation leads to overestimates of Keff because closure stresses cause
the true effective crack length to be somewhat less than (a + p). Burdenkin and
Stone analyzed the strip yield model further and derived a more appropriate
relationship for Ke.ff

Ieff - 0.9 4q'aln sec 2 J] [2.161

Figure Z9 shows a comparison between a pure LEFM analysis, the Irwin plastic zone
correction for plane stress (Eq [2.13]), and the Burdenkin and Stone strip yield
equation. The effective stress intensity, nondimensionalized by crsy4x;, is plotted
against the normalized stress. The LEFM analysis predicts a linear relationship
between K and stress. Both the Irwin and strip yield corrections deviate significantly
from LEFM theory at stresses greater than 0.5 oys. The two plasticity corrections agree
with each other up to approximately 0.85 rys. According to the strip yield model,
Kef approaches infinity as stress approaches the yield strength. Since this model
assumes a nonhardening material, it is not surprising that it predicts that the plate
cannot withstand stresses greater than yield. Obviously, the strip yield model is
conservative for materials that strain harden. More sophisticated analyses that
account for strain hardening are described in Chapter 4.

2.3.3 PlastIc Zone Shape

Figures 2.6 and 2.7 idealize the plastic zone ahead at the tip of a crack tip as circular.
The previous section modeled the plastic zone as a narrow-strip ahead of the crack.
Neither of these viewpoints is correct in the case of metals. (Some plastics, however,
exhibit crack tip damage zones which closely resemble Fig. 2.8a (16).) Figure 2.10
shows the computed plastic zone shape at several stress levels (17). For metals the
plastic zone typically has a "butterfly" shape.
 16
aysfRi?
1.2

0.8

0.4

0
0 0.2 0.4 0.6 0.8
ayys
FIG. Z9 Nondimenalonal plot of K,1, versus stress for a through crack Inan Infinite plate.

0. 5 a
z 0.75-

FIG. 2.10 Plastic mon shape computed by Tuba (17).

16
2.3.4 Crack Tip Opening Displacement
In the late 1950s, Wells attempted to apply Irwin's stress intensity concept to
measure the fracture toughness of a series of medium strength structural steels. He
found that these materials exhibited a high degree of plastic deformation prior to
fracture. This was good news from a design engineer's standpoint because it
indicated high toughness in these steels. However, significant plasticity was bad
news for theoreticians because it meant that linear elastic fracture mechanics was not
applicable to typical structural steels.
Wells noticed that the crack faces moved apart prior to fracture in test specimens
which exhibited a high degree of plasticity. In 1961, he proposed using the
displacement of the crack faces as a measure of fracture toughness (14). The crack
opening displacement (COD) was defined at the original crack tip, as illustrated in
Fig. 2.11. Today Wells' COD is called the crack tip opening displacement (CTOD) to
distinguish it from other displacement measurements along the crack flanks.

a) Sharp crack b) Blunted crack

FIG. 2.11 Crack tip opening displacement (CTOD), defined at the original crack tip.

Wells (14) added credibility to CTOD as a fracture mechanics parameter by

demonstrating a relationship between KI and CTOD for plane stress small scale
yielding conditions:

8 = K12[217]
oys E

where 8 is the CLOD. Of course the intent was to apply CTOD beyond the limits of
LEFM, but it was encouraging to learn that CTOD was consistent with the stress
intensity approach in the limit of linear elastic behavior.

Burdenkin and Stone (18) used the strip yield model to estimate CTOD in an infinite
plate with a through crack:

8 a a Inn [2-18]
xE 2 )
 17

Series expansion of the In sec term yields

8 "a [I Li
0 [Zl9a]
zE 2'(T" ,.f + 1L (C
O +...i~ [2.I9b]

Thus when a/ays is small, the Burdekin and Stone equation reduces to the Wells
relationship for small scale yielding (Eq. [2.17).
2.3.5 The J Contour Integral
Plasticity theory is more complex than the theory of elasticity. When a material
deforms elastically, it is possible to deduce the current stresses from the current
strains, and vice versa. However, material response to plastic deformation is history
dependent. Since a set of plastic strains does not uniquely define the stresses in the
material, a closed-form solution to the crack tip stress field, similar to the
Westergaard solution for linear elastic materials, is not possible for an elastic-plastic
material.
In certain cases, however, approximate stress analyses of elastic-plastic materials are
possible by assuming a nonlinear elastic material response, as illustrated in Fig. 2.12.
On loading, the stress-strain behavior of the nonlinear elastic material is identical to
that of an elastic-plastic material. Only on unloading do the stress-strain curves
differ. The similarity between stress-strain curves of elastic-plastic and nonlinear
elastic materials does not necessarily carry over to three-dimensional loading
situations, but there are many situations where the nonlinear elastic model gives a
good approximation of real material behavior.

Rice (15) utilized a nonlinear elastic material assumption to derive the J contour
integral, a parameter that describes the conditions near the tip of a crack. Prior to
publishing his results, Rice discovered that Eshelby (19) had earlier published a series
of conservation integrals, one of which was identical to Rice's J, but Eshelby did not
apply his results to crack problems.
The J integral is obtained by integrating the following expression along an arbitrary
path around the tip of a crack (Fig. 2.13):
i8

Nonlinear elastic
material
STRESS

icpastic
material

-1 STRAIN
12 CmprionOf
FIQ*IFI-nonlin ar,.~ o nonlinear elastic
fl0~ n arelastic and elastic plao
materila will Unoa
along the t te,
s a e ia8l 1bab ihwavoro
4oae

FIG. -2.13 'Arbitrary~ contu

OUond a crack tip.r
 19

j= J(Wdy - T - ds) [2.20

r
where r is the path of integration, W is the strain energy density, T is the traction
vector, u is the displacement vector, and ds is an increment along r. The coordinates
x and y are as defined in Fig. 2.3. For nonlinear elastic materials, Rice showed that
the value of J is independent of the integration path as long as the contour encloses
the crack tip, as illustrated in Fig. 2.13.

The J integral can also be defined as a nonlinear elastic energy release rate:

J -[2.211

where U is the strain energy per unit thickness, as in Eq. [2.1], and A is the
displacement in the loading direction. Figure 2.14 illustrates the energy release rate
definition of J. For a linear elastic material, it is obvious that J - G. Thus the J
integral reduces to the LEFM energy release rate under small scale yielding
conditions.

LOAD

DISPLACEMENT
FIG. 2.14 Energy release rate definition of the J Integral.
20

In addition to being an energy release rate, J can be viewed as a stress intensity

parameter. Hutchinson (20) and Rice and Rosencren(21) independently derived a
crack tip stress and strain solution for nonlinear elastic materials, which became
known as the HRR singularity. The HRR solution was derived for materials with a
power law stress-strain relationship:

a M a_ [2221

where a0 is a reference value of stress (usually aye), eo = co/E, and a and n are
material constants. The HRR stress and strain fields near the tip of the crack are
given by

(ij
O ; E r )" (n+l) gij(n,O) [2.23]

and

2i-
ac0 Lao2in
E I r ) n/(n+1) hij(n,O) [124]

where In is a dimensionless constant that depends on n, and gij and hij are
dimensionless functions of n and e. These parameters also depend on stress state
(plane stress v. plane strain). For a linear elastic material, n = 1 and stress varies as
J/r (=G/r or K/NO, which is consistent with the Westergaard solution (Eq. [2.3]). As
with the stress intensity factor in linear elastic materials, J completely characterizes
the crack tip stresses and strains in nonlinear elastic materials. Thus J is an excellent
fracture mechanics parameter for elastic-plastic materials, as long as the nonlinear
elastic assumption gives a reasonable approximation of real material behavior. The
limitations of the J integral approach are discussed in Section 2.3.7

2.3.6 Relationship Between J and CTOD

Under small scale yielding conditions, J - G, and CTOD and KI are related (for plane
stress) by Eq. [2.17]. Thus, for linear elastic, plane stress conditions, J and CTOD are
related as follows:

J MOYS 8 [2.251
 21

Shih (22) used the HRR singularity to derive a relationship between J and CTOD for
elastic-plastic conditions:

C[FO6a

where dn is a dimensionless constant that depends on material flow properties (E0 , n,

a) and stress state (plane stress versus plane strain). Shih defined CTOD at the
intersection of.90* vertices drawn from the crack tip, as illustrated in Fig 2.15. Typical
structural steels blunt in a roughly semicircular shape (23). Thus the 90* intercept
definition is approximately equal to the CTOD defined at the original crack tip.

Medium strength structural steels typically have an n value in the range of 8 to 12.
The dn value for such materials in plane strain is approximately 0.5. In the limit of
no strain hardening, where n becomes large, dn approaches 1.0 for plane stress. This
agrees with Eq. [2.251, based on the Wells (14) and Burdekin and Stone (18) analyses.
Both of these analyses assumed a nonhardening material.

FIG. 2.15 The 900 Intercept definition of crack tip opening displacement.

2.3.7 The Effect of Yielding on Crack Tip Stress Fields

Figure 2.16 is a series of schematics showing the effect of plastic deformation on crack
tip conditions. Part (a) illustrates small scale yielding behavior. The region
moderately close to the crack tip where the crack tip stress and strain fields are
described by Eq. [2.3] is called the K controlled region because the stress intensity
factor completely defines the stresses and strains. Somewhat closer to the crack tip,
in the plastic zone, is the J controlled region where Eqs. [2231 and [2.241 apply. The
small area very close to the crack tip is the large strain region where the HRR
solution is no longer valid. A log-log plot of stress versus r has a slope of -1/(n+1)
22

in the J controlled region and a slope of -1/2 in the K controlled region. Outside the
K controlled region, the stress field is influenced by the higher order terms in the
Westergaard solution. In the case of small scale yielding, one can characterize the
severity of the crack tip conditions with either K or J. Since J-G in this case, K and J
are directly related:

K2

As the plastic zone grows with additional loading, the K controlled region
disappears, as .illustrated in Fig. 2.16(b), but there is still a region in which the HRR
solution is valid. Thus the crack tip conditions can be characterized with J, but K is
no longer defined. It is possible to define an effective K based on the J value and Eq.

Kefu= 4TJF [2.28]

This effective K does not describe the actual stress field, but the stress field that would
be present under small scale yielding, as indicated by the dashed line in Fig. 2.16(b).

Note that the large strain region is much bigger in Fig. 2.16(b). The size of this region
is approximately 28 (24). The HRR solution is not valid in the large strain region
because of crack blunting. Equation [2.23] predicts that stresses approach infinity as r
approaches zero. However, the tip of the blunted crack is a free surface which cannot
support a triaxial stress state. Material in the large strain region has partially
unloaded; thus the nonlinear elastic material assumption leads to erroneous
predictions. The stresses in the large strain region are lower than predicted by the
HRR singularity.

Under large scale yielding conditions (Fig. 2.16(c)), the large strain region engulfs the
J controlled region, and neither J nor K characterize crack tip conditions. Because of a
loss in triaxiality, the stresses are lower than predicted by both J theory and K theory.

The sizes of the J and K controlled regions depend on the size and geometry of the
cracked structure. Because J and K controlled conditions exist longer in larger
structures, fracture mechanics test standards usually contain specimen size
requirements. For example, the American Society for Testing and Materials (ASTM)
standard for KIC testing (25) includes the following requirement.

B, b > 2.5 (IC [2.29]

 23

a)small scale yleldl LO a

JYC ontr~edLOG r

b)~~ ~ ~ O ~l ~ ~ O rlsi-aebhvo

field.E
24
where B is the specimen thickmess and b is the uncracked ligament length, as defined
in ig. 2.17. By comparing this requirement with Irwin's estimate of the plane strain
plastic zone size (Eq. [2.12b]) one concludes that the plastic zone size must be no more
than - 1/50 of characteristic specimen dimensions. The ASTM JIC testing standard
(26) also has a size requirement:
B, b 225 J [M1
OY

where ay is the flow stress, defined as the average between the yield and tensile
strengths. For a material with dn = 0.5 and cy = 1.ays, this requirement implies that
the relevant-specmen dimensions should be at least 45 times larger than the CTOD.
Equation [2.30] is much less stringent than Eq. (2.29] because J controlled conditions
exist much longer than K controlled conditions, as illustrated in Fig 2.16.

In many practical situations, it is impossible to satisfy Eq. [2.29] with laboratory

specimens, but a large structure made from the same material may behave in a linear
elastic manner at the design stress. This is illustrated schematically in Fig. 2.18,
where a large structure and a small laboratory specimen are loaded to the same
applied J. The plastic zone is too large for there to be a K controlled region in the
small specimen, but the plastic zone is small compared to structural dimensions.

FIG. 2.17 Critical dimensions of a fracture mechanics specimen.

 25

Thus the structure has both K and J controlled regions, but the specimen has only a J
controlled region. Near the crack tip, the stress fields of the two configurations are
identical. Further from the crack tip, however, the stress fields are different because
the stresses in the small specimen are influenced by its finite size. In situations such
as this, it is possible to measure a JIC value on the small specimen, convert this value
to an equivalent KiC using Eq. [2.28], and apply this KIC to the structure. Linear
elastic fracture mechanics design analyses are much simpler that elastic-plastic
analyses.

The issues raised in this section are discussed further in Chapters 3 and 4.

2.3.8 Effect of Thickness on Crack Tip Stress Fields

The discussion so far has focused on the x-y plane of a cracked structure because
-xisting fracture mechanics analyses are two-dimensional. Equations [2.31, [2.231, and
[2.241 apply only to purely plane stress or plane strain conditions. Real structures and
specimens, however, contain regions that are neither plane stress nor plane strain.

Plane strain conditions at the tip of a crack produce a triaxial stress state, while plane
stress conditions are biaxial by definition. Because of the different stress states, the
plastic zone at a given K value is larger in plane stress, as indicated by Eqs. [2.12a] and
[2.12b]. Both the stress normal to the crack plane, oyy, and the stress parallel to the
crack propagation, oxx, are lower in plane stress than in plane strain. Since the local
stress plays a key role in the failure of most materials, the measured fracture
toughness is usually higher in plane stress.

The ASTM standards for KIc and JIc testing (25,26) have requirements on thickness
as well as ligament length (Eqs. [2.29] and [2.301). The thickness requirements are
needed for the following reasons:

" Both J and K analyses assume two-dimensional deformation: either plane

stress or plane strain. Thus fracture mechanics tests should approximate
one of these two situations.

" Since plane strain is the more severe of the two cases, test specimens
should be sufficiently thick to ensure predominantly plane strain
conditions.

Wellman et al. (27) demonstrated the effect of thickness with three-dimensional

finite element analysis of a fracture mechanics specimen of A36 steel Figure 2.19 is a
plot of ayy along the crack front for two thicknesses and three CTOD values,
corresponding to 0 a 0 and a constant r value near the crack tip. The normalized
thickness refers to the relative position along the crack front; Z/T - 0 corresponds to
26

SMALL
TETSERE
~ LO
FIG. .13 sresa ield In asmal testrllemnd
chema~c alresrcuelae
to the
aplgionJ
ame
 27

the center of the specimen and Z/T -1.0 corresponds to an outer edge. Note that
near the center of the specimen the curves are flat, indicating plane strain conditions.
Near the edge of the specimen, stress decreases rapidly as the plane stress limit is
approached. The size of the plane strain region decreases with increasing CTOD.
The thicker specimen has a larger relative plane strain region at a constant CTOD, an
effect is seen more dearly in FIg. 2.20, which is a plot of the relative size of the plane
strain region versus CTOD/thickness. For this plot, the boundary of the plane strain
region was defined arbitrarily as the point where the stress fell to 90 percent of the
center plane value. The KIC thickness requirement for this material is superimposed
for comparison. According to Fig. 2.20, approximately 85 percent of the crack front is
in plane strain when Eq. [2.291 is satisfied. The thickness requirement of Eq. [2.301 for
this material corresponds to CrOD/B = 0.022, which is well off the scale of Fig. 2.20.
Thus the crack front of a specimen that just satisfies the JIc standard has less than 50
percent plane strain along the crack front. Whether or not this is sufficient to
measure a fracture toughness value indicative of pure plane strain conditions
depends on the micromechanism of fracture.

2.4 MICROMECHANISMS OF FRACTURE IN FERRITIC STEEL

Fracture in steel parent material and welds usually occurs by one of three
mechanisms:

1. Transgranular cleavage

2. Microvoid coalescence

3. Intergranular fracture

Cleavage is rapid, unstable fracture usually associated with brittle materials, while
microvoid coalesence (or ductile tearing) can occur in a slow, stable manner.
lntergranular cracking can be either ductile or brittle. It is usually associated with a
corrosive environment, grain boundary segregation, or both. In the absence of
adverse environmental conditions and detrimental heat treatments such as temper
embrittlement, fracture in ferritic materials nearly always occurs by mechanisms (1)
and (2). Consequently, this section focuses on cleavage and microvoid coalescence.

Cleavage occurs when the local stress is sufficient to propagate a crack nucleus that
forms from a microstructural feature such as a carbide or inclusion. For ductile
tearing, a critical strain must be reached for the coalescence of voids that form
around second phase particles. The fracture toughness will necessarily differ for the
different fracture mechanisms (28).

Figures 2.21 and 2.22 are scanning electron microscope (SEM) fractographs that
compare the appearance of the two fracture mechanisms (23,29). Cleavage produces a
relatively flat, faceted surface because the fracture propagates along specific
28

4.C
CnnwwM CMu - U63
as 127 9 2.4 -m MpsbA
8L 31.3 e 63. -o vpskn

3A

900

CD"VVLNCA
w STEEL*M

1.0 -e * 2;4Wusped 12.

U
06 Bm4. nU; W=64nu

900

70 I
500

0 0.002 0.004
CTODID

FIG. 2.20 Percent of the crack front where nonminally plane strain condition exist as a
function of thickness and CTOD. Based on Wellman, et. at. resuits (27).
 29

crystallographic planes ((100) planes in the case of BCC iron). The microvoid
coalescence fracture surface has a fibrous appearance because of the growth and
coalescence of voids formed at inclusions. Some of the inclusions that nulceated
voids are visible in Fig. 2.22. These two fracture mechanisms are described in more
detail below.

2.4.1 Cleavage
Cleavage on the atomic level involves breaking bonds along the cleavage plane.
Thus, the cohesive strength of the material must be exceeded locally. However, a
macroscopic crack is only capable of producing stresses 3 to 5 times the uniaxial yield
strength (24), well below the stress required to break bonds. Ferritic materials
contain microstructural features capable of nucleating sharp microcracks that
provide sufficient local stress elevation to cause failure.

This mechanism of cleavage nucleation is illustrated in Fig. 2.23. The macroscopic

crack provides a local stress and strain concentration. A second phase particle, such
as a carbide or inclusion, cracks because of the plastic strain in the surrounding
matrix. At this point the microcrack can be treated as a Griffith (2) crack. If the stress
ahead of the macroscopic crack is sufficient, the microcrack propagates into the ferrite
matrix, causing failure by cleavage. For example, if the particle is spherical and it
produces a penny-shaped crack, the fracture stress is given by

Of Ey [/231

where yp is the plastic work required to create a unit area of fracture surface in the
ferrite and C0 is the particle diameter.

The nat.ure of the microstructural feature that nucleates cleavage depends on the
alloy and heat treatment. In mild steels, cleavage usually initiates at grain boundary
carbides (28,30,31). In quenched and tempered alloy steels, the critical feature is
usually either a spherical carbide or an inclusion (28). Various models (28,30-33)
have been developed to explain the relationship between cleavage fracture stress and
microstructure. Most of these models resulted in expressions similar to Eq. [231];
some models differ from Eq. [2.31] because they account for dislocation interactions
with the particles (31,32).

Susceptibility to cleavage fracture is enhanced by almost any factor that increases the
yield strength, such as low temperature, a triaxial stress state, radiation damage, high
strain rate, and strain aging. Grain size refinement increases the yield strength but
also increases Of. There are a number of reasons for the grain size effect. In mild
30

FIG. 2.21 SEM fractagraph of cleavage fracture Inpure Iron (23).

FIG. 2.22 SEM fracographs of microvold coalescence In hot rolled 1040 stee (29).
 31

of

FIG. 2.23 Nucleation of a sharp microcrack ahead of a macroscopic crack.

steels, a decrease in grain size implies an increase in grain boundary area, which
leads to smaller grain boundary carbides and an increase in of. In fine grained steels,
the critical event may be propagation of the microcrack across the first grain
boundary it encounters. In such cases the Griffith model implies the following
expression for fracture stress:

Of = ( E~ybj/ 2 [2.321

where 'Ygb is the plastic work per unit area required to propagate into the adjoining
grain, and d is the grain diameter. Since there tends to be a high degree of mismatch
between grains in a polycrystalline material, Igb > 1p. Equation [2.32] assumes an
equiaxed grain structure. For martensitic and bainitic microstructures, Dolby and
Knott (33) derived a modified expression for of based on the packet diameter.

Figure 2.24 summarizes the the cleavage fracture process. Although this schematic
shows a spherical particle in the center of a ferrite grain, the concepts represented
apply to all types of cleavage nucleation. Part (a) illustrates crack nucleation in the
particle. The particle cracks because it is unable to strain plastically with the matrix.
If the stress is high enough to satisfy Eq. [2.31], the microcrack propagates into the
surrounding matrix, as illustrated in Fig. 224(b). When the propagating crack
32

reaches the grain boundary, it must change orientation to align itself with the nearest
cleavage plane of the next grain (Fig. 224c), requiring additional work, as discussed
above.

In some cases cleavage nucleates, but total fracture of the specimen or structure will
not occur. Figure 2.25 illustrates three examples of unsuccessful cleavage events.
Part (a) shows a microcrack that has arrested at the particle/matrix interface. The
particle cracks due to strain in the matrix, but the crack is unable to propagate because
the applied stress is less than the required fracture stress. This microcrack does not
re-initiate because subsequent deformation and dislocation motion in the matrix
causes the crack to blunt. Microcracks must remain sharp in order for the stress on
the atomic level to exceed the cohesive strength of the material. If a microcrack in a
particle propagates into the ferrite matrix, it may arrest at the grain boundary, as
illustrated in Fig. 2.25(b). This corresponds to a case where Eq. [2.32] governs
cleavage. Even if a crack successfully propagates into the surrounding grains, it may
still arrest if there is a steep stress gradient ahead of the macroscopic crack (Fig. 2.25c).
This tends to occur at low applied KI values. Locally, the stress is sufficient to satisfy
Eqs. [2.31] and [2.32] but the stress decays rapidly away from the macroscopic crack and
eventually can no longer satisfy the Griffith energy criterion.

The phenomena illustrated in Fig. 2.25 have been observed experimentally.

Gerberich (34) monitored fracture toughness tests with acoustic emission and
observed many micro-cleavage events before final fracture. Lin et al. (35) provided
metallographic evidence of cracked carbides and crack arrest at grain boundaries in a
1008 spherodized steel. Irwin (36) observed numerous cleavage initiation sites on
the fracture surfaces of notched round bars which were tested at very high strain
rates. The dynamic loading caused cleavage nucleation at very low KI values. These
early cleavage events arrested, apparently because of the steep stress gradient. Final
failure of each specimen occurred when the applied KI was sufficient for a crack to
propagate through the specimen.

Cleavage fracture is a weakest link phenomenon. A specimen or structure needs

only one critical ndcrostructural feature for catastrophic failure to occur. The local
fracture stress depends on the largest or most favorably oriented particle that occurs
in the material near the tip of a macroscopic crack. A finite amount of material
must be sampled in order to find a critical particle. Ritchie et. al. (37) were among the
first to recognize this when they proposed a simple model for cleavage. Their model
states that cleavage will occur when the critical fracture stress, of, is exceeded over a
critical distance, x, ahead of the crack tip. They assumed that of and xc were single
valued material constants. Curry and Knott (38) used a statistical argument to
develop a model in which a critical sample volume was required in order to cause
failure. Later, Curry (39) demonstrated that their statistical interpretation of cleavage
was essentially equivalent to the Ritchie et al. model. Recently, a number of more
sophisticated statistical models for cleavage fracture have been developed(35,40-44).
These models predict the effect of microstructure on fracture toughness. In addition,
 33

a) Particle cracks.

b) Microcrack 'propagates 'Into the ferrite matrix.

Q) Cleavage crack propagates Into adjoining grains.

FIG. 2.24 Schematic drawings of the cleavage fracture process in foritlc steals.
34

a) Arrest at the partice/matrix Interface, followed by crack blunting.

b) Arrest at a ferrite grain boundary.

e) Arrest due to a steep stress gradient

FIG. 2.25 Examples of unsuccessful cleavage events.
 35

the statistical models quantify the scatter in fracture toughness data, which is a direct
result of the weakest link nature of cleavage. This scatter is particularly severe in the
ductile-brittle transition of steels. Some of the methods for analyzing scatter are
described in Section 3.5.

