STP 559-1974

FRACTURE TOUGHNESS AND
SLOW-STABLE CRACKING
Proceedings of the 1973

National Symposium on
Fracture Mechanics, Part I
A symposium sponsored by
Committee E-24 on
Fracture Testing of Metals,
AMERICAN SOCIETY FOR
TESTING AND MATERIALS
University of Maryland,
College Park, Md., 27-29 Aug. 1973
ASTM SPECIAL TECHNICAL PUBLICATION 559

P. C. Paris, chairman of symposium committee
G. R. Irwin, general chairman of symposium
List price $25.25

04-559000-30
~(~l~ AMERICAN SOCIETY FOR TESTING AND MATERIALS

1916 Race Street, Philadelphia, Pa. 19103
(~) AMERICAN SOCIETY FOR TESTING AND MATERIALS 1974
Library of Congress Catalog Card Number: 7 4 - 8 1 1 5 4
NOTE
Thc Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.
Printed in Baltimore, Md.

August 1974
Foreword
The 1973 National Symposium on Fracture Mechanics was held at the

University of Maryland Conference Center, College Park, Md., 27-29
Aug. 1973. The symposium was sponsored by the American Society for
Testing and Materials through Committee E-24 on Fracture Testing of
Metals. Members of the Symposium Subcommittee of Committee E-24
selected papers for the program. Organizational assistance from Don
Wisdom and Jane Wheeler at ASTM Headquarters was most helpful.
G. R. Irwin, Dept. of Mechanical Engineering, University of Maryland,
served as general chairman. Those who served as session chairmen were
H. T. Corten, Dept. of Theoretical and Applied Mechanics, University of
Illinois; C. M. Carman, Frankford Arsenal; J. R. Rice, Div. of Engineer-
ing, Brown University; D. E. McCabe, Research Dept., ARMCO Steel;
J. E. Srawley, Fracture Section, Lewis Research Center, NASA; E. T.
Wessel, Research and Development Center, Westinghouse Electric Corp.;
and E. K. Walker, Lockheed-California Co.
The Proceedings have been divided into two volumes: Part 1--Fracture
Toughness and Slow-Stable Cracking and Part II--Fracture Analysis.
Related
ASTM Publications
Stress Analysis and Growth of Cracks, STP 513 (1972), $27.50
04-513000-30
Fracture Toughness, STP 514 (1972), $18.75 04-514000-30
Fracture Toughness Evaluation by R-Curve Methods, STP 527 (1973),
$9.75 04-527000-30
Progress in Flaw Growth and Fracture Toughness Testing, STP 536
(1973), $33.25 04-536000-30
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 13:10:55 EST 2015
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized
Contents
Introduction 1
Fracture Toughness Test Methods for Abrasion-Resistant White Cast

Irons Using Compact Specimens--D. E. DIESBUR6 3
Experimental Procedure 5
Results and Discussion 6
Conclusions 13
Development of Fast Fracture in a Low Alloy Steel--J. c. RADON AND

A. A. POLLOCK 15
Material, Specimen Geometry, and Fracture Mechanics Results 16
Acoustic Emission 19
Fractography 25
Discussion 27
Conclusions 29
Acoustic Emission from 4340 Steel During Stress Corrosion Cracking--

H. H. CHASKELIS, W. H. CULLEN, AND J . M. KRAFFT 31
Experimental Procedures 32
Data Reduction 35
Comparison of Acoustic Emission to Crack Growth 36
Comparison of Acoustic Emission to the Fracturing Process 36
Discussion 42
Conclusions 43
Effects of Shot-Peening Residual Stresses on the Fracture and Crack-

Growth Properties of D6AC Steel--WOLF ELBER 45
Nomenclature 45
Residual Stress Model 47
Experiments 50
Conclusion 57
Fracture Properties of a Cold-Worked Mild Steel--E, J. RIFLING 59

Materials and Procedure 60
Conclusions 72
More on Specimen Size Effects in Fracture Toughness Testing~J. 6.

KAUFMANAND F. G. NELSON 74
Material 75
Procedure 75
Results 79
Discussion 79
Implications in Size Requirements 83
Value of Kmax 85
Summary 85
Dynamic Compact Tension Testing for Fracture Toughness---P. c. pARIS,
R. 3". BUCCI~ AND L. L. LOUSltIN 86
Alloys 88
Experimentation 89
Equipment and Instrumentation 91
Test Records 93
Analysis of Test Records 96
Summary 98
Further Aspects of Fracture Resistance Measurement on Thin Sheet Ma-
terial: Yield Stress and Crack LengthwA. M. SULLIVAN ANt)
STOOP 99
Experimental Parameters 101
Degradation of K e with Increased Yield Strength 103
Relationship Between Initial and Final Crack Length 104
Screening Specimen 109
Conclusions 109
Further Studies of Crack Propagation Using the Controlled Crack Propa-
gation Approach--R. H. WEITZMANN AND I. FINNIE 111
General Test Aspects 124
Conclusions 125
Double Torsion Technique as a Universal Fracture Toughness Test
Method--J. o. OUTWATER, M. C. MURPHY, R. G. KUMBLE, AND
J. T. BERRY 127
Fundamental Concepts of Fracture Mechanics 128
Theoretical and Experimental Limitations 129
Theoretical Basis for the Double Torsion Technique 130
Experimental Procedure and Results 133
Discussion 135
Conclusions 136
Author's Note Added in Proof 137
Measurement of KIe on Small Specimens Using Critical Crack Tip Open-
ing Displacement--J. N. ROBINSONAND A. S. TETELMAN 139
Summary 156
Correlation Between Fatigue Crack Propagation and Low Cycle Fatigue