2.4.2 Microvold Coalescence

In ferritic steels, as the temperature increases and-the flow stress decreases it becomes
more difficult to produce high enough stresses to initiate cleavage. When conditions
for cleavage are unfavorable a ductile fracture mechanism, microvoid coalescence,
operates. This is the dominant fracture mechanism of FCC alloys, even at very low
temperatures. The typical microstructural changes which occur during initiation
and growth of a fibrous crack are (28):

1. Formation of a free surface at a second phase particle or inclusion by either

interface decohesion or particle cracking

2. Growth of a void around the particle, with the aid of hydrostatic stress

3. Coalescence of the growing void with the crack tip

Crack growth by microvoid coalescence is illustrated schematically in Fig. 2.26. The

above events occur continuously as the crack advances. That is, as voids at the crack
tip coalescence, additional voids nucleate and grow further away from the crack tip.
A number of models have been developed to describe this fracture process (45-47).
Rice and Tracey (45) proposed the following equation to describe the growth of a
void.

In(R 0283fexp "m'deq M

where R is the void radius, Ro is the initial radius, ceq is the equivalent plastic strain,
and 0 m is the mean (or hydrostatic) stress, defined as

Gi - Oxx + Y4 U-]
zz
3
Rice and Tracey assumed a nonhardening material in their analysis. More recent
models (46,47) have modified the above expression to take account of strain
hardening.
36

Iclusions

Void forationI

with crack tip

FIG. 2.26 Crack growth by microvold coalescence.

The Rice and Tracey model, as well as the more recent versions, indicate that
microvoid coalescence is essentially strain controlled fracture, but that the
hydrostatic stresses play an important role.

2.4.3 The Ductile-Brittle Transition

The fracture toughness of ferritic steels can change drastically over a small
temperature range, as illustrated in Fig 2.27. At low temperatures, the steel is brittle
and fails by cleavage. At high temperatures, the material is ductile and fails by
microvoid coalescence. Ductile fracture initiates at a particular toughness value, as
indicated by the dashed line in Fig. 2.27. The crack grows as load is increased.
Eventually, the specimen fails by plastic collapse or tearing instability. In the
transition region between ductile and brittle behavior, both micromechanisms of
fracture can occur in the same specimen. In the lower transition region, the fracture
mechanism is pure cleavage, but the toughness increases rapidly with temperature
because cleavage is becoming more difficult. In the upper transition region, a crack
initiates by microvoid coalescence but ultimate failure occurs by cleavage. On initial
 37

loading in the upper transition region, cleavage does not occur because there are no
critical particles near the crack tip. As the crack grows by ductile tearing, however,
more material is sampled. Eventually, the growing crack samples a critical particle
and cleavage occurs. Because fracture toughness in the transition region is governed
by these statistical sampling effects, the data tend to be highly scattered.

Recent work by Heerens and Read (48) demonstrates of the weakest-link sampling
nature of cleavage fracture. They performed a large number of fracture toughness
tests on a quenched and tempered alloy steel at several temperatures in the
transition region. As expected, the data at a given temperature were highly scattered.
Some specimens failed without significant stable crack growth while other
specimens sustained high levels of ductile tearing prior to cleavage. Heerens and
Read examined the fracture surface of each specimen to determine the site of
cleavage initiation. The measured distance from the initiation site to the original
crack tip correlated very well with the measured fracture toughness. In specimens
that exhibited low toughness, this distance was small; a critical nucleus was available
near the crack tip. In the specimens that exhibited high toughness, there were no
critical particles near the crack tip. The crack had to grow and sample additional
material before a critical cleavage nucleus was found. Figure 2.28 is a plot of fracture
fracture toughness versus the critical distance, rc, which Heerens and Read
measu-ed. In every case, cleavage initiated near the location of the maximum
tensile stress. A similar fractographic study by Wantanabe et al. (49) also revealed a
correlation between Jc and rc.

Cleavage
+MVC

FRACTURE
TOUGHNESS Plastic collapse
Cleavage or ductile instability

TEM(PERATURE

FIG. 2.27 The ductile-brlItle transition of ferrltic steels.

38

6000

II

2 00 0

Ty-GC
00

00 0.5 1.0 1.5 2.0

Aa. r., mm
FIG 2.2 Conutlatlon between fracture toughness and the distance from the cracit tip to the site of
cleravae initiation (48).
 39

3. FRACTURE TOUGHNESS TESTING

A fracture toughness test measures the resistance of a material to crack extension.

Such a test may yield either a single value of fracture toughness or a resistance curve,
where a toughness parameter such as K, J, or CTOD is plotted against crack extension.
A single toughness value is usually sufficient to describe a test that fails by cleavage,
because this fracture mechanism is typically unstable. The situation is similar to the
schematic in Fig 2.2(a), which illustrates a material with a flat R curve. For reasons
discussed in Section 3.6, cleavage actually has a falling R curve after initiation. Crack
growth by microvoid coalesence, however, usually yields a rising R curve, such as
that shown in Fig. 2.2(b). Thus ductile crack growth can be stable, at least initially.
When ductile crack growth initiates in a test specimen, that specimen seldom fails
immediately. Therefore one can quantify upper shelf fracture toughness either by
the initiation value or by the entire resistance curve.

There are several ASTM standards for fracture toughness testing. The KIC standard,
ASTM E399-83 (25), is intended for relatively brittle materials or thick sections. The
JIC standard, ASTM E813-87 (26), measures a J value near initiation of ductile tearing.
Another standard, E1152-87 (50), gives guidelines for measuring a J resistance curve.
A CTOD testing standard has been published recently: ASTM E1290-89 (51).

Most standard fracture toughness tests are conducted on one of two specimen types:
the compact specimen and the single edge notched bend (SENB) specimen. The Kic
standard permits two additional specimen configurations but the other standards
mentioned allow only the compact and SENB geometries. Figures 3.1 and 3.2
illustrate the SENB and compact specimens, respectively. The compact specimen is
pin loaded; the SENB specimen is loaded in three point bending at a span of 4W.
Both specimen types contain sharp machined notches from which fatigue cracks are
grown.

Fracture properties of a material typically depend on orientation. For example, a

typical steel plate appears to be much tougher if the crack propagates through the
thickness rather than along the rolling direction, parallel to the plate surface. Thus it
is important to specify the orientation of the fracture specimen. Fig. 3.3 illustrates
the standard nomenclature for specimens extracted from rolled plate.

The field of fracture toughness testing is relatively mature, as evidenced by the

numerous standard test methods that have gained world-wide acceptance, but these
standards fail to address a number of important issues. For example, none of these
standards gives guidelines on weldment testing. In addition, the ductile-brittle
transition region presents unique problems for which current standards are
inadequate.
40

P/2 SPAN = 4WP/

wa B

FIG. 3.1 Single edge notched bond (SENS) specimen.

FIG 2.2 Compact specimen.

 41

The following sections summarize the standardized test methods and describe recent
progress and current research in such areas as weldment testing, scatter, size effects,
loading rate effects, and the relationship between fracture tests.

3.1 KIC TESTING

The ASTM Standard E399 was first published in 1970; the most recent revision was
made in 1983. The title, *Standard Test Method for Plane Strain Fracture Toughness
of Metallic Materials, is somewhat misleading. Although plane strain is a necessary
condition for a valid KIC test, it is not sufficient; a specimen must also behave in a
linear elastic manner. The validity requirements in this standard are very stringent
because even a relatively small amount of plastic deformation invalidates the
assumptions of K theory.

Specimens for KIC tests are usually fabricated with the width, W, equal to twice the
thickness, B. They are fatigue precracked so that the crack length/width ratio (a/W)
lies between 0.45 and 0.55. Thus the specimen design is such that all the critical
dimensions, a, B, and b (- W-a), are approximately equal. This design results in
efficient use of material, since each of these dimensions must be large compared to
the plastic zone.

When a test specimen is loaded to failure, load and displacement are monitored.
Three types of load-displacement curves are shown in Fig. 3.4. The critical load, PQ,
is defined in one of several ways, depending on the type of curve. One must
construct a 5% secant line (i.e. a line from the origin with a slope equal to 95% of the
initial elastic loading slope) to determine P5 . In the case of Type I behavior, the load-
displacement curve is smooth and it deviates slightly from linearity before ultimate
failure at Pnux. This nonlinearity can be caused by plasticity, subcritical crack growth,
or both. For a Type I curve, PQ = P5. With a Type II curve, a small amount of
unstable crack growth (i.e. a pop-in) occurs before the curve deviates from linearity
by 5%. In this case PQ is defined at the pop-in. A specimen that exhibits Type M[
behavior fails completely before achieving 5% nonlinearity. In such cases, PQ = Pmax.-

Once PQ is determined, a provisional fracture toughness, KQ, is computed from the

following relationship:

I 0 M- O- f(a/W) [3.1]

where f(a/W) is a dimensionless function of a/W. This function is given in

polynomial form in the E399 standard for both the compact and SENB specimens.
Individual values of f(a/W) are also tabulated in the standard.
42

FIG. 3.3 Definition of specimen orientations In a rolled plate (25).

0 0 0 ~DISPLACEMENT, v -->

FIG. 3A Typical load-dlsplacement curves In a KA tas 2)

 43

Recall from Section 2.2 that expressions for K can always be reduced to the form of
Eq. [2.61. Equation [3.11 is no exception. If we define a characteristic stress as P/(BW),
the geometry correction factor for the compact and SEN1 specimens is given by

Y = f(a/W) (W )1/2 [32]

The characteristic stress, P/(BW), has no physical meaning for these test specimens
since they are loaded predominantly in bending. Thus Eq. [3.1] is a more convenient
form in this case than .q.[2.1].

The KQ value computed from Eq. [3.11 is a valid KIC result only if all validity
requirements in the standard are met. The main validity requirements are as
follows.

0.45! a/W < 0.55 [3.3a]

Pux:< 1.10 PQ [33b]

B, W-a a 2.5 oQ [34

The last requirement is a restatement of Eq. [2.29]. If the test meets all of the above
criteria as well as additional requirements of ASTM E399, then KQ = KIC.

Most structural steels cannot meet the validity requirements of E399 except at very
low temperatures, as demonstrated by a few sample calculations. Consider a
medium strength steel with ays - 50 ksi. On the upper shelf, the fracture toughness
corresponding to initiation of ductile crack growth is typically around 200 ksi -JiW'for
such materials (23). Substituting these values for strength and toughness into Eq.
[3.3c], reveals that a fracture toughness specimen must be 40 in thick to obtain a valid
KIC! If the 50 ksi material is produced as 1 in thick plate, the maximum valid KIC
that can be measured is 32 ksi N[i . If a structure made from a material with this
toughness were loaded to half the yield strength (25 ksi), the critical crack size
(estimated from Eq. [2.5]) would be approximately 0.5 in. If this were a welded
structure with yield magnitude residual stresses, the critical crack size would only be
0.06 in.

Thus it is virtually impossible to perform a valid KIC test on most structural steels at
ambient temperatures. Such a material could meet the validity requirements with a
reasonable section thickness only by poor fabrication practice or by cooling the
material so that it is on the lower shelf of toughness. In either case, the material
would probably be too brittle for structural application. When linear elastic test
.4
methods are invalid, fracture toughness must be quantified by an elastic-plastic
parameter such as J or CTOD.

3.2 JIc AND J-R CURVE TESTING

There are two ASTM standards currently for J testing. The JIc standard, E813(26),
which was first published in 1981 and revised in 1987, outlines a test method for
estimating the critical J near initiation of ductile crack growth. The J-R curve testing
standard, E1152(50), was first published in 1987.

Both test methods produce a J-R curve, a plot of J versus crack extension. The E1152
standard applies to the entire J-R curve; E813 is concerned only with Jic, a single
point on the R curve. The same test can be reported in terms of both standards. This
is analogous to a tensile test, where one can report either the yield strength or the
entire stress-strain curve.

In the case of the JIC standard, the R curve can be generated by either multiple
specimen or single specimen techniques. With the multiple specimen technique, a
series of nominally identical specimens are loaded to various levels and then
unloaded. Some stable crack growth occurs in most specimens. This crack growth is
marked by heat tinting or fatigue cracking after the test. Each specimen is then
broken open and the crack extension is measured. The most common single
specimen test technique is the unloading compliance method. The crack length is
computed at regular intervals during the test by partially unloading the specimen
and measuring the compliance. As the crack grows, the specimen becomes more
compliant (less stiff). Both E813 and E1152 provide polynomial expressions that
relate a/W to compliance. An alternative single specimen test method is the
potential drop procedure, yet to be standardized by ASTM, in which crack growth is
monitored through the change in electrical resistance which accompanies a loss in
cross sectional area. Both single specimen procedures are practical only in
conjunction with a computer data acquisition and analysis system.

Regardless of the method for monitoring crack growth, a corresponding J value must
be computed for each point on the R curve. For estimation purposes, both standards
divide J into elastic and plastic components:

J = Jel + Jp1 [3.4]

The elastic J is computed from the elastic stress intensity:

2
- K (1-v)
E [35]

where K is computed from load with Eq. [3.13. The Jic standard enables the plastic J at
each point on the R curve to be estimated from the plastic area under the load-
displacement curve:
 45

Jp- Bbo

where il is a dimensionless constant, ApI is the plastic area under the load-
displacement curve (see Fig. 3.5), and bo is the initial ligament length. For an SENB
specimen,

Ti= 2.0 [3.7a]

For a compact specimen,

= 2 + 0.522 bo/W [3.7b]

Equation [3.6] was derived from the energy release rate definition of J (Eq. [2.21]).

The J-R curve testing standard requires a somewhat more detailed calculation of JpI,
performed incrementally since the crack length increases and the net cross-sectional
area decreases throughout the test. Since E813-87 is concerned with the initiation
point rather than the shape of the R curve, the more detailed formula that corrects J
for crack growth is optional in JIC tests. In the limit of a stationary crack, both
formulas give identical results. Thus the measured initiation toughness is
insensitive to the choice of J equation.

The E813-87 procedure for computing JQ, a provisional Jic, from the R curve is
illustrated in Fig. 3.6. Exclusion lines are drawn at crack extension (Aa) values of 0.15
and 1.5 mm. These lines have a slope of 2ay, which corresponds approximately to
the component of crack extension that is due to crack blunting, as opposed to ductile
tearing. A horizontal exclusion line is defined at a maximum value of J:
J =boay [3
15

All data that fall within the exclusion limits are fit to a power law expression:

J = CI(Aa) C 2 [3.91

The JQ is defined as the intersection between Eq. [3.9] and a 0.2 mm offset line. If all
other validity criteria are met, JQ - JIc as long the following size requirements are
satisfied:

B,bo a 2[3.10
Uy
46

UNLOADING LINE

LOAD LINE DISPLACEMENT -

PIG. 3.5 Definition of the plastic area for J,,calculations (26).

0 POINTS U29D PON RIGUORS ADMIXES

an INLVRW
~Uam
g~w -

Ib

an3 an an 1.5 In 1. 1.A ?A LU

CXA"c u91Inumim

FIG. 3.6 Typical J-R curve whi1ch Illustrate* the calculation of J. (26).
 47

The J-R curve standard has limits on Jmax and crack extension relative to sp.ldmen
size:

B,ba20 J[x111

and

s 0.10 bo I32
Figure 3.6 (50) shows a typical J-R curve with the E1152-87 validity limits. The
po ion of the J-R curve that falls outside these limits is considered invalid. Note
thailthe limit on Jwx in ASTM E1152-87 is more severe than the Jnax limit of ASTM
E813-87 (Eq. 3.8]), but this limit is less restrictive than the JIc size criterion (Eq. [3.101).

3.3 CTOD TESTING

Because of the strict limits on plastic deformation, the KIc test can only be used on
the lower shelf of toughness in structural steels and welds. The JIC and J-R curve test
methods allow considerably more plastic deformation, but these tests are only valid
on the upper shelf. The CTOD test is currently the only standardized method to
measure fracture toughness in the ductile-brittle transition region.

The first CTOD test standard was published in Great Britain in 1979(52). ASTM
recently published E1290-89, an American version of the CTOD standard. The British
CTOD standard allows only the SENB specimen; the ASTM standard provides for
CTOD measurements on both the compact and SENB specimens. Both standards
allow two configurations of SENB specimens: 1) a rectangular cross section with W =
2B, the standard geometry for KIC and JIc tests; and 2) a square cross section with W -
B. The rectangular specimen is most useful with L-T or T-L orientations (see Fig.
3.3); the square section is generally used in the L-S or T-S orientations.

Experimental CTOD estimates are made by separating the CTOD into elastic and
plastic components, similar to the JIc and J-R tests. The elastic CTOD is obtained
from the elastic K:

2(1-v2)
&dUK 1-3

The elastic K is related to applied load through Eq. [2.11. The above relationship
assumes that dn = 0.5 for linear elastic conditions (see Eq. [2.261). The plastic
component of CTOD is obtained by assuming that the test specimen rotates about a
plastic hinge. This concept is illustrated in Fig. 3.7 for an SENB specimen. The
48

FIG. 3.7 Rotation of an SENB specimen about a plastic hinge.

plastic displacement at the crack mouth, Vp, is related to the plastic CTOD through a
similar triangle construction:

= ry (W-a) Vp [3.14]
rp (W-a) + a + z

where rp is the rotational factor, a constant between 0 and 1 that defines the relative
position of the apparent hinge point. The mouth opening displacement is measured
with a clip gage. In the case of an SENB specimen, knife edges must often be attached
in order to hold the clip gage. Thus Eq. [3.14] must correct for the knife edge height, z.
The compact specimen can be designed so that z = 0. The plastic component of V is
obtained from the load-displacement curve by constructing a line parallel to the
elastic loading line, as illustrated in Fig. 3.8. The plastic rotational factor is given by

rp = 0.44 [3.15a]

for the SENB specimen and

rpM0.41+2[(
sp.e5 )2+a+ 0.5J1/2 -Q [.151,n

for the compact specimen.

 49

* Pait
LOAD

acture
I
I
I
I

I
I
I I
I I

MOUTH OPENING DISPLACMENT

FIG. 3.8 Estimation of VP from the load-displacement curve.

Figure 3.7 indicates that the crack mouth opening displacement, V, on an SENB
specimen is not the same as the load line displacement, A. The latter displacement
measurement is required for J estimation because Api in Fig. 3.5 and Eq. [3.6]
represents the plastic energy absorbed by the specimen. The CTOD standard utilizes
VP because this displacement is easier to measure in SENB specimens. However, if
rp is known, it is possible to infer J from a P-V curve or CTOD from a P-A curve
(53,54). The compact specimen simplifies matters somewhat because V = A as long as
z = 0.
The CTOD standard test method can be applied to ductile and brittle materials, as
well as steels in the ductile-brittle transition. This standard includes a notation for
critical CTOD values that describes the fracture behavior of the specimen:

8- Critical CTOD at the onset of unstable fracture without prior stable crack
growth. This corresponds to the lower shelf and lower transition region of
steels where the fracture mechanism is pure cleavage.

8- Critical CTOD at the onset of unstable fracture which has been preceded by
stable crack growth. In the case of ferritic steels, this corresponds to the
"ductile thumbnail" observed in the upper transition region (see Fig. 2.27).
50
Bt - CTOD at the initiation of stable crack growth. This measure of toughness
is analogous to JiC.

8= - CTOD at the maximum load plateau. This occurs on or near the upper
shelf of steels.

Figure 3.9 is a series of schematic load-displacement curves that manifest each of the
above failure scenarios. Curve (a) illustrates a test that results in a 8 c value; cleavage
fracture occurs at Pc. Figure 3.9 (b) corresponds to a 8u result, where ductile tearing
precedes cleavage. The ductile crack growth initiates at Pi. A test on the upper shelf
produces a load-displacement curve like Fig. 3.9(c). A maximum load plateau occurs
at Pm. The specimen is still stable after maximum load because the material has a
rising R curve and the test is performed in displacement control. (See Section 4.6 for
further discussion on stable and unstable crack growth.) Three types of CTOD result,
8r, 8 u and 8m are mutually exclusive; i.e, they cannot occur in the same test. It is
possible, however, to measure a 81 value in the same test as either a 8 m or 8u result.

Pu
P
PC
LOAD
Fracture

Fracture

(a) ) (c)

MOUTH OPENING DISPLACEMENT

FIG. 3.9 Typical load-displacement curves that Illustrate possible behavior In a

CTOD test.
a) 8c result; b) 8u result; c) 8 m result.
 51

As illustrated in Fig. 3.9, there is usually no detectable change in the load-

displacement curve at Pi. The only deviation in the load-displacement behavior is
the reduced rate of increase in load as the crack grows. The maximum load plateau
(Fig. 3.9c) occurs when the rate of strain hardening is exactly balanced by the rate of
decrease in the cross section. However, the initiation of crack growth can not be
detected from the load-displacement curve because the loss of cross section is
gradual. Thus Si must be determined from an R curve.

As with the JIC standard, the 8-R curve can be generated by either single or multiple
specimen procedures, but the way in which the initiation point is defined differs
from ASTM E813. Figure 3.10 illustrates the E1290-89 procedure for Bj determination
from a 8-R curve. CTOD values are plotted against the physical crack extension, Aap.
Vertical exclusion lines are drawn at 0.15 and 1.5 nun of crack extension. The data
are then fit to an offset power law expression:

8 = C1 (C2 + Aap)C3 3.16J

The initiation toughness, 8t, is then defined at the CTOD value corresponding to AaP
- 0.2 mm.

COD I Exclusion lines

0 01

-C:2 0.15 nan 15 mun

CRACK EXTENSION
FIG. 3.10 CTOD R curve for the determination of 81 according to ASTM E1290-89 (51).
52
The only specimen size requirement of the British and ASTM CTOD standards is a
recommendation to test full section thicknesses. For example, if a structure is to be
made of 1 in thick plate, then B in the test specimens should be nominally I in. If
the specimen is notched from the surface (L-S or T-S orientations), a square section
specimen is required for B to equal the plate thickness. The British CTOD standard
allows a/W ratios ranging from 0.15 to 0.75, while the ASTM standard restricts the
permissible a/W values to the range of 0.45 to 0.55. In an earlier draft, the ASTM
standard allowed a much wider range of a/W in the square section specimen, but
concerns about the possible variation of rp with a/W led the members of the ASTM
CTOD task group to restrict the allowable range until the uncertainties in rp could be
resolved. Future revisions of this standard will probably relax the a/W restriction.
Shallow cracked specimens have certain advantages, particularly for weldment tests
(see Section 3.4).

The ASTM CTOD standard recognizes that since it does not contain size
requirements relative to the degree of plasticity, the measured critical CTOD value
may be geometry dependent. The standard correctly assumes that if a test specimen's
thickness matches that of the structure being evaluated, the crack tip conditions in
the specimen will be at least as severe as in the structure. Experimental data and
numerical analysis indicate that a small scale bend specimen is more highly
constrained than a large panel loaded in tension.

3.4 WELDMENT TESTING

All the standards discussed are suitable for specimens extracted from uniform
sections of homogeneous material, but welded joints have decidedly heterogeneous
microstructures and, in many cases, irregular shapes. Weldments also contain
complex residual stress distributions. Existing fracture toughness testing standards
do not address the special problems associated with weldment testing. The factors
that make weldment testing difficult (i.e. heterogeneous microstructures, irregular
shapes, and residual stresses) also tend to increase the risk of brittle fracture in
welded structures. Thus, one cannot simply evaluate the regions of a structure
where ASTM testing standards apply and ignore the fracture properties of
weldments.

Although there are currently no fracture toughness testing standards for weldments,
a number of laboratories and industries have significant of experience in this area.
The Welding Institute in Cambridge, England which probably has the most expertise,
has recently published detailed recommendations for weldment testing (55). The
International Institute of Welding (MW) has produced a similar document (56),
although not as detailed. The American Petroleum Institute (API) has published
guidelines for heat affected zone (HAZ) testing as part of a weld procedure
qualification approach (57). Committees within ASTM and the British Standards
Institute (BSI) are currently drafting weldment test methods, relying heavily on 20
years of practical experience as well as the aforementioned documents.
 53

Some of the general considerations and current recommendations for weldment

testing are outlined below, with emphasis on the Welding Institute procedure (55)
because it is the most complete document to date. Early drafts of both the ASTM and
BSI documents are similar to The Welding Institute approach.

When performing fracture toughness tests on weldments, a number of factors need

special consideration. Specimen design and fabrication is more difficult because of
the irregular shapes and curved surfaces associated with some welded joints. The
heterogeneous microstructure of typical weldments requires special attention to the
location of the notch in the test specimen. Residual stresses make fatigue
precracking of weldment specimens more difficult. After the test, a weldment must
often be sectioned and examined metallographically to determine whether or not the
fatigue crack sampled the intended microstructure.

3.4.1 Specimen Design and Fabrication

The underlying philosophy of the Welding Institute (55) guidelines on specimen

design and fabrication is that the specimen thickness should be as close to the section
thickness as possible. Thicker specimens tend to produce more crack tip constraint,
and hence lower toughness (See Chapter 2). Achieving nearly full thickness
weldment specimens often requires sacrifices in other areas. For example if a
specimen is to be extracted from a curved section such as a pipe, one can either
produce a subsize rectangular specimen which meets the tolerances of the existing
ASTM standards, or a full thickness specimen that is curved. The Welding Institute
recommends the latter.
Figure 3.11 illustrates the recommended tolerances for misalignment, curvature and
angular distortion in SENB specimens (55). Distortion and curvature in the long axis
of the specimen is limited to 10% of W (Fig. 3.11a,b,c); curvature in the width
diredion (Fig. 3.11d) is limited to 2.5% W. Note that the top and bottom surfaces in
the SENB specimen must still be flat and parallel at the loading points. If a
weldment specimen has more than 10%W curvature or distortion in the long axis,
the specimen can be straightened by bending on either side of the notch to produce a
"gull wing" configuration, which is illustrated in Fig. 3.12. The bending must be
performed so that the three loading points are aligned.

Fabrication of either a compact or SENB weldment specimen is possible, but the

SENB specimen is preferable in nearly every case. Although the compact specimen
consumes less material (for a given B and W) in parent metal tests, it requires more
weld metal in a through-thickness orientation (L-T or T-L) than an SENB specimen.
It is impractical to use a compact geometry for surface notched specimens (T-S or L-
S). Such a specimen would be greatly undersized with the standard B X 2B geometry.
54

IV

FIG. 3.11 Recommended allowances for mlsallgnent, angular distortion and

curvature (65).

FIG. 312 OGuUl wing" configuratio which correcs flor excessive curvature and angluar
distortion (5).
 55

The Welding Institute recommendations cover both the rectangular and square
section SENB specimens. The appropriate choice of specimen type depends on the
orientation of the notch.

3.4.2 Notch Location and Orientation

Weldments have a highly heterogeneous microstructure. Fracture toughness can

vary considerably over relatively short distances.. Thus it is important to take great
care in locating the fatigue crack in the correct region. If the fracture toughness test is
designed to simulate an actual structural flaw, then the fatigue crack must sample
the same microstructure as the structural flaw. For a weld procedure qualification or
a general assessment of a welduent's fracture toughness, location of the crack in the
most brittle region may be desirable, but it is difficult to know in advance which
region of the weld has the lowest toughness. In typical C-Mn structural steels, low
toughness is usually associated with the coarse grained HAZ and the intercritically
reheated HAZ. A microhardness survey can help identify low toughness regions
because high hardness is often coincident with brittle behavior. The safest approach
is to perform fracture toughness tests on a variety of regions in a weldment.

Once the microstructure of interest is identified, a notch orientation must be

selected. The Welding Institute gives two alternatives: a through-thickness notch
and a surface notch, as illustrated in Fig. 3.13. Since full thickness specimens are
desired, the surface notched specimen should be square section (BXB), while the
through thickness notch will usually be in a rectangular (B X 2B) specimen.

For weld metal testing, the through-thickness orientation is usually preferable

because a variety of regions in the weld are sampled. Since brittle fracture is a
weakest link phenomenon (Section 2.4), the fracture behavior of the test specimen is
controlled by the region with lowest toughness. However, there may be cases where
the surface notched specimen is the most suitable for testing the weld metal. For
example, a surface notch can sample a particular region of the weld metal, such as
the root or cap, or the notch can be located in a particular microstructure, such as
unrefined weld metal.

Notch location in the HAZ often depends on the type of weldment. If welds are
produced solely for mechanical testing, for example as part of a weld procedure
qualification or a research program, the welded joint can be designed to facilitate
HAZ testing. The Welding Institute (55) recommends a K or a half-K preparation to
simulate double-V and single-V welds, respectively. These geometries are illustrated
in Fig. 3.14. The plates should be tilted when these weldments are made, to have the
same angle of attack for the electrode as in an actual single- or double-V joint. For
fracture toughness testing, a through-thickness notch is pl- --_d in the straight side of
the K or half-K HAZ.
56

a) Through-thickness notch. b) Surface notch.

FIG. 3.13 Notch orientation for weidment specimens (55).

Notch $zne

a) alf-K preparation.

b) Kpreparation.
Abth piwi.

FIG. 3.14 Special welded joints for HAZ fracture toughness testIng (55).
 57

In many instances, fracture toughness testing must be performed on an actual

production weldment, where the joint geometry is governed by the structural design.
In such cases, a surface notch is often necessary for the crack to sample sufficient
HAZ material. The measured toughness is sensitive to the volume of HAZ material
sampled by the crack tip because of the weakest link nature of cleavage fracture (see
Section 3.5). Figure 3.15 illustrates several types of notches in production welds.
According to the Welding Institute (55), a through thickness notch is acceptable if the
HAZ is sufficiently straight for at least one-third of the crack front to be located in the
HAZ (Fig. 3.15a). The disadvantage of the notch orientation in Fig. 3.15(a) is that the
notched region of the HAZ is on the edge of the specimen, where the constraint is
low.
Another application of the surface notched orientation is the simulation of
structural flaws. Figure 3.16 illustrates HAZ flaws in a structural weld (Part a) and a
surfaced notched fracture toughness specimen that models one of the flaws.

Figures 3.15 and 3.16 demonstrate the advantages of allowing a range of a/W ratios
in surface notched specimens. A shallow notch is often required to locate a crack in
the desired region, but existing ASTM standards do not allow a/W ratios less than
0.45. As stated in Section 3.3, more liberal a/W allowances were removed from the
CTOD standard because of uncertainty on the effect of crack depth on the rotational
factor. An additional problem is the tendency of shallow notched fracture toughness
specimens to have lower constraint than deeply cracked specimens. The question of
the rotational factor can be resolved fairly easily but the constraint issue is more
difficult. Some fracture mechanics specialists argue that shallow notched tests
should be permissible because they simulate the constraint in an actual structure
which contains a shallow flaw. Others believe that a fracture mechanics test should
be performed under conditions where a single parameter characterizes crack tip
conditions (see Section .3.7). Such a test would yield a geometry-independent
fracture toughness value that could be viewed as a material property. A test on a
shallow notched specimen, they argue, is invalid because it does not satisfy these
criteria. The Welding Institute and the Edison Welding Institute are addressing
these issues in a cooperative research program.

3.4.3 Fatigue Precracking

Weldments that have not been stress relieved typically contain complex residual
stress distributions that interfere with fatigue precracking of fracture toughness.
specimens. Figure 3.17, a typical fatigue crack that is produced in an as-welded
specimen, was extracted from a double-V weld and was given a through-thickness
notch. The fatigue crack did not grow in the center of the specimen because the
residual stresses were compressive in this region.
58

NXF

(i

FIG. &IS Various notch orentations for HAZ simlmesh

0) Structure. b)Test specimen.

FIG. &1i6 Skmulation of structural flaws with a surfac notched SENS specimen (55).
 59

Towers and Dawes (58) evaluated the various methods for producing straight fatigue
cracks in welded specimens, including reverse bending, high R ratio, and local
compression.

The first method bends the specimen in the opposite direction to the normal loading
configuration to produce residual tensile stresses along the crack front that
counterbalance the compressive stresses. Although this technique gives some
improvement, it does not produce acceptable fatigue crack fronts.

The R ratio in fatigue cracking is the ratio of the minimum stress to the maximum.
A high R ratio minimizes the effect of residual stresses on fatigue, but also tends to
increase the apparent toughness of the specimen (60). In addition, fatigue
precracking at a high R ratio takes much longer than precracking at R = 0.1, the
recommended R ratio of the various ASTM fracture testing standards.

The only method evaluated which produced consistently straight fatigue cracks was
local compression, which is illustrated in Fig. 3.18. The ligament is compressed to
produce nominally 1% plastic strain through the thickness, mechanically relieving
the residual stresses. However, local compression can reduce the toughness slightly.
Towers and Dawes concluded that the benefits of local compression outweigh the
disadvantages, particularly in the absence of a viable alternative.

3.4.4 Post-Test Analysis

Correct placement of a fatigue crack in weld metal is usually not difficult because this
region is relatively homogeneous. The microstructure in the HAZ, however, can
change dramatically over very small distances (60). Correct placement of a fatigue
crack in the HAZ is often accomplished by trial and error. Because fatigue cracks are
usually slightly bowed, the precise location of the crack tip in the center of a
specimen cannot be inferred from observations on the surface of the specimen. Thus
HAZ fracture toughness specimens must be examined metallographically after the
test to determine the microstructure that initiated fracture. In certain cases, post-test
examination may be required in weld metal specimens.

Figure 3.19 illustrates the Welding Institute's (55) recommended procedure for
sectioning surface notched and through-thickness notched specimens. First, the
origin of the fracture must be located by the chevron markings on the fracture
surface. After marking the origin with a small spot of paint the specimen is
sectioned perpendicular to the fracture surface and examined metallographically.
The specimen should be sectioned slightly to one side of the origin and polished
down to the initiation site. The spot of paint appears on the polished specimen
when the origin is reached.
60

Machined

Fatigue crack

FIG. 3.17 Fracture surface of weidment specimen with Invalid fatigue crack (58).

A(or 0O5% Son ach side?

Sechon; A-A

FIG. 3.18 Local compression applied to the ligament of a weldment specimen to relieve
residual stresses prior to fatigue cracking (55).
 61

Li fMqqt#dV RAI Muclined On69n6W pbiw

a) Through-thickness notch. VO~ fghc,;f

b) Surface notch. inihatien p~i

FIG. 3.19 Post-test sectioning techniques for HAZ specimens (55).

otigue crack hi ~ Fat iq~ecokt

Positintcrk

ils 3.% tb
8) S. a.02 MM. b) S. 67mm

FIG. 3.20 Metallographic sections from two surface notched Specimens (69).
62

Figure 3.20 shows photomicrographs of two typical sections of fracture toughness

specimens. Fracture in the specimen on the left initiated at the fusion boundary at a
Ac of 0.02 mm. The crack propagated a short distance and arrested in the weld metal.
Further loading was required to cause complete failure. The specimen on the right
fractured at a 8 u of 0.67 mm. The initial fatigue crack was in the weld metal, which
has a reasonably high toughness. On loading, the crack blunted and grew by ductile
tearing. Cleavage fracture occurred when the growing crack reached the HAZ
These two cases illustrate how slightly different notch placement can result in vastly
different fracture behavior.

The API document RP2Z (57) outlines a post-test analysis of HAZ specimens which is
more detailed and cumbersome than the procedure outlined above. In addition to
sectioning the specimen, the amount of coarse-grained material at the crack tip must
be quantified. For the test to be valid, at least 15% of the crack front must be in the
coarse-grained HAZ. The purpose of this procedure is to prequalify welding
procedures and steels with respect to HAZ toughness, identifying those that produce
low HAZ toughness so that they can be rejected before fabrication.

3.5 SCATTER AND SIZE EFFECTS IN THE TRANSITION REGION

In the ductile-brittle transition region of steels, ultimate failure occurs by cleavage;
but this is often preceded by extensive plastic deformation and/or ductile tearing.
There are a number of problems associated with fracture toughness testing in the
transition region. Toughness data tend to be highly scattered because of the weakest-
link nature of cleavage fracture, and existing measures of fracture toughness in the
transition region often depend on specimen size.

The CTOD test is the only standardized method that is valid in the transition region,
but it has no specimen size limitations to guarantee a single parameter description of
crack tip conditions. As discussed in Chapter 2, a single parameter description is
required for a geometry-independent measure of fracture toughness. The JIc
standard contains size requirements (Eq. [3.10]) for initiation of ductile tearing. A
material's resistance to cleavage fracture can, in principle, be quantified by a critical J
value. In fact, the nuclear power industry uses such an approach in the transition
region, but it has not been standardized by ASTM. Size requirements for critical J
values for cleavage have not been established. Equation [3.10] is probably not
stringent enough for the transition region because cleavage fracture toughness is
more sensitive to constraint loss than is ductile tearing (61).

Recent research has led to a number of simple, yet accurate analyses of scatter in
cleavage fracture toughness data (44), but they are valid only as long as the crack tip is
characterized by a single parameter such as J or CTOD. Thus the analyses are
restricted to the lower transition region. In the upper transition region, the stresses
at the crack tip relax and the crack grows by ductile tearing; fracture toughness
becomes geometry dependent and the data scatter worsens.
 63

3.5.1 Scatter in the Lower Transition Region

When an experimental measurement displays a high degree of scatter, it is

customary to attempt to reduce this scatter by exercising tighter control over the
experiment. Much of the observed scatter in cleavage fracture toughness data is an
inherent feature of the material, however, that cannot be eliminated by improving
the experimental procedure. Fortunately, fracture toughness scatter obeys certain
rules and its magnitude can be predicted.

Statistical Aspects of Fracture

A number of micromechanical models for cleavage fracture have been proposed in
the past few years(35,38-44); all based on weakest link statistics. Each model relates a
material's microstructure to cleavage fracture toughness.

When fracture occurs by a weakest link mechanism, the failure probability is equal to
the probability of sampling at least one critical microstructural feature. The Poisson
distribution appropriately describes the failure probability in this case (43):

F1- [3.17]

where F is the failure probability, V" is the volume in the process zone ahead of the
crack tip, and

# particles 2 critical size

unit volume

The parameter p contains the inherent fracture properties of the microstructure. The
macroscopic fracture properties of a structure or test specimen can be obtained by
integrating p over the appropriate volume. In a uniaxial tensile specimen, p is
constant throughout the gage section if the stress is constant, but p varies with
position when a crack is present because p depends on stress. The critical particle size
is related to stress through Eq. [2.311. As stress increases, critical size decreases; p
increases with stress because more particles are capable of initiating cleavage.
Anderson (43) modeled the stress dependence of p with a modified power law-

p - ([3.181
V

where (Y is the maximum principal stress at a point, and Yu, m and iV are material
constants that characterize the fracture properties of the microstructure. The
64

quantity ou is the threshold stress, corresponding to the largest conceivable particle

in the material.

If fracture occurs by a weakest link mechanism and the crack tip stresses are
completely characterized by J (e.g. Fig. 2.16 a-b), it can be shown (44) that fracture
toughness data follows a characteristic distribution function:

F=1- B (j)21[3.19a]

or

F;=1 4 Bo (8c2J [3.19b]

or

04- O OK j
for critical J, CTOD and K values, respectively; Bo is a reference thickness. The
quantities Oj, 08, and eK are the 63rd percentile values of the corresponding fracture
toughness parameters when specimen thickness = Bo. These three constants are
related through Eqs. [2.26] and [2.28]. Equation [3.19] is derived in the Appendix. A
number of investigators (44, 62, 63) have derived nearly identical cleavage fracture
toughness distributions.

Equations [3.19 a-c] have the form of a Weibull distribution. The Weibull shape
parameter (or slope) = 2.0 for critical J and CTOD values, and because of the
relationship between J and K, the shape parameter - 4.0 for KIC or effective K values
computed from J (see Eq. [2.28]).

Figure 3.21 (64) gives an experimental validation of a theoretical Weibull slope.

Experimental values of the Weibull slope from published-fracture toughness data are
plotted as a function of data set size. All J and CTOD data were converted to
equivalent KIC values through Eqs. [2.26] and [2.28]. Although the computed
Weibull slopes are widely scattered for small sample sizes, the data approach the
theoretical value of 4.0 as sample size increases. The scatter in computed slopes can
be attributed to small sample effects. The curves in Fig. 3.21 represent 95%
confidence limits which were computed by Monte Carlo simulation. Finite samples
were randomly taken from a hypothetical population of fracture toughness data with
a true slope of 4.0, and Weibull slopes were computed. This process was repeated
numerous times for various sample sizes. Note that the computed confidence limits
 65

describe well the variability of the real experimental data. Thus Fig. 3.21 shows
strong experimental evidence for the validity of Eq. [3.191.

The discovery of a characteristic Weibull slope for fracture toughness distributions

has important ramifications. If the 63rd percentile fracture toughness of a given
material is known, the entire toughness distribution is defined by Eq. [3.19. Thus
scatter in cleavage fracture toughness data can be quantified.

100 .. . . ... .

o Stmadel. 1986
O Shabbits. 1969
0MGowan, o Server, 1978
1978
00 0 Pugh, 1985
Mrand, 198 1
01
0 [ -t
" .A Bryan, 1987

m. Watanabe, 1987

of 4.0T195%confidence
(64). LimitsweeenrtdyMoeCro

simulation.
66

Analsis of Scatter
The value of Eq. [3.191 can be illustrated by introducing two quantities that
characterize scatter. the confidence band ratio and the lower bound ratio. The
confidence band ratio is obtained by dividing the confidence band width by the
median value. For example, the 80% confidence band ratio for CTOD data is given
by

CBRo. 8o.5 . ln(0.1)-

4-ln(o.)--ln(0.9) 1.43 [320]

where 80.9, g0,, and Bo are the 90% upper bound, 10% lower bound and median
CTOD values, respectively. Note that the confidence band ratio is independent of
temperature, specimen thickness, and material properties. Although the scatter in
absolute terms increases with median toughness, the relative scatter is constant.

The lower bound ratio is defined as the ratio of the median to a particular lower
limit in the distribution. This is probably a more useful quantity than the CBR
because the lower tail of a toughness distribution causes the most concern in
structural applications. The 10% lower bound ratio is given by

LB80.5 ,l
LBR .1 = - = 2-57 [3.21a]
1n(0.9)
Similarly, the 1%lower bound ratio can be computed:

LBRo.01 =I (0.99) -8.30 [321b]

Thus, if the median CTOD is known, the 1 percent lower bound toughness can be
estimated by dividing the median by 830.

The confidence band and lower bound ratios would be identical for J and CTOD data
since both parameters have the same Weibull exponent, but K values have a
Weibull exponent of 4.0, causing the data scatter to be somewhat different. For
example, the 1 percent lower bound ratio for critical stress intensity data is given by

Kos = ln(O.5) 1l/4[

LBRMo 1 -K.01 I 2.88 [3.21c]

Critical K data for cleavage appears to be less scattered than J or CTOD data only
because both J and CTOD are proportional to K2 under small-scale yielding
 67

conditions. No benefit would be realized by quantifying toughness with K rather

than J or CTOD. The uncertainty in critical crack size calculations that results from
toughness scatter is the same regardless of the fracture parameter.

Analysis of Small Data Sets

Equations [3.201 and [3.211 estimate relative scatter, but to determine absolute values
in the distribution with these equations, a single value (e.g. the median) must be
known. Stated another way, at least one point on the distribution must be known in
order to compute 0 and thereby define the ertire distribution. The degree of
confidence to which 0 can be estimated depends on the size of the available data set.

In most cases, a relatively small number of fracture toughness tests are used to
characterize a given material at a constant temperature, partly because fracture
toughness testing tends to be expensive and time consuming. In addition, many still
view fracture mechanics analysis as deterministic; i.e., a single value of toughness is
typically used to estimate critical crack size. Consequently, large fracture toughness
data sets are often unavailable

Stienstra et al. (64) developed an approach for inferring useful information from
small fracture toughness data sets by utilizing order statistics to estimate a lower-
bound toughness value from the available data. The analysis assumes that Eq. [3.19]
describes the toughness distribution. The lower-bound value is related to measured
values by a multiplying factor. For example, the 10% lower-bound CTOD is
estimated with 90% confidence as follows.

80109- ko.1,0.9(n) Snmjn [3.221

where kO.1,0.9(n) is a multiplying factor, n is the sample size, and 8min is the lowest
CTOD in a set of n values. Whenever statistical inferences are made from limited
data sets, two probability values must be specified. The above equation corresponds
to an estimate of the 10% lower bound for the toughness distribution. This is a
conservative estimate because the true 10% lower bound lies above the estimate 90%
of the time. Table 3.1 lists multiplying factors for four combinations of confidence
levels and lower bounds. This approach to lower bound estimates has a built-in
retest rational, as illustrated in the following example.

Consider a set of three CTOD values for a given material, with 8rain - 0.10 mn.
According to Table 3.1, the 10% lower bound, estimated with 90% confidence, is 0.037
mm. Imagine a hypothetical application where 80.1 ,.9 must be at least 0.05 mm.
Given the existing data, one must conclude that the material's toughness is
inadequate for the application. However, additional tests may reveal that the
material is acceptable and that the low CTOD value in the first three tests was a rarity.
If a total of seven tests are performed and Smin remains at 0.10 mm, the estimated
lower bound increases to 0.057, which is above the minimum acceptable value.

The use of small data sets to estimate lower bound values is not without cost. In
order to obtain 90 or 95% confidence that the estimate is below the true value, the
estimate must be conservative when the sample size is small. In general, the
estimate is less conservative if more data are available (64). Thus there is a trade-off
between the cost of performing a large number of fracture toughness tests and the
cost of rejecting a material that may be adequate for the application.

TABLE 3.1. Multiplying factors for estimating lower bound toughness from
small data sets(64).

NUMBER OF MULTIPLYINZG FACTOR (k)

VALID 10% Lower Bound 5% Lower Bound
TESTS 90%Confidence 95%Confidence 90%Confidence 95%Confidence

3 0.3705 0.3248 0.2585 0.2266

5 0.4783 0.4193 0.3337 0.2926
7 0.5660 0.4962 0.3949 0.3462
10 0.6764 0.5930 0.4720 0.4138
15 0.8285 0.7260 0.5781 0.5068
20 0.9566 0.8387 0.6675 0.5852

Thickness Effects
The weakest link nature of cleavage leads to a size dependence on fracture
toughness. This is apparent from Eq. [3.19] because thickness appears in the
exponent For a constant failure probability, critical J and CTOD values are
proportional to 1/TB-, and Kic (or Keff) is proportional to B-1/ 4 . Early studies (7,65)
on the size dependence of critical K values concluded that a plateau value is reached
above a critical thickness, but the statistical model predicts that average KIC decreases
by 16% each time thickness is doubled. Early investigators probably lacked sufficient
data to detect the continual decrease in toughness. Statistical thickness effects can be
masked by the scatter in toughness.
Lar.des and Schaffer (66) were the first to apply a weakest-link statistical argument to
relate fracture toughness data from large and small specimens. They fit Weibull
distributions to small specimen data and estimated the average toughness of large
specimens with the following relationship:
 Avg. KIC (large specimen) = Avg. KIC (small specimen) X [D.23)

where 0 is the Weibull slope. The Landes and Schaffer analysis of thickness effects is.
similar to Eq. [3.191, except they left the Weibull slope as a variable. The discovery of
a characteristic Weibull slope for cleavage toughness (44, 62, 63) came several years
after the Landes and Schaffer study.

Specimen size effects on fracture toughness cannot always be explained completely by

weakest link statistics. When fracture is preceded by large scale plasticity or ductile
tearing, additional size effects are observed (67), as discussed in Section 3.5.2

Threshold Toughness
The decrease in toughness with increasing thickness does not continue indefinitely.
Fracture toughness data exhibit an absolute minimum that is above zero (62, 64).
One reason for this is that the fatigue precracking procedure introduces warm
prestress effects at the crack tip. The plastic zone formed during precracking results
in compressive residual stresses when the specimen is unloaded. When the
specimen is reloaded at the test temperature, these stresses must be overcome for
fracture to occur. The compressive residual stresses can be represented by a negative
stress intensity that reduces the effective KI:

Keff= KI- Ko [3.24]

where KI is the apparent stress intensity, and K is the stress intensity of the
compressive residual stresses. Equation [3.19c] can be modified to account for warm
prestress effects:

FP=oo 1K- rKO

'L<o
I-K [3.21

Thus Ko represents the threshold toughness for the material, analogous to the
threshold AK caused by crack closure effects during fatigue.

Experimental and theoretical estimates of Ko indicate that it is approximately 20

MWaN"'for steels. The temperature dependence of Ko is unclear. Current research
at Texas A&M University is establishing more accurate estimates of Ko.

The threshold toughness significantly influences the toughness distribution only on

the lower shelf, where the average toughness is of the same order as Ko. In the
transition region, Ko can be neglected for most analyses. Unless otherwise specified,
Ko is assumed to equal zero in this document, and the cleavage fracture toughness
distribution is described with Eq. [3.19].
0

Local Brittle Zones

Some materials contain localized regions of low toughness. Since cleavage fracture
is a weakest-link phenomenon, these local brittle zones (LBZs) can dominate the
fracture process.

The most common instance of local brittle zones is in the heat-affected zone (HAZ)
of a steel weldment (60, 68, 69). A typical weld HAZ contains a wide range of
microstructures because of the various thermal treatments encountered. For
example, the material at the fusion boundary experiences a very high austenitizing
tanperature and cooling rate, while material a short distance away is subjected to a
much lower temperature and cooling rate. In addition, certain regions are tempered
by subsequent welding passes. Some of the microstructures t produced are extremely
brittle.

The fracture behavior of such materials is complex. The crack tip in a fracture
mechanics specimen is likely to sample a number of microstructural regions. The
fracture toughness distribution depends on the relative volume of each
microstructure that is sampled.