Properties---SAURINDRANATH MAJUMDAR AND JODEAN MORROW 159
Nomenclature 159
Description of the Fatigue Crack Propagation Model 163
Mechanics and Fatigue Analysis of the Crack Tip Region 164
Influence of Material Properties on the Coefficient of Eq 13 168
Effect of Microstructure Size 169
Comparison of Eq 14 with Barsom's Data on Steel 172
Conclusions 173
Effect of Stress Concentration on Fatigue-Crack Initiation in HY-130
Steel--J. M. BARSOM AND R. C. M e NICOL 183
Materials and Experimental Work 184
Summary 199
Evaluation of the Fatigue Crack Initiation Properties of Type 403 Stain-
less Steel in Air and Steam Environments---w. G. CLARK, JR. 205
Material 206
Experimental Results 211
Crack Initiation Data 215
Discussion 218
Summary 222
Conclusions 223
Subcritical Crack Growth Under Single and Multiple Periodic Overloads
in Cold-Rolled S t e e l ~ F . H. GARDNER AND R. I. STEPHENS 225
Material and Experimental Procedures 227
Test Results 229
Discussion of Results 239
Conclusions 242
Effect of Mean Stress Intensity on Fatigue Crack Growth in a 5456-
H l 1 7 Aluminum AlloymH. P. CHU 245
Conclusions 259
Discussion 261
Effects of R-Factor and Crack Closure on Fatigue Crack Growth for
Aluminum and Titanium Alloys--M. KATCHER AND M. KAPLAN 264
Nomenclature 264
Results 270
Discussion 277
Conclusions 280
Proposed Fracture Mechanics Criteria to Select Mechanical Fasteners
for Long Service Lives---A. F. GRANDT, J R . AND 3". P. GALLAGHER 283
Approach 284
Stress Intensity Factor Analysis 285
Application of the Linear Superposition Method to the Fastener
Hole Problem 287
Interference Fit Fasteners 289
Achieving a Fatigue Stress Intensity Threshold 290
Implications to Fastener Design 294
Summary 295
Rapid Calculation of Fatigue Crack Growth by lntegration--T. R.
~RUSSAT 298
The Existing Summation Procedure 299
The Proposed Integration Procedure 300
Application and Comparison 307
Conclusions 310
STP559-EB/Aug. 1974
Introduction
Readers of this volume will not be disappointed with regard to novelties

of current and practical interest in fracture toughness and slow-stable
cracking. These range from unusual test methods to a puzzling effect
of lateral specimen dimensions on Kic values for an aluminum alloy.
Observational techniques include acoustic emission, both in relation to
onset of rapid fracture and stress corrosion cracking, tape recordings as
an assist for rapid load testing, and use of rubber castings to verify
measurements of crack opening stretch. Toughness measurements are
reported for white cast irons and cold-rolled steel. The papers dealing with
fatigue cracking include a low cycle fatigue viewpoint on fatigue crack
growth, effects of shot peening, initiation of fatigue cracking as a function
of notch root radius, as well as effects of overloads, mean K, and mechani-
cal fastener pressure.
The development of technology in this field has prospered over the
years so that often novel approaches soon become routine techniques to
solving problems. This volume is another contribution to the engineer
and metallurgist faced with fracture problems.
With two exceptions, all of the papers in this volume were presented
at the 1973 National Symposium on Fracture Mechanics held at the
College Park campus of the University of Maryland, 2 7 - 2 9 Aug. 1973.
The two exceptions were a paper offered for this symposium but not
presented and a late submission of a paper from the 1972 symposium.
The companion volume, STP 560, covers fracture analysis.
G. R. Irwin
Department of Mechanical Engineering,

University of Maryland,
College Park, Md.
Copyright*1974 by ASTMInternational www.astm.org

D. E. Diesburg 1
Fracture Toughness Test Methods for

Abrasion-Resistant White Cast Irons
Using Compact Specimens
REFERENCE: Diesburg, D. E., "Fracture Toughness Test Methods for

Abrasion-Resistant White Cast Irons Using Compact Specimens," Frac-
ture Toughness and Slow-Stable Cracking, A S T M STP 559, American
Society for Testing and Materials, 1974, pp. 3-14.
ABSTRACT- The fracture toughness of abrasion-resistant white cast irons
has been measured, using precracked compact specimens. Some procedures
used for precracking the brittle cast irons were outside the ASTM Test for
Plane-Strain Fracture Toughness of Metallic Materials (E 399-72)
requirements but still gave valid results. The excellent reproducibility,
combined with a range in toughness values of 17.5 to 28.5 ksiVTff. (19.2 to
31.4 M N / m 8/~) for abrasion-resistant white cast irons, provided the sensi-
tivity necessary to distinguish differences in the toughness of white cast irons
resulting from variations in composition or microstructure. The fracture
toughness of three commonly used irons, 27Cr, 9Cr-6Ni, and 20Cr-2Mo-
1Cu, was compared in the as-cast (and stress-relieved) condition. Heat
treating the 20Cr-2Mo-lCu iron substantially increased the hardness and
reduced the fracture toughness slightly.
KEY WORDS: abrasion-resistant iron, white cast iron, fracture properties,
toughness, evaluation, mechanical tests, fatigue (materials), mechanical
properties.
T h e mining i n d u s t r y uses a b r a s i o n - r e s i s t a n t m a t e r i a l s in large q u a n t i t y

in various stages of o r e beneficiation. T h e m a t e r i a l s u s e d for jaw crusher
plates, c r u s h e r liners, mill liners, grinding balls, c o n v e y i n g systems, a n d
slurry p u m p s all m u s t be able to w i t h s t a n d a b r a s i v e wear. T h e w e a r rate
in these pieces of e q u i p m e n t contributes significantly to the e c o n o m i c s of
m i n i n g o p e r a t i o n s . U n f o r t u n a t e l y , the m o s t a b r a s i o n - r e s i s t a n t m a t e r i a l s
1 Senior research associate, Ann Arbor Research Laboratory of Climax Molyb-

denum Company of Michigan, a subsidiary of American Metal Climax, Inc.
Copyright9 1974 by ASTM lntcrnational www.astm.org