The statistical model described above is capable of characterizing fracture toughness

distributions in composite microstructures such as weld HAZs. The contribution of
each microstructural region to the failure probability can be summed to obtain an
overall probability of failure. For example, if p constituent microstructures are
sampled by the crack tip, the CTOD distribution is given by

F-i-x __ [3.26]

where cti is the fraction of the crack front that samples the microstructural
constituent, and 0 i is a characteristic toughness for the constituent. Note that CTOD
still follows a Weibull distribution with the slope = 2. Equation [3.26] reduces to Eq.
[3.19bJ when p a 1.

Consider a duplex (p=2) microstructure with local brittle zones in a matrix that is
three times as tough as the LBZs (i.e., 01 = 902). Figure 3.22 is a plot of normalized
CTOD versus percent LBZ(44). The median CTOD decreases with increasing volume
fraction of LBZ. Beyond 40 percent LBZ, however, the rate of decrease in toughness
is small. Figure 322 includes plots for matrix/LBZ toughness ratios of three, ten, and
infinity. The curve for toughness ratio a 10 is essentially identical to the curve
corresponding to an infinitely tough matrix. This implies that a constituent ten
times as tough as the most brittle constituent does not contribute to the overall
 71

failure probability. Figure 3.23 (44) illustrates the effect of LBZ on scatter when the
toughness ratio is 3.0.
The above analysis is a means for estimating the probability of fracture initiation in
materials with local brittle zones. Cleavage cracks that initiate in the LBZs may arrest
in the surrounding material. The statistical model will be conservative in these
case because it does not consider crack arrest. The crack arrest behavior of complex
microstructures such as weld HAZs is not well enough understood for reliable
predictions. Thus it is usually advisable to design against cleavage initiation.

MEDIAN CLD

TOUCHNISS RATIO:
4 -0-3.0

81.1 0.0

0 20 40 60 s0 100
PERCENT LBZ

FIG. 3.22 Effect of the percent local brittle zone (LBZ) along the Crack front on median
fracture toughness in the heat affected zone. The toughnoe ratio Is
defined ms the matrix toughnesalLsZ toughness.
12

6 I * I * I I I *

% TOUGHNESS RATIO - 3.0

-4-F O.9

" . . . . . . . . . 13 --------- . . . . . . . . . . . . .
.
0 * I * I , I , I , I * I , I *,

0 20 40 60 80 100
PERCENT LBZ

FIG. 3.23 Predicted scatter In CTOD data for weldments with a. matrx/LBZ toughness
ratio of 3.0 (44).

3.5.2 Large Scale Yielding

The J integral and crack tip opening displacement approaches allow significantly
more plasticity than linear elastic fracture parameters such as stress intensity, but
elastic-plastic fracture theory eventually breaks down. In such cases, J does not
uniquely characterize crack tip conditions (Fig. 2.16c). The apparent fracture
toughness depends on the size and geometry of the specimen or structure. This
effect is most pronounced in the upper transition region, where cleavage is preceded
by large scale yielding and ductile crack growth.

Bending Versus Tension

Figures 3.24 and 3.25 show crack tip stress fields computed from finite element
analysis (70) for a center cracked tension panel and an SENB specimen, respectively.
The results are plotted so that all data collapse onto a single curve when J integral
theory is valid. In the case of the center cracked panel, the computed stresses in large
scale yielding lie well below the HRR solution. The stresses in the
 ayyy OH
C7 Obo1-0
3

1 10 100

FIG. 3.24 Finite element solution for crack tip stress fields In a center cracked panel
loaded In plane strain tenslon(70).

4
BENID SPECIMEN, n=10

.w

FIG. 3.25 Finite element solution for crack tip stress fields In a plane strain SENS
speclmen(70).
74
bend specimen are close to the HRR solution near the crack tip but diverge at greater
distances. Since the stresses at the crack tip control fracture, J is a good characterizing
parameter in bend specimens up to relatively high levels of plasticity. The highest J
value plotted in Figs. 324 and 3.25 is slightly less than the value corresponding to the
ASTM size limit for JIc (Eq. [3.10]). It is apparent from these two figures why the
ASTM size requirements apply only to specimens loaded predominantly in bending.

According to Figs. 3.24 and 3.25, bending produces more constraint than tensile
loading. Because of this constraint effect, the apparent cleavage toughness is higher
in specimens and structures loaded predominantly in tension. Figure 3.26 shows
ClOD data for SENB specimens as well as wide plates with surface cracks and
through- thickness notches (71). The material is a BS 4360-50D steel tested at -65 ° C
(72). Although the scatter bands overlap, the critical CTOD in the wide plates tends
to be significantly higher than in the bend specimens.
Scatter in Toughness
Loss of crack tip constraint not only increases the average toughness, but also
increases the relative scatter (44). This is evident in Fig. 3.26, where the wide plate
data are more scattered than the SENB data. Under large scale yielding conditions,
scatter in fracture toughness data can no longer be described by Eq. [3.19]. This
relationship was derived assuming HRR conditions (see Appendix). When crack tip
constraint relaxes, an incremental increase in J (or CTOD) produces a smaller
increase in stress than predicted by HRR theory. Large increases in J have relatively
little effect on the actual driving force at the crack tip. The failure probability also
increases more gradually with increasing J. Thus critical values of J or CTOD are
widely scattered.

Prior stable crack growth also tends to increase the level of scatter in cleavage
toughness (44). As the crack grows, the crack tip stress field deviates from HRR
theory. The effect on failure probability and the fracture toughness distribution is
similar to that associated with constraint loss at a stationary crack. In a typical
structural steel in the upper transition region, constraint loss and stable crack growth
combine to produce a high degree of scatter in fracture toughness. Wallin (73, 74)
characterized upper transition behavior with a modified statistical model for
cleavage.
Figure 3.27 is a schematic Weibull plot of fracture toughness data in the ductile-
brittle transition region, with distributions for two specimen sizes. At the lower end
of the distribution, Eq. [3.19] applies and critical J or CTOD data have a slope of 2.0 on
a Weibull graph. At higher toughness values, the slope decreases because a
combination of large scale plasticity and ductile crack growth. A low Weibull slope
indicates a high degree of scatter. Under J controlled conditions when the Weibull
slope a 2.0, the distributions for the two thickness can be collapsed onto a single
curve by multiplying the data for the smaller specimens by \B2/B• This
multiplying factor does not correct for thickness at large J values.
 75

2.0
BS 4360 50D)L
1.8 steel
-651"C
1.6 B= 52mm

r1.4 A

1.2
1.0 C
aB
0.6 I
0.4 Median o
of a a
0.2 74 results

Fullthickness C wide Surface notched

SEN specimens plates wide plates
FIG. 3.26 Comparison of critical CTOD values for SENB specimens with data for center
crack tension and surfaced notched tension wide plate specimens (71).

!zo

B! B2

LN Je)or LNI&)
FIG. 3.27 Schematic Welbull plot of fracture toughness data for two specimen
thicknesses.
F6

Constraint loss and crack growth produce a thickness dependence in addition to that
associated with statistical sampling effects.

Size Effects
Size effects on fracture toughness in the transition region can be caused by a
combination of constraint and statistical sampling effects. Figure 3.28 is a schematic
plot of median fracture toughness versus specimen size. If the specimen is large
enough to maintain J-controlled conditions, the size dependence of toughness is due
only to statistical effects. The effect of thickness can be inferred from Eq. [3.191. For
smaller sizes, the loss in crack tip constraint produces an additional size dependence,
as indicated by the shaded area in Fig. 3.28.

The statistical size dependence on fracture toughness cannot be avoided, but it has
been quantified (Eq. [3.19]). Thus it is relatively easy to correct for statistical effects
when analyzing a flaw in a structure. Since there are currently no proven
corrections for crack tip constraint, J and CTOD tests in the transition region should
be performed on sufficiently large specimens to ensure J controlled conditions at the
crack tip. The minimum size requirements for cleavage toughness have not been
established. The development of such criteria should be the subject of future
research.

Corrections for Crack Tip Constraint

In certain instances, it may not be practical to perform a fracture toughness test under
J controlled conditions. For example, an engineer may want to test a specimen with
a shallow crack in order to simulate a surface flaw in a structure. In such cases, the
resulting fracture toughness will be influenced by constraint loss. Although there
are no proven methods to correct for constraint loss, a recently developed analysis
shows some promise. Anderson(61) derived a J-like parameter to characterize the
stress normal to the crack plane (oyy) under large scale yielding conditions:

Jyyo'' (3.2,,1

Recall that dn is a dimensionless constant that relates J and CTOD under HRR
conditions (Eq. [2.26]). Equation [3.271 characterizes the loss in triaxiality through the
change in the J-CTOD relationship from the HRR solution. When the HRR solution
is valid, Jyy" = J. Figure 3.29 shows the finite element results from Fig. 3.24,
normalized in terms of Jyy" rather than J. This new parameter appears to correct the
stress fields for constraint loss. Since oyy near the crack tip governs cleavage, Jyy"
can be viewed as the effective driving force for cleavage. Further validation of this
parameter is needed before it can be applied to practical situations.
 77

NWLAN[ controUed
FRCflRE jCfbcu511
TOUGFNESS

SPECIMErJ SIZE

FIG. 3.28 Schematic plot of the effect of specimen size on ffacture toughness.

4 i i I
I F iI I
i i i

CENTER CRAOCED PANE , lo

COo 0 HRR
0 bMoi7-n2O
X bCo/1-3D

3x

2 x x

1x

1 10 100 1000

FIG. 3.29 Crack tip stress fields from Fig. 3.24, normalized by Jyy*(61).
3.6 DYNAMIC FRACTURE TOUGHNESS AND CRACK ARREST

When fracture is associated with either a high loading rate or a rapidly propagating
crack, the problem is complicated by two factors: the inertia of the structure and the
strain rate sensitivity of the material. With inertia effects, calculation of crack tip
parameters such as J and K is much more difficult (75). When the strain rate varies
over several orders of magnitude, metols behave in a viscoplastic manner, even at
low temperatures. A material's flow stress increases with strain rate; thus quasistatic
tensile properties are not relevant at high strain rates.

The relationship between loading rate and fracture toughness is complex. Inertia
effects tend to increase the apparent toughness of the -materialbecause the total
fracture energy includes., zigidficait kinetic energy component. Viscoplastic
material behavior can either inci-ease or decrease the toughness with loading rate,
depending on the fracture mechanism. Cleavage fracture toughness tends to
decrease with strain rate because cleavage is stress controlled. Increasing the flow
stress makes it easier for material at the cyack tip to reach the critical cleavage stress.
Since microvoid coalescence is predominantly strain controlled, upper shelf
toughness usually increases wiih strain rate. As the flow stress increases, more
energy is required to deform the material near the crack tip to a critical strain.
During rapid propagation of a crack in a large specimen or structure, the crack speed
(and hence the local strain rate) can vary considerably. In such cases it is not unusual
for the fracture mechanism to aiernate between cleavage and ductile tearing, as one
mechanism becomes more favorable than the other (76).

Inertia effects are most pronowic iiA1m,.LiatC.y after application of the load (77).
The material oscillates as sues, 1. i,, at through the specimen, and the
amplitude of these oscillations c;,-.reair rapidly. After a short time, the local inertia
effects are small and the behavior ii don-ti, ated by global inertia effects. At long
times, the behavior of the specimn is sirailar to that of the quasistatic case, with the
exception of viscoplastic material ;1ow priperties. Nakamura et al. (77) developed
the concept of a transition time, t':, to delineate between short time response, where
inertia effects are important, and iong timie behavior, where inertia effects are small
relative to the total energy in the specimen. Dynamic finite element analysis
performed by Nakamura et aL. ii: icat.,s ihat inertia effects are small at times greater
that 2tr. As long as fracture initiation occurs after 2 tT, the dynamic JIC can be
computed from the quasistva~ic i:aions-Kps in ASTM E813-87..

Joyce and Hacket (78, 79) have a:iK '.' :r,..sition time concept to dynamic J
testing. They have perfected ,xpx..ri:i tehniques for high rate fracture
toughness tests on SENB spf-%d.2_is. Th se specimens are typically 25 mm thick and
are heavily instrumented. Lnpt lc~idi. is ipplied by a drop tower. Nakamura et
al. estimated the transition iiii. hr
r,-ir p- cimens to be 300 lIs. Materials that
Joyce and Hacket have tested 7'az>.i' ely ductile. Fracture initiation in these
materials usually occurs be, 0 c; an d Vivo milliseconds after the tup contacts the
specimen. Since the time to f:'r.cvre g-eater than 2 tT in this case, J can be
 79

computed with Eqs. [3.5] to [3.7a]. Joyce and Hacket developed a key curve method to
Infer crack growth during the test. Figure 3.30 (78) compares quasistatic J-R curves for
a 3-Ni steel with dynamic curves obtained front the their method. The viscoplastic
material behavior causes the material resistance to increase with strain rate, as
exlpeced.

When inertia effects are minimal, dynair..- recl,.ccs de;w~age fracture

toughness. Since plastic defonnation if . ,",.miy
'. ariib'ated process,
increasing the strain rate is eui' .znt i - tf-,i.rature. Thus, when
the loading rate increases, thz ducilbrinK: . . c r~c shits upward in
temperature. For example, Fig. 3.31 shows fracture toughness data for an A572 Grade
50 steel at three loading rates (8M. The tansition irves shift approximately 2000F
(110C) between the lowest and highest E'rain rate.

Once cleavage initiates, whether from quasistatic or dynamic loading, a propagating

cleavage crack moves rapidly through the specimen. Material near this propagating
crack is subjected to very high strain rates. Thus the situation is analogous to a
continuous dynamic fracture toughness test. In order for the crack propagation to
continue, the driving force for fracture must be at least as large as the material
resistance. Resistance of a steel to cleavage fracture can be represented by a falling R
curve, as illustrated in Fig. 3.32. When fracture initiates at KIC, the crack propagates
in a rapid and unstable manner. The fracture process then becomes dynamic, and
resistance to cleavage decreases. In many loading situations, the driving force
increases with crack length or remains constant. and the crack continues to
propagate. In some cases, however, the driving force may decrease until it intersects
the material resistance curve, and the crack arrests because further propagation is
impossible. The arrest toughness, Kia, is a measure of a material's ability to stop a
rapidly propagating crack.

There a number of practical situations in which crack arrest can occur. A specimen
or structure loaded in displacement control can have a falling driving force curve. A
falling driving force curve can also result from a ra-nsient load, as when a ship
collides with another vessel. The transient load may be sufficient to initiate
cleavage, but the driving force decrease-s r .vid v a- the stress waves are damped by
the structure. Another common scenario for crack arrest involves a toughness
gradient. For example, a crack may initiate i :t mld region of a structure, where the
toughness is low, and propagate into rmateri - : higher temperature and toughness
level. The crack arrests when the Ki e,.-, driving force. A toughness
gradient could also result from variatior. in microstructure, such as in welded
structures. If a cleavage crack iniftater ir *- '-I on of the weld, it may arrest
when it reaches the parent metal, which .uIuMly has higher toughness.
80

70

DRO IO H IU

CRACK IXTEINM 1ml

FIG. 3.30 Effect of loading rate on J-R curves for a 3-NI steel (78).

AM GRADE 80 STEEL ley -SO bi

* SLOW KNO LOAD O-'I

ai 10
8 *NTURMIDSATS SrTRAIN-RATE LOAD ide 10 3-1
A DThIAWC LOAD Q1a 6 =-I)

MDT

PIG. 3.31 Cleavage fracture toughness at three strain rates for AM7 Grade 50 stee (S0).
 81

CRACK LENGTH
FIG. 3.32 Schematic driving force/resistance curve behavior for cleavage fracture

*sc&

3.3 uecken
FiGTt ppaau o ~ et
kd codn oAT l21O S)
Because several industries, including nuclear power and natural gas, have become
increasingly interested in designing structures to arrest rapidly propagating cracks, a
standardized method to quantify arrest toughness, KIa, was needed. In 1988, ASTM
published E1221 (81), a standard test method for measuring Kla. The test method
utilizes a compact crack arrest specimen that slightly resembles the ordinary compact
fracture toughness specimen. Figure 3.33 illustrates the crack arrest testing
procedure. The specimen is laid on its side and a wedge is driven though a hole in
the specimen until fracture initiates. This produces nearly pure displacement-
controlled loading, which results in a falling driving force curve. The Kia can be
inferred from the final crack length after arrest.

3.7 COMPARISONS BETWEEN J AND CTOD AS TOUGHNESS

PARAMETERS
The CTOD approach developed in Great Britain has been applied primarily to welded
steel structures in the transition region. One of the major driving forces for CTOD
technology was the development of oil reserves in the North Sea. Most of the early
work on J integral testing and analysis was funded by the nuclear power industry in
the United States. Since typical service temperatures for reactor pressure vessels are
well on the upper shelf, where fracture occurs by ductile tearing, most research in J
integral technology has focused on upper shelf behavior.
Because of the history of these two toughness parameters, J is normally viewed as a
ductile fractwre parameter, while CTOD is usually associated with the transition
region and weldments. However, there are no sound technical reasons for not using
J and CTOD for all types of fracture behavior in steels. Both are equally valid elastic-
plastic measures of fracture toughness.

Under small scale yielding conditions, there is a unique relationship between J and
CTOD for a given material. Shih (22) derived this relationship (Eq. [2.26]) from the
HRR crack tip solution. Thus when the assumptions of the HRR analysis are valid, J
and CTOD are equivalent measures of toughness.

When fracture is preceded by large scale yielding, the J-CTOD relationship depends
on the geometry of the specimen or structure. Equation [2.26] implies a constant ratio
between J and CTOD, but under fully plastic conditions this ratio varies with
displacement. Experimental estimates of plastic CTOD are proportional to
displacement (Eq. [3.14]); Jp is proportional to the plastic area under the load-
displacement curve.
Figures 3.34 and 3.35 show the effect of plastic deformation on the J-CTOD
relationship in a center cracked panel and a bend specimen, respectively. Each of
these plots compares the small scale yielding solution (dn) with a fully plastic
solution and an elastic plastic finite element solution. The fully plastic J-CTOD
relationships (61) were derived from remote loads and displacements, obtained from
 83

%\ CENTER CRACKCED PANEL, U010

0.9
800 -U-EA CSHDO
-U-FUY PFLA5TJ

0.7

0.4

0 0L01 0.02 0.0 0.0 0.05

b co
FIG. 3.34 The relationship between J and CTOD for a center cracked panel In plane
atraln(22,61).

BEND SPEO[MEN, a-10

0.9

J -aFEA (SHIM)
-- FUY LASTIC

0.7-

0.5

0 0.01 0a0 0.m 00 LOS

FIG. 3.35 The J-CTOD relationship an SENS specimen In plane strain (22,61).
44

the Electric Power Research Institute (EPR elastic-plastic fracture handbook (82).
The elastic-plastic finite element solutions were performed by Shih (22). In the case
of the center cracked panel (Fig. 3.34), the elastic-plastic J-CTOD relationship deviates
significantly from the small scale yielding value. Note that the finite element
solution and the fully plastic estimate tend to converge at high J values. The elastic-
plastic J-CTOD relationship for the bend specimen is very close to dn; the fully plastic
relationship coincides with dn at moderately high J values. Thus the small scale
yielding solution for the J-CTOD relationship is fairly accurate for bend specimens
well into the plastic range; the small scale yielding solution does not apply to tensile
loading, except at very low J values.

The above observations are consistent with the stress fields for bending-Aid tension
(Figs 3.24 and 3.25). The crack tip stresses in bending are close to the HRR solution at
high J values, but the center cracked panel cannot maintain sufficient constraint with
plastic deformation. A loss of constraint appears to be accompanied by a deviation of
the J-CTOD relationship from Eq. [2.26]. Equation [3.271 implies that constraint loss
is directly related to the J-CTOD relationship.

Neither J nor CTOD appears to have a clear advantage over the other parameter as a
measure of fracture toughness. When the assumptions of the HRR theory are
approximately valid, the two parameters are uniquely related, and either can
characterize crack tip conditions. When the HRR theory breaks down, neither
parameter characterizes crack tip conditions; both critical J and critical CTOD values
depend on the configuration of the specimen or structure in such cases. Equation
[3.27] and Fig. 3.29 indicate that simultaneous measurement of both parameters may
be needed to characterize the critical conditions for fracture in large scale yielding.

3.8 CHARPY-FRACTURE TOUGHNESS RELATIONSHIPS

Fracture toughness testing is a relatively new phenomenon. In many industries, the

Charpy V-notch (CVN) test is still the only accepted method for quantifying a
material's toughness. Most design codes for welded steel structures do not address
fracture mechanics methodology. Typically, these codes specify a minimum CVN
energy for the base material and strict flaw tolerances for the welds. Both the CVN
and flaw acceptance criteria are usually arbitrary and often very conservative. In
some cases, however, fracture control procedures based on CVN toughness can be
nonconservative. For example, a large number of steel bridges which were designed
with such a procedure have sustained brittle fractures over the last 15 years (83).

Since the Charpy test did not become obsolete with the advent of fracture mechanics,
there have been many attempts to correlate CVN toughness measurements with
fracture mechanics parameters such as KIC. The problem with such correlations is
that the CVN test is very different from fracture mechanics tests. The Charpy
specimen contains a blunt notch, while fracture toughness tests are performed on
specimens with sharp fatigue cracks. The Charpy test is performed in impact loading,
 85

but most fracture toughness tests are performed quasistatically. Another difficulty
with the Charpy test is that the specimen is small. Recall that fracture mechanics
tests must be performed on sufficiently large specimens to guarantee a single
parameter description of crack tip conditions. When the specimen is too small,
toughness is size dependent; a small specimen may indicate ductile behavior, but a
large specimen tested at the same temperature may be brittle.

Most CVN-fracture correlations published to date are empirical. That is, the
investigators measured both CVN and KIC, and then curve-fit the results to produce
a mathematical relationship between the two parameters. Because of the differences
between CVN and fracture mechanics tests, each of these empirical correlations
usually works well only for the naterial originally used to generate the correlation.
None of these correlations can be successfully applied to all steels (84).

Recently, some investigators have attempted to develop theoretical relationships

between CVN and fracture toughness. These approaches address the differences
between the two types of tests.

3.8.1 Empirical Correlations

The various empirical correlations can produce widely differing estimates of KIC
when applied to the same CVN data. These approaches differ in three ways:

1. The type of steel from which the correlation was generated

Z The way in which the data were fit; i.e., the form of the equation

3. The fracture behavior analyzed; e.g., transition region or upper shelf.

Some of the more common empirical correlations are summarized below.

Mandaret and Sanz (85) applied a transition temperature approach to data for C-Mn
and low alloy steels. They defined the beginning of the transition at 28 J and 100 MWa
4i-for the CVN and KIC data, respectively. They correlated transition temperature
with the following expression.

T (KIC 100) - 9 + 1.37 T (CVN2s) [328]

where T is the temperature in *C. Assuming thac the KIC and CVN transition curves
had a similar shape, they estimated Kic at other temperatures from the following
relationship:

KIC - 194NrCVN [3.291

16

where KIC is in MPa 4- and CVN is in J. Note that one must apply the
temperature shift (Eq. [3.28]) when using the above expression.

Rolfe and Novak (86) developed the following correlations for upper shelf toughness
in steels:

(KIC) 2 (CVN S) (3.301

where KiC is in ksi , CVN is in ft-lb, and ys isin ksi. The good correlation with
Rolfe and Novak's experimental data was attributed to upper shelf fracture not being
as sensitive to notch acuity as cleavage. An addition explanation is that upper shelf
toughness data are not nearly as scattered as data in the transition region. Pisarski
(84) pointed out that much of Rolfe and Novak's data did not meet the KIC size
requirements and were therefore not valid.

Rolfe and Barsom (87) developed the following correlation for the transition region.

KIC= 2 [(CVN)3/2
331j
E

where KIC is in psi "fiW, CVN is in ft-lb, and E is in psi. The data displayed a
considerable degree of scatter about the proposed correlation. The poor correlation is
probably due to differences in strain rate and notch acuity. However, even a perfect
correlation would not eliminate all data scatter in the transition region because it is
an inherent feature of the material.

Barsom (80) attempted to reduce the scatter in CVN-fracture toughness correlations

by correcting for dynamic effects. He developed a two step correlation for the
transition region. First, he correlated CVN data with dynamic fracture toughness
(KId) tests. He then determined an empirical temperature shift between KIC and KId
data. Figure 3.31 shows some of the data for the latter correlation. Barsom fit KId
and CVN data for a number of steels to the following equation:

KId 2
E -A(CVN) [332]

where A is a constant The units on KId, CVN, and E are the same as in Eq. [3.31].
For standard CVN specimens, A is approximately 5. This constant is approximately 4
if the CVN specimen contains a fatigue precrack. The empirical temperature shift
between KIc and KId data is given by

Tshift = 215 -Oys (fory <140

9 ksi) (333]
 87

where temperature is in OF and oys is in ksi. Thus one must first estimate a KId
value at a given temperature from Eq. [3.321, and then apply the temperature shift
(Eq. [3.33]) in order to convert this value into a KIC estimate.