4 FRACTURETOUGHNESS,AND SLOW-STABLE CRACKING
also tend to be brittle. Not only must the processing equipment withstand
abrasion, but it must do so without fracturing during service. Therefore,
in every application where abrasion is a factor, the material used must
provide abrasive wear resistance and adequate toughness.
Research laboratories have been able to determine the wear rates of
many materials under various types of abrasive environments, but the
fracture resistance of these materials has been difficult to evaluate before
the material is placed into service. Although abrasion-resistant materials
are all relatively brittle, lack of toughness may shorten the service life
by an amount depending on the application. Extensive investigations of
fractured components has led to the conclusion that there can be a large
difference in fracture resistance in materials, which, when tested in a
Charpy impact testing machine, may exhibit less than 2 ft-lb (0.3
kgfm/cm2). Before a research laboratory can investigate the metallurgical
parameters providing the best fracture resistance, it is first necessary to
devise a test method that can reproducibly provide enough sensitivity to
distinguish the various levels of toughness that can be present in materials
that fail at low levels of absorbed impact energy.
Plane-strain fracture toughness is a measure of fracture resistance in
the early stages of crack propagation and has been successfully measured
for a cast steel 2 and gray and ductile cast irons?, ~ Fracture toughness
measurements are sensitive to changes in microstructure, which is exactly
the type of measurement needed to investigate the factors controlling the
toughness of white cast irons.
The method of testing for the plane-strain fracture toughness of metallic
materials (ASTM Test for Plane-Strain Fracture Toughness of Metallic
Materials ( E 3 9 9 - 7 2 ) ) provided the basis for the development of a test
that can reproducibly provide enough sensitivity to measure the toughness
of various abrasion-resistant white cast irons. The attempt was made to
follow exactly the method outlined in ASTM E 399-72, but it was soon
realized that a few of the specifications could be relaxed and still provide
a valid measurement of plane-strain fracture toughness for these brittle
irons. This paper (1) describes the development of the test procedure
used to measure K~e in white cast irons and (2) cites a few examples
to illustrate the spread in values that can be expected.
2 Greenberg, H. D. and Clark, W. G., Jr., Metals Engineering Quarterly, American

Society for Metals, Aug. 1969, pp. 30-39.
3Glover, A. G. and Pollard, G., lournal of the Iron and Steel Institute, Feb. 1971,
pp. 138-141.
"Lazaridis, A., Worzala, F. J., Loper, C. R., and Heine, R. W., Transactions,
American Foundrymen's Society, 1971, Vol. 79, pp. 351-360.
DIESBURG ON ABRASION-RESISTANT WHITE CAST IRONS 5
Experimental Procedure
The irons used to establish the test procedure were available from
previous laboratory investigations and, in many cases, were not in the
recommended heat-treated conditions. These available irons will be referred
to as Irons I, II, III, IV, V, and VI and are further described in the
appendix.
The white cast irons (compositions given in Table 1 ), used to illustrate
typical values of fracture toughness expected of white cast irons in the
properly heat-treated condition, had been cast into baked-sand molds as
1-in. (25-mm) thick plates from 125-1b (57-kg) induction-melted heats.
The irons were tested in the as-cast plus stress-relieved condition. The
20Cr-2Mo-lCu iron was also tested after the matrix microstructure had
been changed from austenite to predominantly martensite through heat
treatment.
Specimen Preparation
Compact test ( C T ) specimens with dimensions as shown in Fig. 1 were
prepared for each iron. The outer dimensions were obtained by grinding,
while electrical discharge machining ( E D M ) was used to prepare the
pin loading holes and the crack initiating notches. The notch was machined
in two steps, the final step producing a 0.01-in. (0.3-mm) wide by 0.09-in.
(2.3-mm) deep slot. A fatigue crack was grown at the base of the 0.01-in.
(0.3-mm) wide slot using an SF-1U Sonntag fatigue testing machine with
a loading cycle that always kept the specimen loaded in tension. The
R-values (ratio of minimum to maximum load) were always less than 0.1.
The crack length was measured on both broad surfaces of the CT speci-
mens (polished through 600-grit paper).
Testing
Once the specimens had been precracked, the load required to extend
the crack was determined by pulling the specimen in a tension testing
machine. The same clevis fixtures used on the fatigue machine were used
to load the specimens in the tension testing machine.
TABLE 1--Chemical analysis of the irons.
Element, weight percent

Iron C Mn Si Ni Cr Mo Cu P S
27Cr 2.68 0 . 8 3 0.36 ... 27.60 ...... 0.009 0.008
9Cr-6Ni 3.42 0.67 1.61 6.17 8.64 . . . . . . 0.029 0.036
20Cr-2Mo-lCu 2.89 0.73 0.55 ... 19.35 2.08 1.02 0.015 0.016
6 FRACTURETOUGHNESS AND SLOW-STABLE CRACKING
L = 2.12 IN.
/
(54.0 mm) ]
W = 1.62 I N . . . J
0.01 IN.
(41.2 mm)
[ h = 0.50 FN.
(0,3 mm) - - - ~ a ~--(12.7 mini
i
1.2W = 1.95
IN.
.25 IN. (6.4 rnml
~ . ~ _ ~ j ~ 0.09 IN.
(2.3 ram)
(49.5 ram)
L ~ - "CLIP GAGE SUPPORT
0.38 IN.
(9.5 ram) DIA.
ii i
',I I I ~(12.7 mm~
FIG. 1--Compact specimen.
A cantilever extensometer, calibrated to meet the requirements of

ASTM, was used to monitor crack extension. The extensometer was
clipped between two parallel knife edges (clip gage support, Fig. 1),
which had been machined separately and attached to the specimen with
LocTite 310 metal bonding adhesive. A special fixture was used to hold
the knife edges in a parallel position until the adhesive cured. The strength
of the, adhesive bond was proven to be sufficient by clipping the extenso-
meter between the knife edges of one of the specimens for a duration of
24 h. No extensometer movement, and therefore no slippage, was observed
under the load applied by the extensometer.
The precracked specimens were installed in the tension testing machine
and the extensometer was attached. A crosshead rate of 0.005 in./min
(0.13) ram/rain) provided a loading rate of approximately 700 lb/min
(320 kgf/min). A typical load versus crack displacement curve is shown
in Fig. 2. Oftentimes Po was found to be Pr..x (Fig. 2).
Results and Discussion

An experimental procedure was developed for measuring the plane-
strain fracture toughness of brittle white cast irons. An attempt was made
DIESBURG ON ABRASION-RESISTANTWHITE CAST IRONS 7
EXTENSIONACROSSKNIFEEDGES(MM)
1.0 2.0 3.0
1800 I I
I j~.-EXTENSIONOF ELASTICREGION-~800
I
l/ //~ L I N E HAVING S~'oLESS SLOPE /
1500 k /f THANELASTICREGION I
J lt,,'~~''-- Pmax l
-tooo
0I//
I/
\
\ 200 /
0.03 0.06 0.09

EXTENSIONACROSSKNIFEEDGES(IN.)
FIG. 2--Typical load versus extension curve used'to calculate [racture toughness
of white irons.
to meet all of the requirements outlined by ASTM E 399-72. The

following (1) discusses which ASTM requirements were not met in the
experimental procedure and (2) reports the fracture toughness of three
commonly used abrasion-resistant white cast irons.
Test R e q u i r e m e n t s
The requirements outlined by ASTM E 399-72 can be divided into
three main categories: (1) specimen geometric requirements, (2) pre-
cracking requirements, and (3) testing requirements. All specimens
did meet the dimensional requirements: B_> 2.5 (K~e/~ys) 2, fatigue crack
c was at least 0.05 in. (1.3 mm) in length and greater than 5 percent of
L, and B was at least 0.25 W but less than W.~ The testing requirements
were also met. The requirements that were not always met involved the
growth of the fatigue crack.
The difficulty in controlling the growth of the fatigue crack often
resulted in cracks that did not quite meet specifications. The lowest stress
5B, c, a, L, and W are definedin Fig. 1.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
intensity required to propagate a fatigue crack was usually very close

to the maximum limit of 0.6 of the stress intensity, KQ, routinely deter-
mined in fracture toughness testing. Often Kf(m,~.~, the maximum stress
intensity used to grow the last portion of the fatigue crack, exceeded 0.6
KQ and reached as high as 0.9 K~. A comparison of fracture toughness
values (KQ) with the values obtained from specimens precracked with a
stress intensity less than 0.6 KQ (Table 2), indicated that Kf(.... ~ could
be as large as 0.84 KQ (Iron VI) without altering the measured fracture
toughness of brittle white cast irons.
The fatigue crack path in white cast irons was very dependent on
dendrite orientation. Nonrandom orientation of dendrites could cause the
paths to deviate from the plane of symmetry for the specimen. Fortunately,
the resulting fracture toughness determinations were not affected, even
when the angle from symmetry reached 20 deg (Iron VI in Table 2,
Specimen 4).
ASTM recommends determining the length of the fatigue crack by
taking measurements from the fractured surface. However, it was very
difficult to make precise measurements of the fatigue crack on the frac-
tured surface because there was usually no distinct boundary between the
fatigued surface and the fractured surface. It was observed, however, that
the crack front usually formed a linear boundary between the surface
traces. This observation was made by exposing five precracked specimens
to moisture to form a slight layer of rust, which clearly outlined the loca-
tion of the crack tip prior to testing. The crack lengths on subsequent
specimens were measured along the two surface traces of the fatigue crack
and then averaged. The nonrandom cast orientation sometimes resulted in
nonuniform crack lengths. About one sample in six could be expected to
have one side precracked to a length less than 90 percent of the average
crack length. Again, the resulting fracture toughness determinations were
not affected (Iron III in Table 2).
The shortest fatigue crack length was 0.18 in. (4.6 mm), which was
substantially greater than the required 5 percent of L. The distance a
from the crack tip to the loading plane was greater than 0.45 W for all
specimens. However, there were two specimens (Iron VI, Specimens 2
and 3, in Table 2) for which this distance was greater than the maximum
limit of 0.55 W, as set by ASTM E 399-72. The effect of this extra
length crack could not be detected. The difference between the KQ obtained
from the specimen having the correct notch depth a and the specimen
having a greater than 0.55 W was 0.9 and 8.6 percent. The fact that
one specimen with a long crack length gave a consistent KQ indicates
that the difference of 8.6 percent was not caused by a exceeding 0.55 W.
T A B L E 2--Fracture toughness o[ white irons.
Average Fatigue
KIe~
Crack Growth Rate Crack KQ,
of Last 0.050 in. Length a, ksiVin__ Kr ksi x/in.
Iron Condition ~ Specimen (1.3 m m ) , mm/cycle in. (mm) Remarks b (MN/m~/0 KQ ( M N / m 3/2)
I HT 1 0.013 0.765 (19.4) A 23.5 (25.8) 0.48 23.5 (25.8)

2 0.013 0.750 (19.0) 24.6 (27.1) 0.44 24.6 (27.1)
II AC 1 0.026 0.816 (20.7) A 20.1 (22.1) 0.69 20.1 ~ (22.1) ~
2 0.163 0.848 (21.5) A, B 20.4 (22.4) 0.66 20.4 ~ (22.4) ~
o
3 0.011 0.793 (20.1) C,D 21.7 (23.9) 0.63
4 0.130 0.797 (20.2) A,B 20.9 (23.0) 0.66 20.9 c (23.0) c
t-
5 0..062 0.810 (20.6) B,C,D 21.3 (23.4) 0.70
8 0.433 0.780 (19'.8) A,B 21.0 (23.1) 0.65 21.0 ~ '(33. I )~
O
IlI HT 1 0.077 0.767 (19.5) A,B 23.8 (26.2) 0.49 23.8 (26.2) z
2 0.006 0.782 (19.9) C 22.6 (24.9) 0.55 22.6 c (24.9) c
IV AC 1 0.036 0.769 (19.5) A,B 21.4 (23.5) 0.55 21.4 (23.5)
t~
2 0.007 0.839 (21.3) D 26.8 (29.5) 0.49 5
3 0.093 0.844 (21.4) B 21.7 (23.9) 0.61 21.7 ~ (23.9) ~ ,z
4 0.018 0.775 (19.7) A 22.0 (24.2) 0.55 22.0 (24.2)
V AC 1 0.261 0.767 (19.5) B,D 18.4 (20.2) 0.66
2 0.100 0.772 (19.6) A,B 17.5 (19.2) 0.57 17.5 (i9.2) z
VI AC 1 0.015 0.769 (19.5) 22.0 (24.2) 0.57 22.0 (24.2)
2 0.186 0.929 (23.6) B,E 22.2 (24.4) 0.84 -r
m
3 0.086 0.9.64 (24.5) B,E,F 20.1 (22.1) 0.86
4 0.026 0.829 (21.0) F 22.0 (24.2) 0.60 22.0 c "(24.2) ~ t"l
H T = heat treated; A C = as-cast and no stress relief. m
b A----straight crack; B=precracked in less than 50 000 cycles; C = o n e surface trace was less than 90 percent of the average O
z
value of a; D--fracture branched; E = c r a c k length a too large; F = p r e c r a c k surface trace at angle greater than 10 deg from the
line of symmetry.
Considered invalid by ASTM E 399-72.
"O
However, since the seemingly low value of KQ for one specimen cannot be
explained, it was decided that the 0.55 W maximum limit for crack
length a should be maintained as a requirement for a valid Kit. test of
white cast irons.
Branching of the fatigue crack was observed in a few specimens (Irons
II, IV, and V). During testing one branch stopped while the other propa-
gated to complete fracture. The measured fracture toughness values were
always higher in the branched specimens compared to the values obtained
from the specimens that did not branch. It was concluded that valid
fracture toughness measurements cannot be obtained with specimens in
which cracks branched during precracking.
According to ASTM E 399-72, about half of the fracture toughness
determinations in Table 2 can be labeled at Kic. However, it was pointed
out that those which did meet all of the requirements did not differ sig-
nificantly from the "so-called" invalid determinations, with the exception
of the specimens that had a branched precrack. It was concluded that
although all ASTM E 399-72 requirements must be met for ductile
materials, they may be relaxed for brittle white cast irons. The only
requirements that must be met are that the crack length not exceed 0.55 W
and the fatigue crack not be branched prior to testing.
Toughness of Common Abrasion-Resistant White Cast lrons