There are several other CVN-KIc correlations including those by Sailors and Corten
(88), Begley and Logsdon (89), and Ito et aL (90). Pisarski (84) reviewed a large
number of correlations and applied them to a range of experimental data. He
concluded that most correlations work well for high strength steels but tend to be
conservative for low and medium strength steels. Wellman and Rolfe (91) also
applied several published correlations to data for various steels; the empirical
correlations worked well on the lower shelf but were excessively conservative in the
transition region. The correlations failed to predict the rapid increase in toughness at
the onset of the ductile-brittle transition.

Observations by Pisarski (84) and Wellman and Rolfe (91) imply that the empirical
CVN-KIC correlations apply only to linear elastic behavior, associated with the lower
shelf or high strength steels; none work in the elastic-plastic regime.

Dolby (92) attempted to develop an elastic-plastic CVN-fracture toughness correlation

for weld metals, relating the 0.1mm CTOD transition temperature to a corresponding
CVN transition temperature. He concluded that acceptable correlations could be
found only when the materials were sufficiently delineated according to welding
process, consumables, etc. No universal CVN-CTOD correlation was found.

3.8.2 Theoretical CVN-Fracture Toughness Relationships

Some investigators have taken a more fundamental approach to CVN-fracture

toughness relationships. Norris et al. (93) developed an elastic-plastic, dynamic finite
element model of the Charpy test and applied a local ductile failure criterion to both
the Charpy specimen and a compact specimen to correlate upper shelf CVN energy
with Jic.

A somewhat similar approach by Anderson and Zapata (94) for the lower shelf and
transition region used a statistical model developed by Anderson (43) to quantify the
inherent microstructural fracture properties of a given material. This information,
in conjunction with Norris' finite element solution for the CVN specimen and
sharp crack solutions by Shih and German (70), relates macroscopic fracture
properties of CVN and fracture toughness specimens. Since this model is statistical,
it predicts not only average toughness, but also any level of lower bound.

Figure 3.36 (94) shows a sample comparison between experimental fracture

toughness and toughness predicted from CVN data. The material was a 25 mm thick
plate of ABS Grade EH36 steel. Experimental CTOD data are plotted with predicted
median and 10% lower bound curves. The agreement between theory and
experiment in the transition region is reasonably good, although the predictions are
slightly conservative.
88

Anderson (95) also developed a theoretical relationship for the upper shelf that is
different from the Norris approach in attempting to characterize the slope of the R
curve rather than JIc. The.R curve slope is estimated by considering the tearing
resistance of a CVN specimen and relating it to initiation and pkopagation energy.
The resulting expression for the initial R curve slope is as follows:

dj 4CVN ( 1 334]

where B and b-are the thickness and ligament length in the CVN specimen,
respectively; f is the fraction of the total energy required for propagation; fA (= 1 -fp) is
the initiation traction; and n is the strain hardening exponent, as defined in Eq. [2.22].
Predictions from Eq. [3.341 are compared in Fig. 3.37 with experimental data from a
variety of steels published by Garwood (96). For the predictions in Fig. 3.37, fp was
assumed to be 0.75 and n was assumed to equal 10. The agreement between theory
and experiment is very good except for two data points at high CVN energy levels.

The theoretical CVN-fracture toughness relationships for both the upper shelf and
the transition region should be investigated further. Although early results are
encouraging, additional validation and refinement is needed to make such
approaches practical.

One practical use of reliable correlations will be the evaluation of the structural
relevance of existing CVN toughness criteria. If more quantitative measures of
fracture toughness can be inferred from CVN values, it will be possible to estimate
critical flaw sizes and allowable design stresses from the appropriate driving force
relationships.
 89

0.3

IOL
EE3

0.0 z i i' , q W&MIIor a=

I

100 *GMMN

FIG. 3.36 Comparison of experimental CI"OD data In the transition region with curve predicted
from CVN awa
data (94). The material Is a 25 mm thick plate of ASS Graft EN.36 steel

100

200 -

0 128 192 26 320

CVN, j
FIG. 3.37 Comparison of predicted and experimental R curve Mlope for various steels (911),
90

4. APPLICATION TO STRUCTURES

A crack will grow when the driving force for fracture exceeds the material's
resistance to fracture. Chapter 3 addressed the material resistance side of the
equation. This chapter focuses on the driving force.

4.1 HISTORICAL BACKGROUND

Fracture mechanics design analyses and driving force equations have evolved
considerably over the last 30 years. The first analyses were based on linear elastic
theory and applied only to simple configurations such as a center cracked panel.
Now there are approaches which take account of most complications which arise in
real structures, such as complex shapes, residual stress and large scale plasticity.

4.1.1 Linear El-tlc Fracture Mechanics (LEFM)

The LEFM approach to design can be summarized by restating Eq. [2.6]:

KI = Ycf [41]

where Y is a geometry correction factor, o is a characteristic stress, and a is a

characteristic crack dimension. The above relationship describes the driving force for
fracture. The material fails when KI > KIC, the material's resistance to fracture.

Since Irwin (5) and Williams (6) first applied the Westergaard (4) analysis to practical
problems in 1957, most of the developments in LEFM have involved the generation
of stress intensity solutions (Y) for specific configurations. Most of the early analyses
were theoretical or experimental. Today, however, stress intensity solutions are
usually determined from finite element analysis.

Several handbooks (8-10) contain compilations of stress intensity factors. The most
recent (and most complete) handbook (10) contains K solutions for several hundred
configurations. It is now possible to find published K solutions that closely match
nearly any practical situation.

A number of variations to the LEFM approach attempt to correct for plasticity, as

discussed in Chapter 2. The Irwin plastic zone correction, where crack size is replaced
by an effective crack length, extends LEFM to approximately 80% of the yield load (see
Fig. 2.9). The Dugdale-Barenblatt strip yield model (12,13) enables an effective K to be
computed up to limit load in nonhardening materials.
 91

The development of J integral test methods enables a linear elastic driving force
relationship to be applied in situations where it is not possible to measure valid KIC
values in laboratory tests. As discussed in Section 2.3.7, the J integral can often
characterize the crack tip conditions in a small specimen, even though that specimen
contains a large plastic zone. A small specimen and large structure loaded to the
same J value experience the same crack tip environment, as illustrated in Fig. 2.18.
If the structure is sufficiently large that the global behavior is elastic, the crack tip
conditions can be characterized by either J or K Thus it is possible to estimate the
KIC of the large structure from the JIC of the small specimen:

KIC C [42
Furthermore, one can use linear elastic driving force relationships (or LEFM with a

plastic zone correction) to estimate critical crack size or failure stress in the structure.

4.1.2 The CTOD Design Curve

The CTOD concept was applied to structural steels beginning in the late 1960s. The
British Welding Research Institute (now known as The Welding Institute) and other
laboratories performed CTOD tests on structural steels and welds. At that time there
was no way to apply these results to welded structures because CTOD driving force
equations did not exist Burdekin and Stone (18) developed the CTOD equivalent of
the strip yield model in 1966. Although their model provides a basis for a CTOD
driving force relationship, they were unable to modify the strip yield model to
account for residual stresses and stress concentrations. (These difficulties were later
overcome when a strip yield approach became the basis of the R6 design method.)

In 1971, Burdekin and Dawes (97) developed the CTOD design curve, a semi-
empirical driving force relationship, that utilized fracture mechanics theory, where
available, and relied on empirical correlations and conservative assumptions for
circumstances theory did not address. For linear elastic conditions, fracture
mechanics theory was reasonably well developed, but the theoretical framework
required to estimate the driving force under elastic-plastic and fully plastic
conditions did not exist until the late 1970s. Burdekin and Dawes based their elastic-
plastic driving force relationship on an empirical correlation between small scale
CTOD tests and wide center cracked tension panels made from the same material.
The wide plate specimens were loaded to failure, and the failure strain and crack size
of a given large scale specimen were correlated with the critical CTOD in the
corresponding small scale test.

The correlation that resulted in the CTOD design curve is illustrated schematically in
Fig. 4.1. The critical CTOD is nondimensionalized by the half crack length, a, of the
wide plate and is shown on the ordinate of the graph. The nondimensional CTOD is
plotted against the failure strain in the wide plate, normalized by the elastic yield
92
strain, ey. Based on a plot similar to Fig. 4.1, Burdekin and Dawes proposed the
following two-part relationship:

0 = t -If2 .for- f O.5 [4.3a]

2xeya (ey) ey

and

= 2xefi a e - 0.25 for L > 0.5 [43b]

eya
eyey

where 0 is the nondimensional CTOD. Equation [4.3aJ, which was derived from
LEFM theory, includes a safety factor of 2.0 on crack size. Equation (43b] represents
an upper envelope of the the experimental data.

1.5

8
crit
27tcy a 1.
1.0 -

0.5 SAFE

USAR 1d 2.0
0 0.5 n LO

FIG. 4.1 The semi-ompirical CTOD design curve.

 93

The applied strain and flaw size in a structure, along with the critical CTOD for the
material, can be plotted on Fig. 4.1. If the point lies above the design curve, the
structure is considered safe because all observed failures are below the design line.
Equations [4.3 a-b] conform to the cassical view of a fracture mechanics analysis, in
relating stress (or strain in this case) to fracture toughness (:rit) and flaw size (a).
The CTOD design curve is conservative, however, and does not relate critical
combinations of these variables.

The CTOD design curve approach was modified in the mid 1970s to include stress
concentration effects, residual stresses, and flaws other than through thickness. In
1980, this approach was incorporated into the British Standards document PD6493
(98).

The PD6493:1980 approach addresses flaws of various shapes by relating them back to
an equivalent through-thickness dimension, I. For example, if a structure contains
a surface flaw of length I and depth t, the equivalent through-thickness flaw
produces the same stress intensity when loaded to the same stress as the structure
with the surface flaw. Thus &is a generalized measure of a flaw's severity. The
CTOD design curve can be applied to any flaw by replacing a with I in Eq. [4.3].

The original CTOD design curve was based on correlations with flat plates loaded in
tension. Real structures, however, often include complex shapes that result in stress
concentrations. In addition the structure may be subject to bending and residual
stresses, as well as tensile (membrane) stresses. The PD6493:1980 approach accounts
for complex stress distributions simply and conservatively by estimating the
maximum total strain in the cross section and assuming that this strain acts through
the entire cross section. The maximum strain can be estimated from the following
equation:
1k
El = '[kt(Pm + Pb) + Q] [4.4]

where kt is the elastic stress concentration factor, Pm is the primary membrane stress,
Pb is the primary bending stress, and Q is the secondary stress, which may include
thermal or residual stresses. Since the precise distribution of residual stresses is
usually unknown, Q is often assumed to equal the yield strength in an as-welded
weldment.

When Burdekin and Dawes developed the CTOD design curve, the CTOD and wide
plate data were limited; the curve they constructed lay above all available data. In
1979, Kamath (99) reassessed the design curve approach with additional wide plate
and CTOD data generated between 1971 and 1979. His analysis was essentially
identical to that eventually adopted in PD6493:1980. In most cases, there were three
CTOD tests for a given condition. Kamath used the lowest measured CTOD value to
predict failure in the corresponding wide plate specimen. When he plotted the
94

results in the form of Fig. 4.1, a few data points lay below the design curve, indicating
Eq. [4.3] was nonconservative in these instances. The CTOD design curve, however,
was conservative in most cases. Kamath estimated the average safety factor on crack
size to be 1.9, although individual safety factors ranged from less than 1 to greater
than 10. With this much scatter, the concept of a safety factor is of little value. A
much more meaningfu quantity is the confidence level. Kamath estimated that the
CTOD design curve method corresponds to a 97.5% confidence of survival. That is,
the method in PD6493:1980 is conservative approximately 97.5% of the time.
4.1.3 The R-6 Failure Assessment Diagram
Structures made from materials with sufficient toughness may not be susceptible to
brittle fracture, but they can fail by plastic collapse if they are overloaded. The
PD6493:1980 approach does not explicitly address plastic collapse, but implicitly seeks
to avoid collapse through an arduous flaw recategorization procedure.

The Central Electricity Generating Board (CEGB) in Great Britain developed a design
methodology that takes account of fracture and collapse, as well as the interaction
between these two failure mechanisms. The CEGB approach, known as the R-6
method (100), is based on a modified strip yield model due to Heald et al. (101). The
R-6 method was first published in 1976, with minor revisions in 1977 and 1980. (A
third, more radical revision published recently is discussed in Section 4.2)

The R-6 method introduced the concept of a failure assessment diagram (FAD) to
describe the interaction between fracture and collapse. The modified strip yield
model and the failure assessment diagram that form the basis of the early versions of
the R-6 procedure are described below.

The effective stress intensity for a through-thickness crack in plane stress, according
to the Burdekin and Stone (18) strip yield model, is given by

sefc2 2 [45]

which is a restatement of Eq. (2.16]. As shown in Fig. 2.9, the above relationship is
asymptotic to the yield strength. This expression can be modified for real structures
by replacing ays with the collapse stress, ac for the structure. This would ensure that
the strip yield model predicts failure as the applied stress approaches the collapse
stress. For a structure loaded in tension, collapse occurs when the stress on the net
cross section reaches the flow stress of the material. Thus Gc depends on the tensile
properties of the material and the flaw size relative to the total cross section of the
structure. The next step in deriving a failure assessment diagram from the strip yield
model entails dividing the effective stress intensity by the linear elastic K:
 95

Keff Ors
[7in SC(I (y 1/2
Kj 02 2e FL6IK

This modification not only expresses the driving force in a dimensionless form but
also eliminates the square root term that contains the the half crack length of the
through crack. Thus Eq. [4.61 is more general, in that it applies to all configurations,
and not just a through-thickness crack in an infinite plate. This is analogous to the
PD6493 approach, where the driving force relationship was generalized by defining
an equivalent through thickness flaw, L As a final step, we can define the stress
ratio, Sr, and the K ratio, Kr, as follows:
Kr Y [4.7]

and

O
Sr L8
The failure assessment diagram is then obtained by inserting the above definitions
into Eq. [4.5] and taking the reciprocal:

Kr a 5r in S r 1/ [49]

Equation [4.9] is plotted in Fig. 4.2 The curve represents the locus of predicted failure
points. Fracture is predicted when Kef = KIC. If the toughness is very large, the
structure fails by collapse when Sr a 1.0. A brittle material will fail when Kr = 1.0. In
intermediate cases, collapse and fracture interact, and both Kr and Sr are less than 1.0
at failure. All points inside of the FAD are considered safe; points outside of the
diagram are unsafe.

In order to assess the significance of a particular flaw in a structure, one must

determine the applied values of Kr and Sr and plot the point on Fig. 4.2 The stress
intensity ratio for the structure is given by
K1
Kr (sv0]e)i=KI [4L101
16

,, kus 2

Kr

0.2

0 - -

0 0.2 0.4 0.6 0.8 1 1.2

Sr
FIG. 4.2 The R-6 failure assessment diagram based on the strip yield model.

The applied stress ratio can be defined as the ratio of the applied stress to the collapse
stress. Alternatively, the applied S, can be defined in terms of axial forces or
moments. If the applied conditions in the structure place it inside of the FAD, the
structure is safe.
The 1980 R-6 document (100) offers practical advise on how to apply thie strip yield
FAD to real structures. For example, it recommends that stresses be taken into
account through a secondary stress intensity. The total stress intensity is obtained by
adding the primary and secondary components:

KI - KIP + Kis [4.11

Only the primary stresses are used to compute Sr because secondary stresses, by
definition, do not contribute to collapse. Note that KI is the LEFM stress intensity; it
does not include a plastic zone correction. Plasticity effects are taken into account
through the formulation of the failure assessment diagram (Eq. [4.9J).
The R-6 procedure recommends that the fracture toughness input be obtained
through testing the material according to ASTM E399 or the equivalent British
 97

Standard. When it is not possible to obtain a valid KIC value experimentally, one
can measure Jic in the material and convert this toughness to an equivalent KIc by
means of Eq. [4.2].

4.1.4 The EPRI J Estimation Procedure

The R-6 failure assessment diagram is based on a strip yield model. Since it assumes
elastic-perfectly plastic materiai behavior, it is conservative when applied to strain
hardening materials.

In 1976, Shih and Hutchinson (102) proposed a more advanced methodology for
computing the fracture driving force which takes account of strain hardening. Their
apoach, developed further and validated at the General Electric Corporation in
Schenectady, New York in the late 1970s and early 1980s, was published as an
engineering handbook by the Electric Power Research Institute (EPRI) in 1981 (82).

The EPRI procedure provides a means for computing the applied J integral under
elastic-plastic and fully plastic conditions. The elastic and plastic components of J are
computed separately and added to obtain the total J:

Jtot = Jel + Jpi [4.12]

The elastic J is equal to G, the energy release rate. The EPRI handbook states that Jel
should include a modified first order plastic zone correction, but this correction
typically affects the total J by no more than 5% (102). Since this plastic zone
correction has no theoretical basis and severely complicates the analysis, many
engineers omit the correction on Jel.

For the fully plastic J, Shih and Hutchinson assumed a power law stress-plastic strain
curve (Eq. [2.22]). The nonlinear elastic assumption on which J integral theory is
based implies proportional loading; local stresses must increase in proportion to
remotely applied forces and moments. Hence the HRR singularity (Eqs. [2.23J and
[2.24]) implies that JpI is proportional to p", where P is a remote load. Shih and
Hutchinson invoked proportional loading and dimensional analysis to show that Jpi
can be expressed in the following form:

Jpi - (xeo b hi(a/W, n) (Pi)1 [4.13]

where Po is a reference load and hl is a dimensionless constant that depends on the

configuration and the strain hardening exponent. The terms a, eo, and cro are
material flow properties, as defined in Eq. [2.22], and b is the uncracked ligament
length in the structure or test specimen. Shih and Hutchinson also derived
expressions for the plastic crack mouth opening displacement, Vp, and the plastic
load line displacement, Ap:
 VP - aceo a h2 (a/W, n)()" [41

ALp = aeo a h3(a/W, n) (Pno

where h2 and h 3 are dimensionless constants for the corresponding displacement

measurements. The EPRI handbook lists hi, h2 and h3 values for a variety of
configurations and strain hardening rates. These constants were determined by
finite element analysis.

Figure 4.3 is a plot of applied Jin a center cracked panel with a/W = 0.5 and n - 7.
Note that the elastic term dominates at low loads and the plastic term dominates at
high loads.
The elastic-plastic driving force estimated from the EPRI procedure can also be
expressed in terms of P failure assessment diagram, an idea first proposed by Bloom
(103) and Shih et al (104). The J ratio and stress ration are defined as follows.

r G [416a]

and

P
Sr M -o [4.16b

The equivalent Kr is equal to the square root of Jr.

Figure 4.4 shows the applied J for center cracked panels with a/W - 0.5 and 0.75,
plotted in terms of failure assessment diagrams. The R-6 diagram is included for
comparison. Note that the shape of the FAD changes when strain hardening is taken
into account. The R-6 diagram is conservative in this case because it assumes
collapse will occur when the net section stresses equal ao; a panel made from a strain
hardening material could withstand somewhat greater stresses. Figure 4.4 also
illustrates that failure assessment diagrams derived from the EPRI procedure are
geometry dependent, while the R-6 diagram is geometry independent. This makes
fracture analyses with the EPRI approach more complicated, because a different FAD
must be generated for each configuration analyzed.
 1lemw Ste,a 7a/

_ Il " -ad
I.~-TOW I
/I
/.1
J
4- f l/

0 06 -

PO

FIG. 4.3 Applied J In a center crack panel, as computed from the EPRI J estimation
procedure(82).

12 I I I I I I I I I I I

=lane Sum, 3-7

0.6

0.2

0
* 0.2 OA 0. 0. 1 1.2 1A4

Sr
Fig. 4.4 Failure assessment diagrams generated with the EPRI approach(82),
compared with the R-6 FAD.
100
Section 4.1.5 Recent Advances In Elastic-Plastic Analysis

The CTOD design curve was the first elastic-plastic fracture analysis, developed when
elastic-plastic fracture mechanics was in its infancy. The theoretical work of Rice and
others had not yet filtered down to eng;-eers. Burdekin and Dawes had to rely on
empirical correlations and conservativ, assumptions to develop a comprehensive
elastic-plastic fracture design approach. The CEGB R-6 method, which was
developed a few years later, was based on an oversimplified theory of crack tip
plasticity. The work of Shih and Hutchinson in 1976 was the first sound theoretical
framework for analyzing nonlinear fracture in structures. Their work became the
basis of the EPRI approach, published in 1981.

One problem with the EPRI procedure has been that it can only be applied to
relatively simple configurations. In addition, this method gives no guidance on the
treatment of secondary stresses. Recent advances have improved the EPRI approach
somewhat;, additional configuations have been analyzed (105-107) since the original
EPRI report was published, and Bloom (108) has proposed a failure assessment
diagram approach that utilizes EPRI J solutions and a secondary stress analysis
similar to that of the R6 procedure.

Both the R-6 method and PD6493 have been revised recently to incorporate some of
the theoretical work upon which the EPRI handbook is based. These new documents
combine the flexibility and wide applicability of previous versions with more sound
elastic-plastic driving force analyses.

Recent work by Ainsworth (109) allowed the two British procedures to be updated
while retaining their flexibility. Ainsworth's main contribution was to simplify the
EPRI approach without sacrificing accuracy. The basis of Ainsworth's reference stress
model is described below.

4.2 THE REFERENCE STRESS APPROACH

The EFRI equation for fully plastic J, Eq. [4.13], assumes that the material's stress-
plastic strain curve follows a simple power law. Many materials, however, have
flow behavior that deviates considerably from a power law. For example, most low
carbon steels exhibit a plateau in the flow curve immediately after yielding.
Applying Eq. [4.131 to such a material, results in significant errors. Ainsworth (109)
modified Eq. [4.13] to reflect more closely the flow behavior of real materials. He
defined a reference stress as follows:

Gief=(P/Po) (;o [4.17]

He further defined the reference strain as the total axial strain when the material is
loaded to a uniaxia stress of aref. Substituting these definitions into Eq. [4.13] gives
 101

jpl = aref b hl (ref -2 "0) [4.18]

For materials that obey a power law, Eq. [4.18] agrees precisely with Eq. [4.13], but Eq.
[4.18] is more general, in that it is applicable to all types of stress-strain behavior.

Equation [4.18] still contains hi, the geometry factor which depends on the power law
hardening exponent n. Ainsworth proposed redefining Po for a given configuration
to produce another constant, hl', that is insensitive to n. He noticed, however, that
even without the modification of Po, hi was relatively insensitive to n except at high
n values (low hardening materials). Ainsworth was primarily interested in
developing a driving force procedure for high hardening materials such as austenitic
stainless steels. The strip yield failure assessment diagram was considered suitable
for low hardening materials. He proposed the following approximation.

hi(n) = hi(1) [4.19]

where hi(n) is the geometry constant for a material with a strain hardening exponent
of n and hi(1) is the corresponding constant for a linear material. By substituting
hi() into Eq. [4.18] (or [4.19]), Ainsworth was able to relate the plastic J to the linear
elastic stress intensity factor:

J 'K - f-
El ,re 1 [4.20]

where ji = 0.75 for plane strain and g. = 1.0 for plane stress.

Ainsworth's work has important ramifications. When applying the EPRI approach,
one must obtain a stress intensity solution to compute the elastic J, and a separate
solution for hi in order to compute the plastic term. The hl constant is a plastic
geometry correction factor. However, Eq. [4.20] makes it possible to estimate Jpl from
an elastic geometry correction factor. The original EPRI handbook (82) and
subsequent additions (105-107) contain hl solutions for a relatively small number of
configurations, but there are several hundred stress intensity solutions in handbooks
and the literature. Thus Eq. [4.20] is not only simpler than Eq. [4.13], but also more
widely applicable. The relative accuracy of Ainsworth's sifnplified equation is
examined in Section 4.4.

Ainsworth made additional simplifications and modifications to the reference stress

model in order to express it in terms of a failure assessment diagram. This FAD has
been incorporated into a revision of the R-6 procedure. The new document also
contains more accurate procedures for analyzing secondary stresses. The revised R-6
approach still permits application of the strip yield FAD to low hardening materials.
102
The reference stress FAD has also been included in the revised PD6493 procedure, to
be published in late 1989 or early 1990. This approach, described below, is broadly
similar to the revised R-6 method.

4.3 THE THREE-TIER APPROACH (REVISED PD6493)

The CTOD design curve, which is the basis of PD6493:1980, suffers from a number of
shrtcomings. For example, the driving force equation is mostly empirical and has a
variable level of conservatism. In addition, this approach does not explicitly
consider failure by plastic collapse. Improved driving force equations became
available with the R-6 and EPRI procedures, but the CTOD design curve had already
been widely accepted by the welding fabrication industry in the United Kingdom and
elsewhere. Many engineers are reluctant to discard PD6493:1980. Structures
designed according to the original PD6493 method might have to be re-analyzed with
the new procedure if the CLOD design curve were rendered obsolete. An additional
problem with improving the procedure the tendency for the more accurate elastic-
plastic fracture analyses to be more complex and require more training.