The range of fracture toughness values (Table 2) obtained for the
irons Used in establishing the testing procedure, combined with the ex-
cellent reproducibility, permitted further investigation of the effect of
microstructure and heat treatment on the resistance of white cast irons to
fracture.
Three commonly used abrasion-resistant white cast irons are 27Cr,
9Cr-6Ni, and 20Cr-2Mo-lCu. The compositions of the irons tested are
given in Table 1. All three irons are used in certain applications in the
as-cast (and stress-relieved) and appropriately heat-treated conditions.
These irons were all tested in the as-cast condition. The 20Cr-2Mo-lCu
iron was also tested in two heat-treated conditions. The heat treatments
and resulting matrix microstructure and hardness are given in Table 3.
Representative microstructures of the irons in the conditions tested are
shown in Figs. 3 and 4.
The fracture toughness values given in Table 3 are each an average of
three determinations. The spread in values obtained for any given iron
was always within 3 percent of the average value. Heat treating the
20Cr-2Mo-lCu iron at 1750~ (950~ and 1850~ (1010~ sub-
stantially increased the hardness and slightly reduced the fracture tough-
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
FIG. 3--As-cast white irons: (a) 27Cr iron, eutectic carbides in a predominantly
austenitic matrix; (b) 9Cr-6Ni iron, eutectic carbides in a matrix o] austenite and
large plates of martens#e; (c) 20Cr-2Mo-lCu iron, eutectic carbides in a matrix of
austenite containing a few patches of fine carbide particles. Etched with 1 percent
picral and 5 percent HCI ( •
11
]2 FRACTURETOUGHNESS AND SLOW-STABLE CRACKING
FIG. 4--Microstructures in 20Cr-2Mo-lCu white cast iron after heat treating at

(a) 1750~ (950~ and (b) 1850~ (lOlO~ Etched in 1 percent picral and
5 percent HCI ( X 5 0 0 ) .
TABLE 3--Fracture toughness of common white cast irons.
Fracture
Matrix ~ Average Toughness
Micro- Hardness, ksiVin.
Iron Heat Treatment structure Rc (MN/m 8/~)
27Cr Tempered 2 h at 410~ (210~ A 49.7 27.7 (30.5)
9Cr-6Ni Tempered 2 h at 410~ (210~ A, M 50.3 23.2 (25.5)
20Cr-2Mo-lCu Tempered 2 h at 410~ (210~ A 50.3 28.5 (31.4)
20Cr-2Mo-lCu 1 h 1750~ (950~ air-cooled, M,A 63.3 27.5 (30.2)
Tempered 2 h at 410~ (210~
20Cr-2Mo-lCu 1 h 1850~ (1010~ air-cooled, M,A 62.9 25.6 (28.2)
Tempered 2 h at 410~ (210~
A=austenite, M=martensite (listed in the order of decreasing volume percent).
DIESBURG ON ABRASION-RESISTANT WHITE CAST IRONS 13
ness from the value the iron had when in the as-cast (and stress-relieved)
condition.
The choice of iron for a given application will depend on the com-
bination of hardness and fracture toughness desired. The iron with the
highest fracture toughness would be the 2 0 C r - 2 M o - l C u or 27Cr iron in
the as-cast condition. However, if a hardness of 63 Rc was desired, the
2 0 C r - 2 M o - l C u iron could be heat treated and very little fracture tough-
ness would be lost.
Conclusions
1. A valid fracture toughness of abrasion-resistant white cast irons
can be measured even if there is a slight relaxation of the requirements
outlined by A S T M E 399-72.
2. An average fracture toughness, as determined from three compact
tension specimens, can be expected to be accurate to within at least
3 percent.
3. The range of fracture toughness values, obtained from the irons
tested, provides more than enough sensitivity to permit the testing tech-
nique to be used in the evaluation of the effect of various microstructures
on toughness.
TABLE 4--Description of Irons 1 through VI.
Predominant
Matrix Average
Micro- Hardness,
Iron Type Condition structure ~ Re
I 15Cr-3Mo 2 h 1850~ (1010~ air-cooled
(2.5% C) Tempered 2 h at 450~ (203~ M, A 62.0
II 15Cr-3Mo
(3.6% C) as-cast A 53.8
III 15Cr-3Mo 1 h 1650~ (900 ~ C),
(3.6% C) furnace-cooled
Tempered 2 h at 400~ (200~ P 42.4
IV 16Cr-2Mo-2Cu-
1Ni (3.3% C) as-cast A 53.2
V 2Cr-2Mo-2Cu-
1Ni (3.3% C) as-cast A, M 53.3
VI 4Ni-2.2Cr
(3.19% C) as-cast A, M 53.2
"M---martensite; A=austenite; P=pearlite (listed in the order of decreasing

volume percent).
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorize
APPENDIX
The white cast irons used to establish plane-strain testing techniques
were not necessarily in a recommended heat-treated condition. Irons II,
IV, V, and VI had not been stress relieved, and Iron III had been furnace
cooled. Table 4 gives the heat treatment and resulting matrix microstruc-
ture and hardness of each iron.
J. C. R a d o n 1 a n d A . A . P o l l o c k 2
Development of Fast Fracture

in a Low Alloy Steel
REFERENCE: Radon, J. C. and Pollock, A. A., "Development of Fast

Fracture in a ' Low Alloy Steel," Fracture Toughness and Slow-Stable
Cracking, ASTM STP 559, American Society for Testing and Materials,
1974, pp. 15-30.
ABSTRACT: Fracture mechanics, acoustic emission, and fractography
were used to study the process of crack growth in 2-in. thick double canti-
lever beam specimens of a low-alloy steel over a wide temperature range.
Kzc values ranged from 33 ksiVin, at --200~ to 240 ksiVTff, at -t-18~
The acoustic emission rate rose steadily with load up to fast fracture.
Changes in the emission amplitude distribution were observed shortly before
fracture at one of the warmer temperatures. Emission during load holds was
measured and emission waveforms were photographed. Fractography
showed a small region of ductile tearing prior to fast fracture at warmer
temperatures. A model of the interaction of ductile tearing and cleavage
fracture is proposed. Ductile tearing is seen as a process taking time and
strongly dependent on temperature. Cleavage is seen as a rapid process
averted by ductile flow and tearing. Emissions are believed to be produced
by plastic zone growth and by cleavage. The experimental facts can be con-
sistently interpreted through this small set of assumptions.
KEY WORDS: fracture properties, acoustic emission, crack propagation,
fractography, brittle fracture, mechanical properties.
T h e objective of this p r o j e c t was to c o m b i n e three techniques: f r a c t u r e

mechanics, acoustic emission, a n d f r a c t o g r a p h y - - i n o r d e r to gain a
detailed picture of the process of c r a c k growth in a l o w - a l l o y steel. F r a c -
ture m e c h a n i c s provides the basic tool for relating the p h e n o m e n o n of
c r a c k growth to the a p p l i e d stresses a n d is used in engineering design.
A c o u s t i c emission p r o v i d e s a m e a n s of detecting the m i c r o s c o p i c defor-
1Research fellow, Mechanical Engineering Dept., Imperial College, London,

England.
~Manager, Technical Services, Dunegan/Endevco (Europe), Royston, Herts.,
England.
15
Copyright*1974 by
Downloaded/printed byASTM International www.astm.org
TABLE 1 Material characteristics.
Composition
Composition, %
Steel C Si S P Mn Ni Cr
DUCOL 0.155 0.24 0.021 0.024 1.35 0.22 0.67
Mechanical Properties
Yield Ultimate
Strength, Tensile Strength, Charpy
Steel tonf/in ~ tonf/in 2 50% FATT Remarks
DUCOL 33.6 41.8 20 to 35~ normalized and
tempered
mation events that precede fracture and is used in engineering inspection.

Fractography provides insight into the microstructural and metallurgical
factors that steer the course of crack growth and is used in failure investi-
gations.
The project shows that this combination of techniques is productive.
In considering the results of experimental work in the three areas, we
were able to develop a model of interacting deformation mechanisms,
providing a simple basis for understanding the development of fast frac-
ture in this material.
Material, Specimen Geometry, and Fracture Mechanics Results

The material used for these tests was a low alloy steel whose chemical
composition and mechanical properties are shown in Table 1.
The specimens used for these tests were of the double cantilever beam
(DCB) parallel edge type, which had previously been found particularly
suitable for the study of fast fracture in materials of this type. The de-
velopment of this geometry is described in detail in Ref. 1. 3 Due to the
fact that the DCB specimen could b e designed in any convenient size,
there was no need for an excessively large testing machine. A specimen
length of 30 in. and a width of 2 in. was used, as shown in Fig. 1. The
other dimensions were carefully balanced to avoid yielding of the aims
and to give satisfactory performance over a wide temperature range
(-200~ to + 1 8 ~
After sharpening the starter saw-cut by fatiguing in a three-point bending
rig at one-fourth nominal yield and 6 0 / 1 0 0 kc, the specimen was mounted
in a standard liquid nitrogen/petroleum ether cooling bath and acoustic
a The italic numbers in brackets refer to the list of references appended to this
paper.
RADON AND POLLOCK ON FAST FRACTURE 17
SAW 0.)1' STAR'rER P L O P A G ~ " r l N ~ ~ RACI~
~ m O , ~ 6 t tr
/ "
I I -~.
I
' 4 4 ~ ~ ~o'ro, J (.oo~' ~oo'r ~A~
I . . . . . .
FIG. 1--DCB specimen (straight edge type).
emission transducers were attached. The specimens were tested in a Tinius-

Olsen 120 000 Ib capacity testing machine at a crosshead speed of 0.02
in./min. On reaching the critical load the crack propagated rapidly along
the median plane until a new equilibrium condition had been attained.
The crosshead speed was increased to 0.05 in./min and further crack
jumps followed, typically 2 to 4 per specimen according to temperature.
At the lowest temperatures ( - 135~ and below) a type of slow tear was
observed, with the crack advancing in a large number of small unequal
steps. A typical record of load versus extension is shown in Fig. 2. The
exact length of the crack jumps was measured directly from the fracture
surface after the test.
In calculating Kic values, the critical loads for crack initiation and arrest
were read directly from the load record and values for crack opening
were corrected for the amount of machine extension. The slope of the
elastic loading lines, which may be extrapolated to the origin, was used in
the calculation of specimen compliance ~. Alternatively, d~/da was
measured on another test piece by introducing sharpened slots of increas-
ing length. The derivative of the compliance, which therefore includes
the effect of the side grooves, can be used in the calculation of strain
energy release rate from which K~c values were derived. Calculated values
F ~- d~
G~--
2W da
for initiation and arrest are shown as a function of temperature in Fig. 3.
The curves show some similarity with the behavior of medium strength
steels [2], but the arrest curve is much closer to the initiation curve, indi-
cating that the crack does not propagate so freely as might be expected.
18 FRACTURE TOUGHNESS A N D SLOW-STABLE CRACKING
80
70
"rk4ic~ O u c o L .
Oc B Nu~s~ 21 (wrrw A r -'r,=~,-r)
&o
A ' r - 7~osC ANO CROSSk4S.AC, S p . S O t O2. O~S ~ . ~ . . ~c.~c
C~< I
5c kCIC = I S S , o o o Ib/iN3/t
I~a, . 5 ~ o o o I bliwl sii
40
5o
2o
J
%0
I -I
0-1 o.t o.~ o,4 0.5 O.G 0.7 0.8 O.~ i.O
~XT~NSION ~iN)
FIG. 2--Typical load-extension curve.
i i I
260I i
240 FractureToughnessKIc A
(ksi/in)
200
I n i t i a t i o n KIc Tests )#
valid by ASTME39gcriteria
Initiation
160 Initiation) 2 in. plate &