The conflicting goals of improving PD6493 and maintaining continuity with the past
have been largely satisfied by a three-tier approach. This idea was originally
suggested by Anderson et al. (110). The three tier philosophy assesses fracture
problems at a level of complexity and accuracy appropriate for the situation. All
three levels of PD 6493 are expressed as failure assessment diagrams. Level 1 is
consistent with the CTOD design curve approach; Level 2 utilizes a strip yield model
and is similar to the 1980 version of the R-6 method; Level 3 is based on the
reference stress model and is nearly identical to the revised R-6 procedure. Unlike
the R-6 approach, however, the revised PD 6493 permits assessments based on CTOD
as well as J and K.

Although the new PD6493 procedure has not been published, a number of recent
articles (111, 112) describe its salient features. Information in the following
subsections is based on these articles.

4.3.1 Level I

Level 1 is consistent with the CTOD design curve in the 1980 version of PD6493. The
main differences are that the equations are expressed in terms of a failure assessment
diagram, and an explicit collapse analysis is included. Level 1, which is
conservative, is intended as a screening tool.

If KIC data are used (or equivalent K values from J data), the K ratio is defined by Eq.
[4.10]. For CTOD data, Kr is replaced by 48g, defined as
 103

41
N6 crit
where 81 is the applied CTOD obtained from a modified form of the CTOD design
curve:

6=----
"~E foral/"os0.5 [4MAI

where a/E - ei, the maximum membrane strain defined in Eq. [4.31. Recall that ei
(and thus al) takes residual stresses, bending stresses, and stress conceatk;W on into
account by assuming that the maximum value of the total stress acts uniformly
through the cross section. Unlike Eq. [4.2a], the above expression does not include a
safety factor of two on crack size. In the revised approach, this safety factor is
included in the formulation of the FAD, which is a horizontal line at "1i - 1/N" for
a1/oyS 9 0.5. The Level I failure assessment diagram is illustrated in Fig. 4.5 (111,
112). For higher stress levels, the assessment line is defined from the empirical
portion of the CTOD design curve:

81- K I2 (") 2 (i-0-.25 foro,/oys >0.5 [4.22b]

a - 7sEa)l ( j
The influence of Equation [4.22b] on the FAD is illustrated in Fig. 4.5. The revised
CTOD design curve contains a conservative collapse check in the form of a
maximum stress ratio, Sr. For Level 1, Sr is defined as

Sr a ---- [4.23l
Oflow

where an is the effective primary net section stress and aflow is the flow stress,
defined as (ys + ars)/2 or 1.2 oys, whichever is less. As Fig. 4.5 indicates, the Level I
approach is restricted to 0.8 Sr because Eq. [4.22] can be nonconservative near limit
load (110).
104
L2

1---
O.4

0.2

0
0 0.2 0.4 0.6 0U 1 1.2

Sr
FIG. 4.5 The failure assessment diagrams for Levels 1 and 2 of PD6493(111,112).

4.3.2 Level 2
Level utilizes a strip yield failure assessment diagram. The assessment equation is
identical to the original R-6 relationship (Eq. (4.8]), except that it allows CTOD based
analyses:

Kr r=Sr[Jjlf (Sr)2 19[424] (

Figure 4.5 compares the Level I and Level 2 failure assessment diagrams. Note that
the Level I FAD is always conservative compared to the Level 2 method.
The treatment of stress concentration effects and secondary stresses is more complex
in the upper two levels. The procedure recommends that accurate stress intensity
solutions be obtained for the actual primary and secondary stress distributions. If this
is not feasible, an approximate solution can be obtained by linearizing the stress
distribution and separating the stresses into bending and membrane components.
For example, consider a surface crack of depth a. If the primary and secondary
stresses are resolved into bending and membrane components, the approximate
stress intensity factors are computed from the following expressions:
 105

KIP - " MgPm + MbPb) [4.25a1

K?-' (MmQm + M ) [425b

where * is the faw shape parameter, Pm and Pb are the prinary membrane and
bending stresses, Qm and Qb are the secondary stresses, and Mm and Mb are constants
obtained from the Newman and Raju stress intensity solutions (113). If, as In many
cases, the actual distribution of secondary stresses is unknown, one should assume
that Q acts uniformly across the section. The document recommends that Q be
assumed to equal the material's yield strength in the case of as-welded components.
For thoroughly stress relieved weldments, the estimate of Q can be reduced to 30% of
yield parallel to the weld and 15% of yield transverse to the weld.

The total K is the sum of the primary and secondary contributions. For assessments
based on CTOD, 81 is estimated from KI by assuming plane stress conditions:
2
81=--K1 [4.261
oysE

There is a plastic interaction between primary and secondary interaction that must be
taken into account in Level 2. This is achieved with the correction factor, p, based on
the work of Ainsworth (114). The applied toughness ratios for the structure are
given by

Kr= K!ttt + p [4.XI

and

r a + p [t
[4.281

The value of the p factor decreases as the limit load is approached because secondary
stresses are relieved by plastic strain. When a structure is proof tested, a portion of
the residual stresses are relieved. The revised PD6493 approach lets the user
incorporate the benefits of proof testing into the analysis.
106
The stress ratio, St, is defined as the ratio of the effective net section stress to the flow
stress, as in Level 1. When the point defined by Eqs [4.23], [4.27] and [4.28] falls inside
of the Level 2 assessment line (Eq. [4.241), the structure is considered safe.

4.3.3 Level 3
The Level 3 failure assessment diagram is based on Ainsworth's r, ference stress
approach (109). The FAD is related to the material's stress-strain behavior.
KVrI. 4rI E In (I + eIf) + ref)3 -1/2

£I(Frf(I +yef) 2 oys2 EIn (I + ef)J

The above quantity is plotted against the load ratio, Lr, defined as

S= [430

Note that the reference stress and the effective net section stress are equivalent.
Since the load ratio is defined in terms of the yield strength rather than the flow
stress, Lr can be greater than I. The load ratio cannot exceed aflow/ays, where aflow
is defined as the average between yield and tensile strengths. For Level 3, the
alternate definition of flow stress (Oflow = 1.2 ays) does not apply. For Lr > Cflow/OYS,
Kr-0.
If the stress-strain curve for the material is not available, such as would be the case
when analyzing a flaw in the heat affected zone, the following FAD equation can be
applied at Level 3:

Kr, V - (I - 0.14 Lr2 ) [0.3 + 0.7 exp (-0.65 Lr6)] [431]

This expression also has a cut-off at Lr = aflow/ays • Equation [4.31) was adopted from
the latest version of the R-6 procedure (115). This alternate FAD requires a
knowledge of only the yield and tensile strengths of the material, but this
relationship can be excessively conservative (112). For many materials, Eq. [4.311 is
more conservative than the Level 2 FAD. Figure 4.6 is a plot of Eq. [4.31]. Note that
the upper cut-off on Lr depends on the hardening characteristics of the material. If
the material has a yield plateau, the revised PD6493 recommends that Eq. [4.31] be
restricted to Lr s 1.

The Level 3 analysis of Kr (or s/i ) for the structure is identical to the Level 2
procedures (Eqs. [4.261 - [4.281), but Level 3 includes guidelines for ductile instability
and tearing analysis. This type of analysis is described in Section 4.5.
 107

12

0 L2) [ 0.7 -
Kr,

" TypWa cut-off for"

0.6 - uln mlm.
" TypWa .__
cut-off for

low C 1&
0.4

0.2

0 0.4 0.8 1.2 1.6 2

Lr
FIG 4.6 Alternate Level 3 failure assessment diagram for PD6493 (106).

4.4 COMPARISON OF DRIVING FORCE EQUATIONS: A PARAMETRIC

STUDY
The primary advantage of Ainsworth's reference stress approach is in accounting for
the geometry of a cracked structure through a linear elastic stress intensity solution.
The hi factor is replaced by an LEFM geometry factor. The other contribution of
Ainsworth's analysis, the generalization to stress-strain laws other than power-law,
is of secondary importance.

In most cases, the EFRI procedure and Ainsworth's simplified approach produce
nearly identical estimates of critical flaw size and failure stress. This section presents
the results of a parametric study of the accuracy of Ainsworth's approach relative to
that of the EPRI procedure. The relative accuracy of the strip yield model was also
evaluated. A power law hardening material was assumed for all analyses, since the
main purpose of this exercise was to evaluate the errors associated with the LEFM
geometry correction factor in the elastic-plastic regime.

For a power-law material, the Ainsworth model gives the following expression &-r
the total J:
08

J= [(-v2) + a gL (,P71] [432

Since stress intensity is proportional to load, this relationship has the form

J=Clp2 + C2pn+.

The first term dominates under linear elastic conditions; the second term dominates
under fully plastic conditions. The EPRI approach (Eqs. [4.12] and (4.14]) has the same
form. The only difference between the EPRI equation and Eq. [4.32] is the value of
the constant C2; the equations agree precisely in the linear elastic range. Thus any
discrepancies between the two approaches are observed only when the plastic term is
significant.

When the plastic term dominates, j is very sensitive to the applied load. A small
error in the estimate of P results in a large error in the estimate of J, but this
sensitivity be an advantage when predicting failure. Since J is very sensitive to load,
P is very insensitive to J, implying that the failure load in the plastic range is
insensitive to the fracture toughness; i.e. the critical J. In the fully plastic range,
failure is essentially collapse controlled rather than fracture controlled, as seen
dearly in the failure assessment diagram in Fig. 4.2. As the stress ratio approaches
unity, the diagram becomes nearly vertical, indicating that the failure stress is
independent of the toughness ratio. At these high stress ratios, the failure stress is
governed by the material's flow properties rather than its fracture toughness.

The predicted failure stress in the fully plastic range is also insensitive to the
differences between the EPRI approach and the Ainsworth model. The latter
approach assumes that the geometry factor, hi(n), is equal to the linear elastic value,
hI(1). Errors in J that result from applying Eq. [4.32] are proportional to the ratio
hj(n)/hj(1), plotted against n in Fig. 4.7 for a center cracked panel in plane strain with
a/W - 0.75. Note that the hl ratio in this configuration is sensitive to the hardening
exponent. Thus Eq. [4.321 leads to significant errors in J, particularly at high n values.
However, when the hl ratio is raised to the power 1 /(n+l ), it is insensitive to n. This
latter ratio is indicative of the differences in the predicted failure stress between the
Ainsworth and EPRI approaches.

A design engineer often wishes to use a fracture mechanics analysis to estimate the
critical flaw size at a given applied stress. To determine the sensitivity of critical
flaw size estimates to the driving force equation, a series of calculations were
performed with the EPRI, Ainsworth, and strip yield models on center cracked
panels and edge cracked bend specimens. Hardening exponents of 5, 10 and 20 were
assumed. Only plane strain conditions were considered.
 109

The material was assumed to follow perfectly the power law expression (Eq. [2.22]) for
stress versus plastic strain. The constants a and eo were fixed 1.0 and 0.002,
respectively, for all n values; thus c0 corresponds exactly to the 0.2% offset yield
strength.

Two separate strip yield analyses were performed for each case: one assumed that the
collapse stress was equal to the yield strength; the other based collapse on the flow
stress, defined as the average of yield and tensile strengths. For a material whose
true stress-true strain curve follows a power law, the flow stress can be estimated
from the following expression:

MF 1 +
LI [4.331

where N = 1/n.

The results of the analysis of the center cracked tension panel are plotted in Figs. 4.8
to 4.13. Critical crack size, normalized by W, is plotted against critical J, normalized
by width and yield strength. Figures 4.8 to 4.10 compare the EPRI, Ainsworth, and
strip yield equations for a constant nominal stress equal to 2/3 yield. At low
toughness levels, all predictions agree because linear elastic conditions prevail. At
high toughness levels, the curves are relatively flat, indicating that critical crack size
in insensitive to toughness. In this region, failure is controlled primarily by plastic
collapse of the remaining cross section. The EPRI and Ainsworth equations agree
well at all hardening rates.

The strip yield model is nonconservative for the high hardening material (Fig. 4.8)
when it is based on the flow stress. For this material, the flow stress, as estimated
from Eq. [4.33], was 1.53 Oys, but PD6493 restricts the flow stress to 1.2 times yield in
Level 2 assessments. Thus the nonconservatism observed in Fig. 4.8 would not
occur with the restrictions in PD6493. For n - 10 and n - 20, the estimated flow
stress was 1.17 and 1.06 times yield, respectively. In both cases (Figs. 4.9 and 4.10), the
strip yield model, when based on flow stress, produces good estimates of critical crack
size. When the strip yield model utilizes the yield strength, it is always conservative.

Figures 4.11 to 4.13 compare the EPRI and Ainsworth approaches at various stress
levels. In all cases, the agreement between the two approaches is very good. The
Ainsworth model tends to be slightly conservative for the center cracked
configuration.
The bend specimen results are presented in Figs. 4.14 to 4.19. In Figs. 4.14 to 4.16, the
load is fixed at 50% of Pu, the limit load of an unnotched plate in plane strain. The
load is varied in Figs 4.17 to 4.19. The bend results are similar to the center cracked
paneL The strip yield model works best for low and moderate hardening materials.
to
In addition, the EPRI and Ainsworth analyses agree well in most cases, but the
Ahworth model appears to be slightly nonconservative for the high hardening
material (Fgs. 4.14 and 4.17), especially at high stresses.
These analyses Indicate that the Ainsworth model can predict either critical crack size
or failure stress in the elastic, elastic-plastic, and fully plastic regimes. The strip yield
model gives reasonable results, provided the flow stress is restricted to 1.2 times

Pkedictions of applied J in the fully plastic regime can vary widely for the three
approaches analyzed. Initially one might conclude that the EPRI model always gives
de best predictions of J because it is based on the most sound theory. This is not
necessarily the case, however. Because the plastic component of J is very sensitive to
load, a slight variation in flow properties leads to major variations in Jpi at a fixed
load. The yield strength is proportional to Po, which is raised to a power of n+1 in
both the EPRI and Ainsworth equations. Thus errors in J resulting from the
Ainsworth approximation may lie within the scatter band of errors resulting from
variations in material flow properties. The more complex EPRI formulation may
lead to no more accurate predictions than the simpler approaches.

Figure 4.20 illustrates the effect of varying flow properties on predictions with the
EPRI approach. A 10% overestimate or underestimate of co for the material leads to
approximately a 10% error in the prediction of acrit in the plastic regime. Such an
uncertainty in flow properties corresponds to a 260% uncertainty in the applied J
when n=10. Since flow properties typically vary by several percent with in a given
plate and as much as 20% between heats, accurate predictions of the applied J in a
structure in the plastic regime appear to be impossible.
 CENTU CAOCX PANE1W
od e.,s

DA

0
1 2 3 4 S 6 7 S 9 10 11 2.2 13 14 15 16 2V U 1V 20
HARDVNINC EXPNINT (an)

FIG 4.7 The effect of strain hardening exponent on the 11 factor Inacenter cracked panel In
plane strain with u/W a30.75.

0.7

0.6 OIC = 0.67

035

Based on Yield.SqVOgt

*-AtISWORTH MODEL

*STRIPY~M.DbMDEL

0.1 I

0 0OAS 000 001

FIG. 456 The sBect of fracture toughness on critica crack sin Ina center cracked panel with n 85
aNd nominal stress equal to 0.57 timee yield.
 0.7
CETER CRACKED PANJl9 nalO,

0F/a =0.67
acrt O

-- Bandon Fow Sess -

Bandon Y1id Sra

0.0
-DIIOUA1ON

0.2 -- USOT
- -7W Y=EDMODEL

0.1
0 0=00 0.01 0.015 0.02

W 0.
FIG. 4.9 The effect of fracture toughness on critical Crack size Ina center cracked panel with n x10
and nominal stress equal to 0.67 times yield.

0.7
CENTER CRACKED PANEL, n=20

OA

AN4WORTH MODEL
0.2 ::RE --

YMLD
E~ MODEL

0..

0 0.005 0.01 0.01 0.02

FIG. 4.10 The effect of fracture toughness on critical crack size In a center cracked panel with n a=0
and nominal stress; equal to 0.67 times yield.
 113

0.7
0.7 oo 0 -O50

OA

0.1

Wa

0.7

ajat
WW0

0.5

0.4
006 0.00500001 .02
03i
EW WAIO
FI.412Teefeto ratr ouhes nciiclcak iei acnercakd.ae2it 1
at various
leDis tress
114

i tm0.7?/ 6 05

0/o 06

0.2
-
-EPI EQUJATION
0.1 0005--IS
0 0. 05 .010 .0 13 0.02

FIG. 4.13 The effect Of fracture Wa

toUghness On clifuc"r ck size Ina center cracked
at various stress levels. panel with n =20

0.7
Asa BEND SPECINEN, n=.5

W 0.2
PP - 2 W0 M D0

0- 0- ---- . -0 ------ ---0 - .-

0 .4.. .. . B se o Fl w t m
ff ctoffr03ct
FIT .e .1 ase o YildWtrn
gh es
reto lcrtia.cr cksie.n..bn.s..i
11. . . . . . .e .
and~
~ ~~~~~~~ER
wit .
thEoaQqaUtAhlTheIONce
lmt od
 0.7
SEND SPEOMEN, 3.10
P/P. - 030
OA
acit
w

OA .......

0.2 IL--AR45WORTH MODE.

4 .- SRW YEW MODEL

0.1
0 0=00 0.01 0.015 0M0

W 0.
FIG. 4.15 The effect of fracture toughness an critical crack size In a bend specimen with n .10
and the load equal to half the unnotched limit load.
0.7 1
BEND SPECMUJ, n=20
P/P. .0.50
OA

0.4

-EVE EQUATION

0.2 *-A26VORTH MODEL

*-5REP YIELD MODEL

0 0.005 om1 0.015.0

1J
rit-
W a.
FIG. 4.16 The effect of fracture toughness on critical crack size Ina bond specimen withN~ 20
and the load equal to half the unnotched limit load.
116

0.7

0.1
0 0.005/P 0 0 1005.1

0.7

0.0 o S.0

0.4a

0.7

O 0ci R 0.0 .100002

ww
Fi.41 Teefc o rcue ogns o rtca rc iz nabndseiejwt 1
Arou lodlees
 I117

BEND SPECIEN, m=

NdILP/P. 0.25

*A

0.2O

0.1
0 0.005 0.01 0.03 0.0

FIG. 4.19 The efec of fracture toughness on critca cracktsize In a bend specimen with nl 20
at various load levels.

0.7
CENTER CRACKED PANELv nalo

OA

5.1 I I
* .05501 0.01 0.02

-l
W CF
FIG. 4.20 The effect of a 10% error Inflow properties on crilcal crack os estimates from the
EPRI approach.
118
4.5 PROBABILISTIC FRACTURE MECHANICS

Most fracture mechanics analyses are deterministic; Le.. a single value of fracture
toughness is used to estimate failure stress or critical crack size. Much of what
happens in the real world, however, is not predictable. Since fracture toughness data
in the ductile-brittle transition region are widely scattered, it is not appropriate to
view fracture toughness as a single-valued material constant. Other factors also
introduce uncertainty into fracture analyses. A structure may contain a number of
flaws of various sizes, orientations and locations. Extraordinary events such as
hurricanes, tidal waves and accidents can result in stresses significantly above the
intended design level. Because of these complexities, fracture should be viewed
probabilistically rather than deterministically.

Figure 4.21 is a schematic probabilistic fracture analysis. The curve on the left
represents the distribution of flaws in the structure; the curve on the right is the
critical flaw distribution. This distribution depends on the applied stress and the
distribution of fracture toughness. The critical flaw size corresponding to a particular
stress and toughness can be computed from one of the driving force equations
described in previous sections. When the distributions of critical and actual flaws
overlap, there is a finite probability of failure, indicated by the shaded area. Time-
dependent crack growth, such as fatigue and stress corrosion cracking, can be taken
into account by applying the appropriate growth law to the flaw distribution. This
distribution moves to the right with time, increasing the failure probability.

Critical Fla
FREQUENCY Distribution

of Failure

FLAW SIZE

F10.4.21 Schematic of a probabilistic fracture analysis.

 119

The mathematics of probabilistic analysis is well established (116). Reliability

engineering is currently applied in a variety of circumstances, ranging from quality
control in manufacturing to structural integrity. Probabilistic fracture analyses are
rare, however, because the input data are usually not available.

Scatter in fracture toughness data is the main culprit for uncertainty in critical flaw
size. A probabilistic analysis traditionally requires performing a large number of
fracture toughness tests to define the toughness distribution, but recent research
results may greatly reduce the amount of testing required.

Section 3.5 describes a statistical model for fracture in the transition region that
indicates that cleavage fracture toughness data follow a characteristic distribution.
With knowledge of the shape of the toughness distribution, one needs only to
establish its absolute position. This requires far fewer tests than needed if nothing is
known about the distribution. An added benefit is a means for analyzing the
probabilistic effect of microstructural variations and local brittle zones. This model
has recently been applied to probabilistic fracture analysis of nuclear reactor pressure
vessels.

4.6 TEARING INSTABILITY ANALYSIS

So far, this chapter has focused on initiation of fracture at a critical value of J, K, or

CTOD. However, a ductile material with a rising R curve does not usually fail when
JIc is reached. A tearing instability analysis is required to predict ultimate failure of
structures made from materials with rising R curves.

Chapter 2 introduced the concept of stable and unstable crack growth. As Fig. 2.2
illustrates, unstable crack propagation occurs when the rate of change in driving
force is greater that the rate of change in material resistance. Cleavage fracture is
usually unstable because the material resistance decreases with crack extension (Fig.
3.32). However, crack growth in structural steels on the upper shelf is usually stable
because the R curves are relatively steep. On the other hand, high strength
aluminum alloys can fracture in an unstable manner because the R curves are fairly
flat.

Figure 4.22 illustrates a tearing instability analysis based on J. The relative stability of
the structure depends on how it is loaded. The solid driving force lines represent
load control; the dashed lines indicate displacement control. When load is fixed, the
driving force increases with crack extension, but the driving force decreases with
crack extension when the displacement is fixed. The load controlled structure
eventually becomes unstable as load increases, but the same structure in
displacement control may remain stable.
120

P/
P3
J hstabfilty in %%
Load Control
A5

A2.

Aa
FIG. 4.22 Comparison of crack growth resistance (heavy line) with driving force In load
control (solid curves) and displacement control (dashed curves).

Most structures exist somewhere between pure load control and pure displacement
control. The general case can be represented schematically as a spring in series with
the structure, as in Fig. 4.23. In displacement control, the system compliance, CM, is
zero; i.e., the system is infinitely stiff. Load control corresponds to an infinitely
compliant system. Elastic energy stored by the "spring" helps to drive the crack.
Thus the more compliant the system, the more available potential energy, and the
less stable the system tends to be.

The load-displacement behavior of a hypothetical structure is illustrated in Fig. 4.24.

If the structure is load controlled, it becomes unstable at the maximum load plateau.
In pure displacement control, the structure remains stable, even after significant
amounts of crack growth. At the intermediate condition, instability occurs after
maximum load.

Paris and Hutchinson (117) have developed a detailed instability analysis for ductile
materials that takes account of system compliance. Their approach has been
incorporated into the EPRI elastic-plastic fracture handbook.

Paris and Hutchinson introduced the concept of the stability assessment diagram.
Their approach requires that slopes of the driving force and resistance curves be
nondimensionalized by means of the tearing modulus. The tearing modulus for the
material is defined as
 121

AT P

CM

FIG. 4.23 Schematic structure loaded to a fixed remote displacement. CM is the

system compliance.

LOAD

DISPLACEMEINT

FIG. 4.24 Schematic load-djsplacement Curve. The relative stability of the structure
depends on CM.
122

E dJR [434]

where JR is the value of J on the R curve. The applied T depends on the partial
derivative of J at a fixed remote displacement.

Ev ia 4351
Tj'. " la ) AT

The conditions for stable and unstable crack extension are as follows.

Tj < TR (stable) [4.36a]

Tj k TjR (unstable) [436b]

The above conditions for stability are analogous to those presented in Chapter 2 for
linear elastic conditions (Eq. [2.21). The partial derivative in Eq. [435] is obtained from
a rather complex expression:
ia_a ( / Al 75CM+(
p.]" [437]

(,aal)T (aajp (aPa

Each of the terms in the above expression is obtained separately from the
relationships for P, A, and j in the EPRI handbook. The computation is laborious if
done manually; but for most practical situations, one can make the conservative
assumption that the structure is load controlled and CM = 0s simplifying Eq. [4371
considerably:

=(a. [4.38]
(a AT (a)p

The situations in which one would perform a complete analysis with Eqs. [4.341 to
[4.371 are rare. As illustrated in in Fig. 4.4, instability occurs after maximum load if
the system compliance is finite, but no competent designer would let a structure to
go beyond limit load under normal service conditions. The only time a detailed
tearing analysis is appropriate is for prediction of the behavior of the structure under
extraordinary loading. The above approach was originally developed to analyze
hypothetical accidents in nuclear reactors. Actual service stresses in a reactor
pressure vessel are low enough to ensure that the structure is loaded well below the
Jlc of the material.
 123

4.7 SIMPLIFIED ELASTIC-PLASTIC ANALYSES

Complete tearing instability analyses can be extremely complex and there is no

guarantee of accuracy, since computed J values are highly sensitive to the input data
(Section 4.4). Even the three-tier approach can be complicated.

There are many instances when simple screening criteria would be useful. The
analyses in Section 4.4 illustrate that if fracture toughness is above a certain level,
structural failure is controlled primarily by the material's flow properties. In such
cases, the designer can avoid failure merely by ensuring that the structure is stressed
well below its limit load. Thus one purpose of a fracture toughness screening
criterion is to define the boundary between collapse controlled failure and fracture
controlled failure.

Two simple fracture toughness screening criteria have been been published recently
(71,118). One approach pertains to fracture initiation, either by cleavage or ductile
tearing, and the other relates to tearing instability.

4.7.1 The Yield-Before-Breek Criterion

The yield-before-break criterion (71) is analogous to the leak-before-break approach, a

toughness criterion designed to ensure that a crack in pressure vessel or pipe grows
through the wall before it becomes unstable. As its name suggests, the yield-before-
break criterion quantifies the required level of toughness to ensure that fracture
initiation occurs after net section yield.

Consider a structure that contains a surface flaw of depth a. If the flaw is small
compared to the total cross section of the structure, the CTOD at net section yield can
be estimated conservatively from the following equation.

8MT= - (1.25ey+.002) [4391

where Mm and 0 are as defined in Eq. [4.251. This equation was derived from the
reference stress approach. This simple result was obtained by assuming nomimal
stress, reference stress, and yield strength are equal.

The yield-before-break concept can be illustrated with a simple example. If a 25 mm

(1.0 in) thick plate contains a semicircular surface flaw 12 mm deep, 8 NSY a 0.09 mm,
assuming ey = 0.002. Thus if the material is at least this tough, the structure will not
fail, as long as the applied stresses are below yield.

Equation [4.391 can be modified to take account of residual stresses and stress
concentration effects. If yield magnitude residual stresses are present and there is a
local stress concentration, kt, the yield-before-break CTOD is given by
24

kt2 Mm 2 a
8NSY - (2.25 ey + 0.002) [4.401

This modified equation is based on a number of conservative assumptions.

Therefore it will tend to overestimate 8NSY.

A word of caution is necessary when applying this approach in the transition region.
Since cleavage fracture toughness data in the transition region are invariably
scattered, it it is not sufficient for the average toughness to be above the yield-before-
break threshold. In order to ensure a high probability of yield-before-break
conditions, the designer should compare a lower bound toughness (from Table 3.1)
to 8NSY.

4.7.2 Critical Tearing Modulus

Not only does the reference stress model simplify fracture analyses in the transition
region, it can also simplify tearing instability analysis. Anderson et al. (118) showed
that it is possible to obtain dosed-form solutions for ductile fracture in a number of
simple configuration -TIhese solutions can define the boundary between fracture
controlled failure ann collapse controlled failure on the upper shelf.

Consider a canter cracked panel in plane strain. The following closed form
expression was derived for the critical tearing modulus at maximum load:

2 a x (n+1) (n a) a ra( n+[441

Tc sec I- n(4.41]

By fixing the ratio an/ays , it is possible to calculate the necessary tearing modulus to
reach on. Figure 4.25 is a plot of sample calculations with Eq. [4.41] The net section
stress was fixed at yield and 10% above yield. The flow property constants n and a
were fixed at 10 and 2.0, respectively. According to this plot, the structure will reach
net section yield before tearing instability, as long as the crack is small relative to the
width and the material's tearing modulus is greater than approximately 10. Since
typical structural steels have tearing moduli greater than 50, nearly all steel
structures on the upper shelf must completely yield before tearing instability is a
possibility.
A similar expression to Eq. [4.41J has been derived for surface flaws (71). The
computed Tc values are similar to the center cracked results.
 125

so
CENTER CRACKED PANEL
n=10, a=2.0

60

40 ays

(In_ 1.0
S20 y

0
0 0.2 0.4 0.6 0.8
a/W

FIG. 4.25 Crilcal tearing modulus for a center cracked panel In plane straln(112).

With a philosophy similar to that of the yield-before-break approach, if the material

is on the upper shelf and its tearing modulus is greater than the critical value, the
structure will not fail as long as the net section stresses are kept below the yield
strength or flow stress.

4.8 UNRESOLVED ISSUES IN ELASTIC-PLASTIC FRACTURE ANALYSIS

Although some of the fracture design analyses discussed in this chapter are complex,
they still do not take into account all aspects of the problem. These analyses are two-
dimensional and assume that the material is homogeneous. In addition, all fracture
design analyses contain an inherent assumption that the computed driving force
parameter (K,J, or CTOD) uniquely characterizes crack tip conditions.
The items discussed in this section do not constitute an exhaustive list of unresolved
issues, but are key areas which need to be understood better before fracture analysis
methods can be improved.
It should be noted that current methods of fracture analyses are generally safe. The
effects discussed below tend to make current predictions conservative. A better
understanding of these complexities would merely make analyses of critical
conditions more accurate.

4.8.1 Driving Force In Weldments

A steel weld invariably has different flow properties than the parent metal. In most
cases, the yield strength of the weld metal overmatches that of the parent metal,
although undermatching sometimes occurs. Analyses such as the EPRI approach are
unable to handle structures whose flow properties are heterogeneous. If a crack
occurs in or near a weld, it is impossible to determine the driving force accurately
without performing an elastic-plastic finite element analysis of the component.

This problem can be illustrated by citing an example from Ref. (112). Three welded
aluminum center cracked panels were fabricated, with the weld metal
undermatching the strength of the parent metal. The crack in each panel was in the
weld, which ran perpendicular to the loading direction. Narrow strips were extracted
from each weld for tensile testing. Each tensile specimen was loaded perpendicular
to the welding axis, so as to be representative of loading in the wide plates. The gage
section contained parent metal on either side of the weld. A strain gage was
mounted on the weld to measure the effective flow properties of the weldment.
These flow properties, along with crack growth resistance curves from small scale
specimens, were used in conjunction with the reference stress model to perform a
tearing analysis of each wide plate specimen. These analyses underestimated the
failure stress by as much as 32%, despite extra care in measuring the true flow
properties of these panels. The apparent explanation for the discrepancy is that the
weld metal in the narrow tensile specimens was less constrained than in the wide
plates. Hence the effective yield strength was higher in the wide panels.

4.8.2 Residual Stresses

The assumptions for secondary stresses have a significant effect on predictions with
either the R-6 method or PD6493, but accurate information on the distribution of
residual stresses is rarely available for the weld in question.

Conventional methods for measuring the through-thickness distribution of residual

stress are destructive. Material from one side of the welded plate is typically
removed by a milling machine while strain gage readings on the other side of the
plate are recorded. Such an approach is obviously impractical for a structure in
service. The center hole drilling technique does minimal damage to the structure,
but it only provides information on the surface stresses. The only available method
for through-thickness residual stress measurement is neutron diffraction, a
technique is not portable.
 127

A reliable, portable, nondestructive method for measuring residual stresses is

desperately needed. Accurate finite element models that predict the residual stress
distribution from the joint geometry and welding procedure are also desirable.

4.8.3 Three-Dimensional Effects

Existing elastic-plastic analyses do not account for the variation of the driving force
along the crack front. Parks (113) has conducted numerical studies and Reuter et al.
(114) have performed experiments that demonstrate that both J and CTOD vary
considerably along the tips of surface flaws. Thus the crack tip conditions cannot be
uniquely characterized with a single value of J or CTOD.

These three-dimensional effects influence both cleavage and ductile tearing. Since
cleavage is statistical, good predictions will come only from summing the
incremental failure probabilities along the crack front. Such a calculation must take
account of the variation in the crack driving force with position. Since ductile crack
growth occurs fastest where the driving force is highest, accurate tearing predictions
are possible only with a three-dimensional analysis.

4.8.4 Crack Tip Constraint

Constraint is related to the three-dimensional issue. Plane strain fracture analyses

assume that J or CTOD uniquely characterizes crack tip stresses and strains. If the
entire crack front is not in plane strain, however, there are regions where a single
parameter does not characterize crack tip conditions. Similarly, the single parameter
assumption breaks down under large scale plasticity, which decrease the crack tip
constraint. This constraint loss can occur at very low J (or CTOD) values in
structures loaded predominantly in tension (see Section 3.5.2). In such cases, the
structure has a higher apparent toughness than the small scale fracture toughness
tests, which were loaded predominantly in bending.

The experiments and analyses of surface cracks by Reuter et al. (120) and Parks (121)
demonstrate the complexities of constraint The apparent driving force, quantified
by either J or CTOD, was highest at the maximum crack depth; i.e., along an axis
perpendicular to the plate surface. The J and CTOD values decreased smoothly along
the crack front, reaching a minimum at the free surface, but the maximum crack
growth occurred along an axis approximately 450 from the free surface. Parks' finite
element analyses indicated that this was also the location of the highest normal
stresses, even though J and CTOD were highest at the maximum crack depth. The
crack tip constraint had apparently relaxed at the point of maximum depth, which
was dose to the back surface of the plate. Because of constraint effects, the highest
true driving force and the maximum apparent driving force were at two different
locations.
128

4.8.5 Gross-Section Yielding

The presence of a crack reduces the cross section in the structure; the net section is
defined as the portion of the cross section not occupied by the crack. Net-section
yielding refers to the point when the plastic zone spreads throughout the net cross
section. Grost.-section yielding occurs when plasticity encompasses the entire cross
section.

Net-section yielding tends to occur with deep cracks, while gross section yielding is
more common with shallow flaws. Since deep cracks are usually avoided in
structures, gross-section yielding is much more common. Unfortunately, most
elastic-plastic fracture analyses are not equipped to handle this type of deformation.

Mastic-plastic fracture analyses such as the EPRI and R-6 methods assume net-section
yielding in the structure, and are conservative when gross-section yielding occurs.
According to the EPRI approach, I in the fully plastic range should scale with Pn+l.
However, experimental results of Read (121) indicate that gross-section yielding
causes the applied J to increase much more gradually than expected. Thus an elastic-
plastic fracture analysis can greatly overestimate the applied J in the case of gross-
section yielding. Read's results indicate that gross-section yielding is most likely
when the crack comprises less than 5% of the cross section.
 129

5. DISCUSSION

This chapthf L' intended to provide a perspective on the state of the art in elastic-
plastic fracture mechanics. Many of the issues raised in Chapters 3 and 4 are
disc s< -, ,,.k Where appropriate, recommendations for future directions are
made.

The field of fracture mechanics is relatively mature. The current state of technology
is sufficiently advanced to make a major contribution to industry. Not all of the
important questions have been answered. But enough is known to apply fracture
mechanics to practical situations.

One sign of the field's maturity is that all of the simple problems have been solved;
those that remain are very complex. Future researchers must be very creative to
make significant contributions.

5.1 FRACTURE TOUGHNESS TESTING

Chapter 3 demonstrates that the KIC test is of limited value for testing low- and
mediun- strength steels. If a steel can satisfy the size requirements of ASTM E399-83
(25), it is probably too brittle for structural applications. Thus fracture toughness in
such materials must be quantified by elastic-plastic tests.

Fracture toughness testing procedures for materials on the upper shelf are well
established. The JIc and J-R curve standards (26, 50) provide guidelines for
measuring the material's resistance to ductile fracture initiation and crack growth.
One problem receiving some attention is the crack growth limits in ASTM E1152-87.
Ernst (122) proposed a modified J integral, which has been moderately successful in
producing geometry independent R curves beyond the normal crack growth limits.
This research is driven primarily by the nuclear industry, where accurate tearing
instability analyses are important, but this problem is only marginally important to
the rest of the welding fabrication community.

Just as materials that satisfy the KIC size criterion are usually too brittle, materials on
the upper shelf are sufficiently tough so that fracture is not often a significant
problem. The fracture research area most important to the welding fabrication
industry is the ductile-brittle transition region.

Until recently, the transition region has received little attention from the fracture
mechanics community in the United States. The CTOD standard (51), the first
standardized method which can be applied to the transition region, was published in
130

1989 by ASTM, whereas the British Standards Institute published a C'TOD standard in
1979, and CTOD data were applied to welded structures 15 years earlier.

Because J integral test methods were originally developed for the upper shelf, there is
no standardized J-based test that applies to the transition region. Such a standard
should be developed so that J-based driving force approaches can be applied to
structures in the transition region.

One problem with both J and CTOD testing in the transition region is the lack of size
criteria to guarantee a single parameter characterization of fracture. The Jic size
requirements are probably not restrictive enough for cleavage, and the KIC
requirements are too severe for elastic-plastic fracture parameters. The appropriate
size requirements can be established through a combination of finite element
analysis and micromechanics models, such as that described in Section 3.5.

When a single parameter description of fracture toughness is not possible, as in

shallow notched specimens and tensile panels, the issue of crack tip constraint
becomes important. This is a very difficult problem. Unless a simple analysis is
developed that characterizes constraint loss, these effects will be impossible to
quantify without performing three-dimensional, elastic-plastic finite element
analyses on every configuration of interest. The two-parameter approach described
in Section 3.5.2 shows some promise but requires further study.

Another important issue is fracture toughness testing of welirents. Existing

standards do not address the special considerations required for weldment testing.
The Welding Institute and other organizations have developed informal procedures
over the years, but they need to be standardized.

Fracture toughness data in the transition region are invariably scattered, whether the
tests are performed on welds or base materials, although the problem is worse in the
heat-affected zone of welds. The nature of scatter in the lower transition region is
reasonably well understood; procedures have been developed which allow for
estimating lower-bound toughness with as few as three fractire toughness values.
The problem of scatter in the upper transition region is more complicated; constraint
loss and ductile crack growth combine to increase the level of scatter. Further work
is necessary to quantify these effects.

An accurate correlation between Charpy energy and fracture toughness would be

extremely useful. The empirical correlations developed to date are unreliable. Some
progress has been made in developing theoretical correlations, but these models do
not take into account all factors. If an accurate relationship can be developed,
material toughness criteria based on Charpy energy can be established rationally.
 131

5.2 APPLICATION TO STRUCTURES

Although linear elastic fracture mechanics is of limited use in fracture toughness
testing of structural steels, LEFM driving force relationships are suitable for many
situations. A structure of interest, if it is sufficiently large or the stresses are low, may
be subjected to nearly pure linear elastic conditions. Fracture toughness can be
characterized on a small specimen by a critical J value, which can then be converted
to an equivalent Kic and compared to the applied Ki in the structure.

Pure LEFM analysis does carry risks, however. If the stresses are above
approximately half the yield strength, plasticity effects can be significant. If the LEFM
analysis does not contain some type of plasticity correction, it gives no warning when
the linear elastic assumptions become suspect. Sufficient skill is necessary to
determine whether or not an LEFM analysis is valid in a given situation.

It is perhaps better to apply an elastic-plastic driving force relationship to all

problems; then, the appropriate plasticity corrections are available when needed.
When a linear elastic analysis is acceptable, the elastic-plastic approach will reduce to
the LEFM solution. Thus the analysis decides whether or not a plasticity correction
is needed.

Several types of elastic-plastic fracture analyses are available. The CTOD design
curve, based primarily on an empirical correlation between wide plate tests and
CTOD data, is largely obsolete. Analyses based on the strip yield model are still
useful for low hardening materials. The EPRI procedure is the probably the most
advanced analysis, but it is currently applicable to a limited range of configurations.
The reference stress model, which is a modified version of the EPRI approach, is
widely applicable. Any of these approaches can be expressed in terms of a failure
assessment diagram. This is done merely for convenience, and has no significant
effect on the outcome of the analysis.

The parametric study in Section 4.4 produced some interesting results. As expected,
the strip yield, Ainsworth, and EPRI analyses all agreed in the linear elastic range. In
the elastic-plastic and fully plastic ranges, where the three analyses might be expected
to differ, predictions of failure stress and critical crack size were quite close in most
cases; the only exception was when the strip yield model was applied to the high
hardening (n-5) material. All the analyses predicted similar failure stresses and
critical crack sizes because failure in the fully plastic range is governed by the flow
properties of the material. Above a certain level of toughness, critical values of stress
and crack size are insensitive to fracture toughness.

The analyses do differ in the prediction of the applied J, but for a designer, critical
crack size and failure stress are much more important quantities. Accurate
predictions of the applied J may be impossible, even with an analysis that is
theoretically perfect. The applied driving force in the plastic range is highly sensitive
132

to the P/Po ratio. A slight overestimate or underestimate of Po significantly effects

the results. If the flow properties vary even by a few percent, the resulting error in Po
leads to a large error in the J calculation.

In summary, the driving force expression probably does not matter in most cases.
The only requirements are that the expression reduce to the LEFM solution for small
scale yielding and predict the correct collapse limit under large scale yielding
conditions. An additional proviso is that the strip yield approach or other
nonhardening models should not be applied to high hardening materials.

Since the Ainsworth model works as nearly wel as the EPRI approach, there is little
justification for the EPRI approach in non-nuclear applications. The EPRI procedure
is more cumbersome because it requires a fully plastic geometry correction factor.
The Ainsworth model produces similar results to the EPRI analysis and has the
advantage of a geometry factor based on stress intensity solutions. Currently, there
are many more published K solutions than hi solutions.

There are other not to worry about applying accurate plastic geometry factors. Real
structures, especially welded structures, pose many complex problems that existing
analyses cannot address. As Section 4.8 indicates, the elastic-plastic driving force in a
weldment cannot be represented accurately by a solution for a homogeneous
structure. Additional factors such as residual stresses, three-dimensional effects, and
crack tip constraint, and gross-section yielding combine to increase the uncertainty
and potential errors in fracture analyses. These errors are much more significant
that those that might arise from choosing the strip yield or Ainsworth analysis over
the EPRI approach. Until the complexities discussed in Section 4.8 can be addressed,
one may as well adopt a simple elastic-plastic analysis.

As a first step in a fracture analysis, a simple screening criterion may be appropriate.

Two such approaches are introduced in Section 4.7. The yield-before-break criterion
estimates of the level of toughness required for the structure to reach net section
yielding before fracture initiation. If the toughness is adequate to ensure yield-
before-break conditions, fracture can be avoided simply by ensuring that the structure
is loaded well below its limit load. An analogous quantity, the critical tearing
modulus, is designed to ensure that the tearing resistance is adequate to avoid a
tearing instability below the limit load.
 133

6. CONCLUSIONS AND RECOMMENDATIONS

6.1 FRACTURE TOUGHNESS TESTING

" KIC testing is of little value for structural steels. Fracture toughness should
be quantified by elastic-plastic parameters such as J and CTOD.

" The ductile-brittle transition region is the critical area of concern in welded
steel structures. If a material is on the lower shelf of toughness, it is too
brittle for structural application. If the material is on the upper shelf,
fracture is not a significant problem.

A standard for J testing is the transition region should be developed.

Size requirements for both J and CTOD measurements in the transition

region should be established. Constraint loss should be quantified for cases
where these size requirements are violated.

* Sufficient service experience exists to develop a standard for weldment

testing, which is desperately needed.

The nature of scatter in the lower transition region is understood reasonably

well. Further work is necessary to characterize the upper transition region,
where crack growth and constraint loss increase the level of scatter.

* Accurate and reliable correlations between Charpy energy and fracture

toughness are not available.

6.2 APPLICATION TO STRUCTURES

" Many structures behave in a linear elastic manner but it is safer to apply a
more general analysis that corrects for plasticity when necessary.

• Predictions of critical crack size and failure stress are insensitive to the
driving force equation.

" Of the available elastic-plastic driving force relationships, the strip yield
model and the Ainsworth reference stress approach are the most suitable for
welded steel structures, but the strip yield model should not be applied to
high hardening materials.
134

" If the toughness of the material is above a certain level, failure of the
structure is controlled by the flow properties of the material rather than the
fracture properties. Simple approaches to estimate the required toughness
for collapse controlled failure are available.

" Future research should focus on a number of areas, including driving force
in welded structures, residual stress measurements, three-dimensional
effects, crack tip constraint, and gross-section yielding.
 135

7. REFERENCES

1. Harrison, J.D., Garwood, S.J., and Dawes, M.G. "Case Studies and Failure
Prevention in the Petrochemical and Offshore Industry." Presented at
Fracture84, 2nd National Conference on Fracture, Johannesburg, South
Afirica, November 26-28, 1984.

2. Griffith, A.A. -The Phenomena of Rupture and Flow in Solids."

Philosophical Transactions,Series A, Vol. 221, 1920, pp. 163-198.

3. Irwin, G.R., "Onset of Fast Crack Propagation in High Strength Steel and
Aluminum Alloys." Sagamore Research Conference Proceedings, Vol. 2, 1956,
pp. 289-305.

4. Westergaard, H.M., "Bearing Pressures and Cracks." Journal of Applied

Mechanics, Vol. 6, 1939, pp. 49-53.

5. Irwin, G.R., "Analysis of Stresses and Strains near the End of a Crack
Traversing a Plate." Journal of Applied Mechanics, Vol. 24, 1957, pp. 361-364.

6. Williams, M.L, "On the Stress Distribution at the Base of a Stationary Crack."
Journal of Applied Mechanics, Vol. 24, 1957, pp. 109-114.

7. Brown W.F. Jr. and Srawley, J.E., "Plane Strain Crack Toughness
Testing of High Strength Metallic Materials." ASTM STP 410, 1966.

8. Rooke, D.P. and Cartwright, D.J., Compendium of Stress Intensity

Factors. Her Majesty's Stationary Office, London, 1976.

9. Paris, P.C., and Irwin, G.R. The Stress Analysis of Cracks

Tada, -L,
Handbook. Del Research Corporation, Hellertown, Pa, 1973.

10. Murakami, Y. Stress Intensity Factors Handbook. Pergamon Press,

New York, 1987.

11. Irwin, G.R., "Plastic Zone Near a Crack and Fracture Toughness."
Sagamore Research Conference Proceedings, Vol. 4, 1961.

12. Dugdale, D.S., "Yielding in Steel Sheets Containing Slits." Journal of

the Mechanics and Physics of Solids, Vol 8, pp. 100-104.
136

13. -arenblatt,G.I., "The Mathematical Theory of Equilibrium Cracks in

Brittle Fracture." Advances in Applied Mechanics, Vol VII, Academic
Press, 1962, pp. 55-129.

14. Wells, A.A., "Unstable Crack Propagation in Metals: Cleavage and Fast
Fracture." Proceedings of the Crack Propagation Symposium, Vol 1,
Paper 84, Cranfleld, UK, 1961.

15. Rice, J.P. "A Path Independent Integral and the Approximate Analysis
of Strain Concentration by Notches and Cracks." Journal of Applied
Mechanics, Vol. 35, 1968, pp. 379-386.

16. Hertzberg, R.W., Deformation and Fracture of Engineering Materials. Third

Edition, John Wiley & Sons, New York, 1989.

17. Tuba, I.S.., "A Method of Elastic-Plastic Plane Stress and Strain
Analysis." Journal of Strain Analysis Vol. 1, pp. 115-122, 1966, as
referenced by Broek, David, Elementary Engineering Fracture
Mechanics (4th ed.), Kluwer Academic Publishers, Norwell, MA, 1988.

18. Burdekin, F.M. and Stone, D.E.W., "The Crack Opening Displacement
Approach to Fracture Mechanics in Yielding Materials." Journal of Strain
Analysis, Vol. 1, 1966, pp. 144-153.

19. Eshelby, J.D., "The Continuum Theory of Lattice Defects." Solid State Physics,
Vol. 3, 1956.

20. Hutchinson, J.W., "Singular Behavior at the End of a Tensile Crack Tip in a
Hardening Material." Journal of the Mechanics and Physics of Solids, Vol. 16,
1968, pp. 13-31.

21. Rice, J.L and Rosengren, G.F., Plane Strain Deformation near a Crack Tip in a
Power-Law Hardening Material." Journal of the Mechanics and Physics of
Solids, Vol. 16,1968, pp. 1-12.

22. Shih, C.F. "Relationship between the J-Integral and the Crack Opening
Displacement for Stationary and Extending Cracks." Journal of the Mechanics
and Physics of Solids, Vol 29, 1981, pp. 305-326.

23. Anderson, T.L "'heEffect of Crack-Tip Region Constraint on Fracture in the

Ductile-to-Brittle Transition Region." Ph.D. Dissertation, Colorado School of
Mines, Golden, CO, October, 1983.
 137

24. McMeeking, L.M. and Parks, D.M., "On Criteria for J-Dominance of Crack-Tip
Fields in Large-Scale Yielding." ASTM STP 668, American Society of Testing
and Materials, Philadelphia, PA, 1979, pp. 175-194.