1:1 Arrest )
A
Arrest~.Wi
/
'I~ II -- D
n
I
q-200 -~50 -I00 o
Temperature~
FIG. 3--Fracture toughness versus temperature: Ducol steel DCB specimens (2-in.
thick), crack propagation at 90 deg to rolling direction and plate surface.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authoriz
Acoustic Emission
The equipment used to monitor acoustic emission in these tests was a
conventional Dunegan system, counting threshold crossings, and a Cam-
bridge Consultants amplitude sorter which allows emissions to be batched
according to peak amplitude, the ranges being 20 dB apart. Dunegan
140 Series transducers were used with both instruments.
As a simple cheek on some possible sources of spurious emission (for
example, pin rotation), several of the tests were interrupted midway and
the load was reduced to a low level then reapplied. A good Kaiser effect
was observed in all these cases, that is, emission was negligible during
unloading and during reloading until the previous maximum load was
reached. However, there was some other evidence that sources of irre-
versible spurious emission (possibly fixture noise) were acting at low
stress levels, so the data acquired below 30 kips may be somewhat
unreliable.
Amplitude sorter results for five tests are shown in Fig. 4. The parameter
plotted is the number of events per second at the first level of the ampli-
tude sorter (50 to 500 ~V at the transducer). This emission rate data is
averaged over appropriate intervals up to the first crack jump, and plotted
as a function of load. Since the specimens are behaving in an essentially
elastic manner, the plots are virtually the same as would be obtained by
plotting emission rate against time. Emission rate increases with rising
Emission rate
(level I events/see)
-~ .135~ ~ -55~ ~ -32~ g8 -19~ a -8~
FIG. 4--Emission ]rom first set of five specimens.
load. These experiments did not show any systematic dependence of low-
load emission on temperature. The cumulative emission to failure was
much higher for the warmer-temperature tests, because they proceeded to
higher loads with a rising emission rate.
Analysis of the amplitude distribution yielded some interesting results.
The tests at warmer temperatures tended to yield a larger proportion of
high-amplitude emissions. A simple empirical parameter for describing
the amplitude distribution is
nz=number of emissions with amplitude exceeding 500 ~V

nl = number of emissions with amplitude exceeding 50 ~V
This parameter, averaged over the load rise to the first crack jump, is
plotted as a function of test temperature in Fig. 5. One test gave an
anomalous result which is believed to be due to transducer mounting
problems encountered in that test. Apart from this anomaly, there is good
0.4
Amplitude Distribution
Parameter n21n]
0.3 §
0.2
0.1
§
O.O * i | | t i
-140 -120 -I00 -80 -60 -40 -20
Temperature (~
FIG. 5--Variation of amplitude distribution with temperature.
TABLE 2--Variation of amplitude distribution parameter during test at --IO~
Emission Count Amplitude

distribution
Load Level Level Level parameter
(kip) 1 2 3, 4 n2/nl
0-10 55 11 0 0.167
10-17.5 71 20 1 0.23
17.5-30 115 39 5 0.28
35--45 160 61 13 0.32
45-50 115 36 6 0.27
50-55 150 52 8 0.29
55-60 176 54 3 0.24
65-70 108 61 7 0.39
70-73.6 116 63 8 0.38
Total 1066 397 51 0.296
evidence that deformation events prior to fast fracture tend to be larger at

higher temperatures.
The drop in the 50 to 500 ~V emission rate prior to failure at - 8 ~
(Fig. 4.) was confirmed in a second test at - 1 0 ~ The second test
showed, however, that this effect was outweighed by an increase in the
rate of high amplitude emissions. The amplitude distribution changed
markedly at 85 to 90 percent of the failure load. Data from this test is
shown in Table 2. The change in the parameter n2/nl above 65 kips is
statistically significant at the 0.1 percent fiducial level. The approach to
failure was difficult to recognize by other means and it appears that in
this material at least, amplitude distribution analysis[3,4] offers the
most promising diagnostic of incipient failure. At the lower temperatures
this effect was not observed; the amplitude distribution parameter was
constant, apart from statistical variations, up to the point of instability.
Figure 6 shows a graph of cumulative emission count (smoothed)
against load for the test at - 1 0 ~ This data was taken from the Dune-
gan system at a gain of 86 dB. The concave form of the graph is typical
for flawed specimens. There is no feature in this graph that would serve
as a precursor to failure, but it is possible that by operating at a lower
gain, a recognizable feature would be obtained as a result of the pre-
viously described change in amplitude distribution.
Figure 6 also illustrates the effect of "hold" periods in the test at
-10~ The machine crosshead was stopped for several minutes at 30
and 60 kips. Large quantities of emission were produced in both these
periods. On resumption of loading, the emission rate was low, as if to
compensate for the "extra" emission produced during hold. Figure 6
I I i i l i i
1000 Cumulative emission counts
(Dunegan system, X1000, 86 db)
800 //
600
40(
20(
J
Load (kips)
I I I I I #
10 20 30 40 50 60 70
FIG. 6--Emission from Specimen D21 (--IO~
shows that the effect of an extended "hold" period is not entirely relieved
until the load has risen nearly 10 kips above the "hold" load.
The implication is that we were conducting the tests at quite a high
strain rate for this material, in that the material was lagging well behind
the equilibrium state even at relatively low loads.
Emission during hold is of interest because in some situations it may
serve as a warning of incipient failure[5,6]. Parameters which may be
used to characterize "relaxation emission" are (1) counts and (2) time
distribution.
In the test on Specimen D21, the 60 kips "hold" yielded nearly twice as
much emission as the 30 kips "hold." The time distribution of the relaxa-
tion emission is shown in Figure 7. The statistical quality of the data is
not very good, but the emission rate appears to fall off with a time con-
stant of the order of 1 min. There are no obvious differences in this
respect between the two "hold" periods.
With only two "hold" periods to work with, it would be premature to
say whether relaxation emission could be used for incipient failure diag-
nosis in this material. The theory of relaxation has still to be developed;
these tests provide interesting grounds for thought but no obvious con-
clusions can be drawn.
A number of emission waveforms were photographed as the tests pro-
University of Washington (University of Washington) pursuant to License Agreement. No further reproducti
I ~ r f t
EmlSSlon counts - ~
(Dunegan, XIO00, 86 db) / ~ /
IO0 +.I+
_._~//~-'I"~
6o //
Time (m~nutes)
0 I | I i I
0 1 2 3 4 S
FIG. 7--Emission during hold at 30 kips (lower curve) and 60 kips (upper curve)
during the test of Specimen D21 (--10~
ceeded. A two-transducer technique allowed the rising edges of emission