25. E399-83, "Standard Test Method for Fracture Toughness of Metallic Materials."
American Society for Testing and Materials, Philadelphia, PA, 1983.

26. E813-87, "Standard Test Method for Jc, a Measure of Fracture Toughness."
American Society for Testing and Materials, Philadelphia, PA, 1987.

27. Wellman, G.W., Sorem, W.A., Dodds, Jr., R.H., and Rolfe, S.T., "Specimen
Thickness Effects for Mastic-Plastic CTOD Toughness of A36 Steel." ASTM STP
945, American Society of Testing and Materials, Philadelphia, PA, 1988 pp. 535-
554.

28. Knott, J.F., "Micromechanisms of Fracture and the Fracture Toughness of

Engineering Alloys, Fracture 1977, Vol 1, ICF4, Waterloo Canada, June 1977,
p. 61.

29. Fractdgraphyand Atlas of Fractographs,Metals Handbook, Eight Edition,

Vol. 9, American Society for Metals, Metals Park, Ohio, 1974.

30. McMahan, C.J. Jr and Cohen, M., "Initiation of Cleavage in Polycrystalline

Iron." Acta Metallurgica, Vol 13, 1965, p. 591.

31. Smith, E., 'The Nucleation and Growth of Cleavage Microcracks in Mild
Steel." Proceedings of the Conference on the Physical Basis of Fracture,
Institute of Physics and Physics Society, 1966, p. 36.

32. Cottrell, A.H., "Theory of Brittle Fracture in Steel and Similar Metals."
Transactions of the ASME, Vol 212, 1958, p. 192.

33. Dolby, R.E. and Knott, J.F., Journal of the Iron and Steel Institute, Vol 210,
1972, p. 857.

34. Gerbererich, W.W. "Metallurgical Aspects of Crack Tip Failure Processes."

ASTM STP 945, American Society of Testing and Materials, Philadelphia, PA,
1988 pp. 5-18.

35. Lin, T., Evans, A.G. and Ritchie, R.O., "Statistical Model of Brittle Fracture by
Transgranular Cleavage." Journal of the Mechanics and Physics of Solids, Vol
34,1986, pp. 477496.

36. Irwin, G.R., Presentation at the meetings of ASTM Committee E24 on Fracture
Testing, Reno, NV, April 1988.
38

37. Ritchie, R.O., Knott, J.F., and Rice, J.R. "On the Relationship between Critical
Tensile Stress and Fracture Toughness in Mild Steel." Journal of the
Mechanics and Physics of Solids, Vol. 21, 1973, p. 395.

38. Curry D.A. and Knott, J.F., "Effect of Microstructure on Cleavage Fracture
Toughness in Mild Steel." Metal Science, Vol. 13, 1979, p. 341.

39. Curry, D.A., "Comparison between Two Models of Cleavage Fracture." Metal
Science, Vol 14, 1980, p.78.

40. Wallin, K, Saario, T., and Tcrronen, K, "Statistical Model for Carbide Induced
Brittle-Fracture in Steel." Metal Science, Vol. 18, 1984, p. 13.

41. Beremin, F.M., "A Local Criterion for Cleavage Fracture of a Nuclear Pressure
Vessel Steel." Metallurgical Transactions,Vol. 14A, 1983, p. 2277.

42. Evans, A.G.. "Statistical Aspects of Cleavage Fracture in Steel.", Metallurgical

Transactions,Vol. 14A, 1983, p. 1349.

43. Anderson, T.L., "A Combined Statistical/Constraint Model for the Ductile-
Brittle Transition Region." ASTM STP 995, American Society of Testing and
Materials, Philadelphia, PA, 1989, pp. 11 -5 6 3 - U- 58 3 .

44. Anderson, T.L and Stienstra, D., "A Model to Predict the Sources and
Magnitude of Scatter in Toughness Data in the Transition Region." Journal of
Testing and Evaluation, Vol. 17, 1989, pp. 46-53.

45. Rice, J.R. and Tracey, D.M., "On the Ductile Enlargement of Voids in Triaxial
Stress Fields." Journal of the Mechanics and Physics of Solids, Vol. 17, 1969,
pp. 201-217.

46. d'Escata, Y. and Devaux, J.C., "Numerical Study of Initiation, Stable Crack
Growth, and Maximum Load with a Ductile Fracture Criterion Based on the
Growth of Holes." ASTM STP 668, American Society of Testing and Materials,
Philadelphia, PA, 1979, pp. 229-248.

47. Devaux; J.C., Rousselier, G., Mudry, and Pineau, "An Experimental Program
for the Validation of Local Ductile Fracture Criteria Using Axisymmetrically
Cracked Bars and Compact Tension Specimens." Engineering Fracture
Mechanics, Vol. 21, 1985, pp. 273-283.

48. Heerens, J. and Read, D.T., "Fracture Behavior of a Pressure Vessel Steel in the
Ductile-to-Brittle Transition Region." NISTIR 88-3099, National Institute for
Standards and Technology, Boulder, CO, December, 1988.
 139

49. Watanabe, J., Iwadate, T., Tanaka, Y., Yokoboro, T. and Ando, K, "Fracture
Toughness in the Transition Region." Engineering Fracture Mechanics,
November 1987.

50. E1152-87 "Standard Test Method for Determining J-R Curves." American
Society for Testing and Materials, Philadelphia, PA, 1987.

51. E1290-89 "Standard Test Method for Crack Tip Opening Displacement
Testing." American Society for Testing and Materials, Philadelphia, PA, 1989.

52. BS5762: 1979, "Methods for Crack Opening Displacement (COD) Testing."
British Standards Institution, 1979.

53. Dawcs, M.G. "Elastic-Plastic Fracture Toughness Based on the COD and
J-Contour Integral Concepts." ASTM STP 668, American Society of Testing
and Materials, Philadelphia, PA, 1979, pp. 306-333.

54. Anderson, T.L., McHenry, H.I. and Dawes, M.G., "Elastic-Plastic Fracture
Toughness Testing with Single Edge Notched Bend Specimens." ASTM STP
856, American Society of Testing and Materials, Philadelphia, PA, 1985. pp.
210-229.

55. Dawes, M.G., Pisarski, H.G. and Squirrell, H.G., "Fracture Mechanics Tests on
Welded Joints" ASTM STP 995, American Society of Testing and Materials,
Philadelphia, PA, 1989, pp.11-191 - 11-213.

56. Satok K. and Toyoda, M., "Guidelines for Fracture Mechanics Testing of
WM/HAZ." Working Group on Fracture Mechanics Testing of Weld
Metal/HAZ, International Institute of Welding, Commission X, 11W
Document X-1113-86.

57. RP 2Z, "Recommended Practice for Preproduction Qualification of Steel Plates

for Offshore Structures." American Petroleum Institute, 1987.

58. Towers O.L. and Dawes, M.G., "Welding Institute Research on the Fatigue
Precracking of Fracture Toughness Specimens." ASTM STP 856, American
Society of Testing and Materials, Philadelphia, PA, 1985. pp. 23-46.

59. Steffen, A.A., Packman, P.F. and Dawes, M.G., "The Effect of Pre-cracking
Variables R and KFMAX on Fracture Toughness." Proceedings of the Seventh
International Conference on Fracture (ICF7), 1989, pp. 1445-1452.

60. Fairchild, D.P., "Local Brittle Zones in Structural Weldments." in Welding

Metallurgy of Structural Steels, The Metallurgical Society of AIME,
Warrendale, PA, 1987.
0

61. Anderson, T.L, "Crack Tip Parameters for Large Scale Yielding and Low
Constraint Configurations." Accepted for publication in International Journal
of Fracture,to appear 1989.

62. Wallin, K, "The Scatter in KIC Results." Engineering Fracture Mechanics,

VoL 19,1984, p. 1085.

63. Slatcher, S., and Evanc:t, 0., "Practical Application of the Weakest-link Model
to Fracture Toughness Problems." Engineering Fracture Mechanics, VoL 24,
1986, p. 508.

64. Stienstra, D., Anderson T.i, and Ringer, UJ., "Statistical Inferences on
Cleavage Fracture Toughness Data", to be published in Journal of Engineering
Materials and Technology.

65. Jones M.I- and Brown, W.F. Jr., "The Influence of Crack Length and
Thickness in Plane Strain Fracture Toughness Tests." ASTM STP 463,
American Society of Testing and Materials, Philadelphia, PA, 1970, p. 63.

66. Landes J.D. and Shaffer, D.H., "Statistical Characterization of Fracture in the
Transition Region." ASTM STP 700, American Society of Testing and
Materials, Philadelphia, PA, 1980, p. 368.

67. Anderson, T.L. and Williams, S., "Assessing the Dominant Mechanism for
Size Effects in the Ductile-to-Brittle Transition Region.", ASTM STP 905,
American Society of Testing and Materials, Philadelphia, PA, 1986, pp. 715-740.

68. Royer, C.P., "A User's Perspective on Heat-Affected Zone Toughness."

Welding Metallurgy of Structural Steels, The Metallurgical Society of AIME,
Warrendale, PA, 1987.

69. Pisarski, H.G. and Pargeter, R.J., "Fracture Toughness of HAZs in Steels for
Offshore Structures." Metal Construction, Vol. 16, 1984, pp. 412-417.

70. Shih, C.F. and German, M.D., "Requirements for a One Parameter
Characterization of Crack Tip Fields by the HRR Singularity." International
Journal of Fracture,Vol. 17, 1981, pp. 27-43.

71. Anderson, T.L, "Ductile and Brittle Fracture Analysis of Surface Flaws Using
CTOD." Experimental Mechanics, June 1988, pp. 188-193.

72. Towers, O.L., Williams, S., and Harrison, J.D., ECSC Collaborative Elastic-
Plastic Fracture Toughness Testing and Assessmen.t Methods." The Welding
Institute Report 3571 /10M/84, June 1984.
 141

73. Wallin, K, "'heEffect of Ductile Tearing on Cleavage Fracture Probability in

Fracture Toughness Testing." Engineering FractureMechanics, in press.

74. Wallin, K, Engineering Fracture Mechanics, in press.

75. Kanninen, M.F. and Popelar, C.H., Advanced FractureMechanics. Oxford

Science Publications, New York, 1985.

76. Naus, D.J., Nanstad, R.K., Bass, B.R., Merkle, J.G., Pugh. C.E., Corwin, W.R.,
and Robinson G.C., "Crack-Arrest Behavior in SEN Wide Plates of Quenched
and Tempered A 533 Grade B Steel Tested under Nonisothermal Conditions."
NUREG/CR-4930, U.S. Nuclear Regulatory Commission and Oak.Ridge
National Laboratory, August 1987.

77. Nakamura, T., Shih, C.F. and Freund, LB., "Analysis of a Dynamically Loaded
Three-Point-Bend Ductile Fracture Specimen." Engineering Fracture
Mechanics, Vol. 25,1986, pp. 323-339.

78. Joyce J.A. and Hacket, E.M., "Dynamic J-R Curve Testing of a High Strength
Steel Using the Multispecimen and Key Curve Techniques." ASTM STP 905,
American Society of Testing and Materials, Philadelphia, PA, 1984, pp. 741-774.

79. Joyce J.A. and Hacket, E.M., "An Advanced Procedure for J-R Curve Testing
Using a Drop Tower." ASTM STP 995, American Society of Testing and
Materials, Philadelphia, PA, 1989.

80. Barsom, J.M., "Development of the AASHTO Fracture Toughness

Requirements for Bridge Steels." Engineering Fracture Mechanics, Vol. 7, 1975,
pp. 605-618.

81. E1221-88, "Standard Method for Determining Plane-Strain Crack-Arrest

Toughness, KIa, of Ferritic Steels." American Society of Testing and Materials,
Philadelphia, PA, 1988.

82. Kumar, V., German, M.D., and Shih, C.F.,"An Engineering Approach for
Elastic-Plastic Fracture Analysis." EPRI Report NP-1931, Electric Power
Research Institute, Palo Alto, CA, 1981.

83. Anon, "Brittle-Ductile Transition of Bridge Steels." Request for Proposal No.
DTFH-61-86-R-00028, Federal Highway Administration, Washington, D.C.,
November 1985.

84. Pisarski, H.G., "A Review of Correlations Relating Charpy Energy to KIC." The
Welding Institute Research Bulletin, December, 1978, pp. 362-367.
142

85. Marandet, B. and Sanz, G., "Evaluation of the Toughness of Thick Medium
Strength Steels by LEFM and Correlations Between KIC and CVN." ASTM STP
631, American Society of Testing and Materials, Philadelphia, PA, 1977, pp. 72-
95.

86. Rolfe, S.T. and Novak, S.T., "Slow Bend ICC Testing of Medium Strength
High Toughness Steels." ASTM STP 463, American Society of Testing and
Materials, Philadelphia, PA, 1970, pp. 124-159.

87. Barsom, J.M. and Rolfe, S.T., "Correlation Between KIC and Charpy V Notch
Test Results in the Transition Temperature Range." ASTM STP 466, American
Society of Testing and Materials, Philadelphia, PA, ! 97, pp. 281-301.

88. Sailors, R.H. and Corten, H.T., "Relationship between Material Fracture
Toughness Using Fracture Mechanics and Transition Temperature Tests."
ASTM STP 514, American Society of Testing and Materials, Philadelphia, PA,
1973, pp. 164-191.

89. Begley, J.A. and Logsdon, W.A., "Correlation of Fracture Toughness and
Charpy Properties for Rotor Steels." Westinghouse Report, Scientific Paper 71-
IE7, MSLRF-P1-1971.

90. Ito, T., Tanaka, K. and Sato, M. "Study of Brittle Fracture Initiation from
Surface Notch in Welded Fusion Line." 11W Document X-704-73, September
1973.

91. Wellman, G.W. and Rolfe, S.T., "Engineering Aspects of Crack-Tip Opening
Displacement Fracture Toughness Testing." ASTM STP 856, American Society
c,. Testing and Materials, Philadelphia, PA, 1985, p. 230.

92. Dolby, R.E., "Some Correlations between Charpy V and COD Test Data for
Ferritic Weld Metals." The Welding Institute Report 109/1980, April 1980.

93. Norris, D.M., Reaugh, J.E., and Server, W.L, "A Fracture-Toughness
Correlation Based on Charpy Initiation Energy." ASTM STP 743, American
Society of Testing and Materials, Philadelphia, PA, 1981, p.2 0 7 .
94. Anderson, T.L. and Zapata, J.E. "A Theoretical Charpy-Fracture Toughness
Correlation for the Transition Region." to be published.

95. Anderson, T.L, unpublished work.

%. Garwood, S.J., A Crack-Tip Opening Displacement Method for the Analysis of

Ductile Materials, ASTM STP 945, American Society of Testing and Materials,
Philadelphia, PA, 1988, pp. 957-985.
 143

97. Burdekin, F.M. and Dawes, M.G., "Practical Use of Linear Elastic and Yielding
Fracture Mechanics with Particular Reference to Pressure Vessels."
Proceedings of the Institute of Mechanical Engineers Conference, London, May
1971, p.28.

98. PD6493:1980, "Guidance on Some Methods for the Derivation of Acceptance

Levels for Defects in Fusion Welded Joints." British Standards Institution,
March 1980.

99. Kamath, M.S., 'he COD Design Curve: An Assessment of Validity Using
Wide Plate Tests." The Welding Institute Report 71/1978/E, September 197&

100 Harrison, R.P., Loosemore, K, Milne, I, and Dowling, AR., "Assessment of the
Integrity of Structures Containing Defects." Central Electricity Generating
Board Report R/H/R6-Rev 2, April 1980.

101. Heald, P.T., Spink, G.M., and Worthington, P.J., Post Yield Fracture
Mechanics." Materials Science and Engineering, Vol. 10, 1972, Vol. 10,
pp. 129-137.

102. Shih, C.F. and Hutchinson, J.W., "Fully Plastic Solutions and Large-Scale
Yielding Estimates for Plane Stress Crack Problems." Journal of Engineerink
Materials and Technology, Vol. 98, 1976, pp. 289-295.

103. Bloom, J.M., "Prediction of Ductile Tearing Using a Proposed Strain

Hardening Failure Assessment Diagram." International Journal. of Fracture,
Vol. 6., 1980, pp. R73-R77.

104. Shih, C.F., Kumar, and German, M.D., "Studies on the Failure Assessment
Diagram Using the Estimation Scheme and J-Controlled Crack Growth
Approach." ASTM ST 803, American Society of Testing and Materials,
Philadelphia, PA, 1988, pp.1-239 -11-261.

105 Kumar, V., German, M.D., Wilkening, W.W., Andrews, W.R., deLorenze,
R.G., and Mowbray, D.F., "Advances in Eastic-Plastic Fracture Analysis." EPRI
Report NP-3607, Electric Power Research Institute, Palo Alto, CA, 1984.

106. Kumar, V. and German, M.D., "Elastic-Plastic Fracture Analysis of Through-

Wall and Surface Flaws in Cylinders." EPRI Report NP-5596, Electric Power
Research Institute, Palo Alto, CA, 1988.

107. Zahoor, A. "Ductile Fracture Handbook, Volume 1: Circumferential

Throughwall Cracks." EPRI Report NP-6301-D, Electric Power Research
Institute, Palo Alto, CA, 1989.
144
108. Bloom, J.M., Simplified Procedures for Handling Self-Equilibrating Secondary
Stresses in the Deformation Plasticity Failure Assessment Diagram Approach."
ASTM STP 995, American Society for Testing and Materials, Philadelphia, PA,
1989, pp.11-280 -11-305.
109. Ainsworth, LA., "The Assessment of Defects in Structures of Strain
Hardening Materials." Engineering FractureMechanics, Vol. 19, 1984, p. 633.

110. Anderson, T.L., Leggatt, R.H., and Garwood, S.J., "The Use of CTOD
Methods in Fitness for Purpose Analysis." The Crack Tip Opening
Displacement in Elastic-Plastic Fracture Mechanics, Springer-Verlag,
Berlin, 1986, pp. 281-313.

111. Garwood, S.J., Willoughby, A.A., Leggatt RH., and Jutla, T., "Crack Tip
Opening Displacement (CTOD) Methods for Fracture Mechanics Assessments:-
Proposals for Revisions to PD 6493." Presented at ASFM 6, Ispra, Italy, October
1987.

112. Burdekin, F.M., Garwood, S.J., and Milne, I, "The Basis for the Technical
Revisions to the Fracture Clauses of PD 6493." Presented at the International
Conference on Weld Failures, London, November 1988.

113. Newman, J.C and Raju, I.S., 'Stress Intensity Factor Equations for Cracks in
Three Dimensional Finite Bodies." ASTM STP 791, American Society of
Testing and Materials, Philadelphia, PA, 1983, pp. 238-265.

114. Ainsworth, LA., "The Treatment of Thermal and Residual Stresses in

Fracture Assessments." Central Electricity Generating Board Report
TPRD/0479/N84, 1984.

115. Milne, I., Ainsworth, ILA., Dowling, A.RL, and Stewart, A.T., "Assessment of
the Integrity of Structures Containing Defects." Central Electricity Generating
Board Report R/H/R6-Rev 3, May 1986.

116. Bain, L.J., Statistical Analysis of Reliability and Life-Testing Models, Marcel
Dekker, Inc. New York, 1978.

117. Hutchinson, J.W. and Paris, P.C., "Stability Analysis of J-Controlled Crack
Growth." in Elastic-Plastic Fracture Mechanics, ASTM STP 668, American
Society for Testing and Materials, Philadelphia, PA, 1979, pp. 37-64.

118. Anderson, T.L, Gordon, J.RL and Garwood, S.J.,"On the Application of R-
Curves and Maximum Load Toughness to Structures." ASTM STP 969,
American Society for Testing and Materials, Philadelphia, PA, 1988, pp. 291-
317.
 145

119. Parks, D., "A Surface Crack Review. Elastic and Plastic Behavior." Proc.
Symposium on Surface-Crack Growth: Models, Experiments, and
Structures. Reno, NV, April 1988.

120. Reuter, W.G. and Lloyd, W.R., "Measurements of CTOD and CTOA
Around Surface-Crack Perimeters and Relationships Between Elastic
and Plastic CTOD Values." Proc. Symposium on Surface-Crack Growth:
Models, Experiments, and Structures. Reno, NV, April, 1988.

121. Read, D.T., "Applied J-Integral in HY130 Tensile Panels and

Implications for Fitness for Service Assessment." Report NBSIR 82-
1670, National Bureau of Standards, Boulder, CO, 1982.

122. Ernst, H.A., "Material Resistance and Instability Beyond J-Controlled Crack
Growth." ASTM 51P 803, American Society for Testing and Materials,
Philadelphia, PA, 1983, pp. 1-191 - 1-213.

u.s. Govj M PAp916T1GOW'ICE 1991/519-S27/21167

 SHIP STRUCTURE COMMITTEE PUBLICATIONS

SSC-332 Guide for Ship Structural Inspections by Nedret S. Basar

& Victor W. Jovino 1985

SSC-333 Advance Methods for Ship Motion and Wave Load Prediction
by James C. Oliver, 1989

SSC-334 Influence of Weld Porosity on the Integrity of Marine

Structures by William J. Walsh , Brian N. Leis, and J. Y.
Yung, 1989

SSC-335 Performance of Underwater Weldments by R. J. Dexter, E. B.

Norris, W. R. Schick, and P. D. Watson 1986

SSC-336 Liquid Sloshing in Cargo Tanks by N. A. Hamlin 1986

SSC-337 Part 1 - Ship Fracture Mechanisms Investigation by Karl

A. Stambaugh and William A. Wood 1987

SSC-337 Part 2 - Ship Fracture Mechanisms - A Non-Expert's Guide

for Inspecting and Determining the Causes of Significant
Ship Fractures by Karl A. Stambaugh and William A. Wood
1987

SSC-338 Fatigue Prediction Analysis Validation from SL-7 Hatch

Corner Strain Data by Jen-Wen Chiou and Yung-Kuang Chen
1985

SSC-339 Ice Loads and Ship Response to Ice - A Second Season by

C. Daley, J. W. St. John, R. Brown, J. Meyer, and I. Glen
1990

SSC-340 Ice Forces and Ship Response to Ice - Consolidation Report

by C. Daley, J. "° St. John, R. Brown, and I. Glen 1990

SSC-341 Global Ice Forces and Ship Response to Ice by P. Minnick,

J. W. St. John, B. Cowper, and M. Edgecomb 1990

SSC-342 Global Ice Forces and Ship Response to Ice - Analysisof

Ice Ramming Forces by Yung-Kuang Chen, Alfred L. Tunik,
and Albert P-Y Chen 1990

SSC-343 Global Ice Forces and Ship Response to Ice - A Second

Season by P. Minnick and J. W. St. John 1990

SSC-344 Development of an Onboard Strain Recorder by Eric Greene

and William A. Wood 1987

None Ship Structure Committee Publications - A Special

Bibliography 1983
 COMMITTEE ON MARINE STRUCTURES

Commission on Engineering and Technical Systems

National Academy of Sciences - National Research Council

The COMMITTEE ON MARINE STRUCTURES has technical cognizance over

the interagency Ship Structure Committee's research program.

Stanley G. Stiansen (Chairman), Riverhead, NY

Mark Y. Berman, Amoco Production Company, Tulsa, OK
Peter A. Gale, Webb Institute of Naval Architecture, Glen Cove, NY
Rolf D. Glasfeld, General Dynamics Corporation, Groton, CT
William H. Hartt, Florida Atlantic University, Boca Raton, FL
Paul H. Wirsching, University of Arizona, Tucson, AZ
Alexander B. Stavovy, National Research Council, Washington, DC
Michael K. Parmelee, Secretary, Ship Structure Committee,
Washington, DC

LOADS WORK GROUP

Paul H. Wirsching (Chairman), University of Arizona, Tucson, AZ

Subrata K. Chakrabarti, Chicago Bridge and Iron Company, Plainfield, IL
Keith D. Hjelmstad, University of Illinois, Urbana, IL
Hsien Yun Jan, Martech Incorporated, Neshanic Station, NJ
Jack Y. K. Lou, Texas A & M University, College Station, TX
Naresh Maniar, M. Rosenblatt & Son, Incorporated, New York, NY
Solomon C. S. Yim, Oregon State University, Corvallis, OR

MATERIALS WORK GROUP

William H. Hartt (Chairman), Florida Atlantic University, Boca Raton, FL

Fereshteh Ebrahimi, University of Florida, Gainesville, FL
Santiago Ibarra, Jr., Amoco Corporation, Naperville, IL
Paul A. Lagace, Massachusetts Institute of Technology, Cambridge, MA
John Landes, University of Tennessee, Knoxville, TN
Mamdouh M. Salama, Conoco Incorporated, Ponca City, OK
James M. Sawhill, Jr., Newport News Shipbuilding, Newport News, VA