waveforms to be photographed in real time. Commonly, the natural
emission waveforms had a slower risetime than artificial waveforms pro-
duced by pulse injection from another transducer. This rather surprising
result suggests that many of the deformation events had a fine structure
extending over 100 to 300 ~s. A case was also recorded where the initial
event was followed by other short-rise events after 1.3 and 2.3 ms.
Figure 8a shows a typical emission waveform. Figure 8b shows a
more unusual waveform, which was produced by the second crack jump in
a test at - 1 0 8 ~ Instrumentation systems always saturate for the largest
emissions, so to capture this event we connected an accelerometer (nominal
resonance 50 kHz, sensitivity 60 m V / g 3 to 10 kHz) directly to an oscillo-
scope at 10 V / c m without intermediate amplification. The resulting wave-
form is a remarkably perfect exponential decay. Peak-to-peak voltage is
plotted against time in Fig. 9. Extrapolating back to the time of origin, the
original signal level may be estimated at 140 V peak-to-peak, an unprece-
dentedly high figure.
Emission waveform photography of rapid fracture events might welt be
used as a method of determining crack velocity. We would expect the
transducer response to follow a perfect exponential decay only after the
crack has arrested; by examining the earlier part of the waveform it
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions autho
FIG. 8--(a) Typical emission---0.5 ms/cm; (b) last [racture--lO V/cm, 0.5 ms/cm.
Amplitude
(volts)
I00~
50
Time{milliseco~
0.5 1 0 1.5
FIG. 9--Peak-to-peak amplitude of decaying waveform as a function of time.
should be possible to determine the length of time over which the crack is
in motion.
Fractography
A number of stereoscan micrographs of selected regions of the fracture
surface were prepared. Interest concentrated on slow growth bands which
were visible at the points of initiation and arrest. These bands are wider
and more distinct at higher temperatures (width of the order of 0.015 in.
at - 1 0 ~ and are almost imperceptible at temperatures below
-100~ Figure 10 shows micrographs of the slow growth bands and of
the fast fracture region at two temperatures, - 1 1 5 and - 1 0 ~ The slow
growth bands are in clear contrast: ductile tearing is evident at - 1 0 ~
but at - 1 1 5 ~ quasi-cleavage is seen with only traces of ductile tearing.
The fast fracture regions are essentially similar: failure is by transgranular
cleavage. However, there are local traces of ductile tearing even in the
fast fracture region at - 1 0 ~ Above - 1 0 0 ~ the amount of ductile
tearing prior to failure increased strongly with temperature.
FIG. lO---(a) Slow growth band at --115~ (>(1000), (b) slow growth band at
--10~ (• (c) [ast #acture at --115~ ( X 5 0 0 ) , and (d) last fracture at
- l o o c (•
26
Discussion
The main findings of these tests can be summarized as follows:
(a) K~e and K~rr increase with rising temperature, as is common with
steels.
(b) The acoustic emission rate rises with K and the cumulative emission
prior to fracture increases rapidly with rising temperature.
(c) In a test at - - 1 0 ~ the emission amplitude distribution changed
shortly before fast fracture. In low-temperature tests this did not happen.
(d) Fast fracture was by transgranular cleavage, with small amounts of
ductile tearing at warmer temperatures.
(e) A slow growth zone of ductile tearing was well formed at warmer
temperatures. At lower temperatures, this ductile tearing at the edge of
the fatigue precrack was progressively replaced by quasi-cleavage and the
slow growth zone became imperceptible.
Before turning our attention to events at the crack tip itself, we should
consider the plastic zone which plays such an important part in fracture
mechanics. In these tests it was impractical to examine the plastic zone
directly, either before or after the tests. We can, however, make a rough
estimate of its size from the formula:
K2
d _ m
37ray2
where
, = 33.6 t o n / i n ? , ignoring the variation of yield stress with temperature,
K = 34 ksi~/in., corresponding to fast fracture initiation at - 135~
we find d = 0 . 0 2 2 in., and
K = 228 ksi~/in., corresponding to fast fracture initiation at - 7 ~
we find d = 0 . 9 8 in.
This latter figure is probably so large as to invalidate the simple formula
just used. However, there can be no doubt that yielding extended well into
the specimen and well ahead of the region of slow crack growth, before
rapid fracture took place.
Plastic zone considerations are important because in many materials of
this type, acoustic emission is associated with yielding and the growing
edge of the plastic zone is considered to be the most likely source of
emissions. This is the basis of a well-known model which gives a power-law
relationship between cumulative emission count and stress intensity.
Ductile tearing is a relatively quiet process, in low-strength steels at least
[6]. Cleavage, on the other hand, is a credible source of relatively high-
amplitude emissions.
The wide variations in supposed plastic zone size are broadly consistent
with the wide variations in cumulative emission count prior to fast frac-
ture. Comparing results of tests run at different temperatures, the cumu-
lative emission count to failure varies approximately as the square of K~,,.
Also, the curve shown in Figure 6 approximates closely to a parabola.
Turning to the crack tip itself, we have an interplay between two con-
flicting mechanisms: cleavage and ductile fracture. A simple model is
proposed in the following as a description of the interaction on micro-
scopic and submicroscopic levels.
(a) Ductile rupture is a relatively slow process, taking time and involv-
ing substantial flow of the material on the fracture surface. This process is
strongly dependent on temperature, probably because some of the micro-
scopic flow mechanisms are thermally activated.
(b) Cleavage is a relatively rapid process. For activation it requires a
stress concentration that is very high on the microscopic level; this stress
concentration will be relieved by local ductile flow. Cleavage produces
acoustic emissions of relatively large amplitude. There is relatively little
flow of material on the fracture surface.
The different microscopic characteristics of the two mechanisms, as just
described, give an understanding of the macroscopic effects that were
measured in the course of these experiments. As the temperature rises,
cleavage is increasingly inhibited by plastic flow which relieves stresses in
the microscopic regions of highest stress concentration. As the temperature
rises, the macroscopic stress intensity can rise to much higher levels
before cleavage finally supervenes. Cleavage events at the crack tip have a
tendency to lead on to fast fracture, since relatively little energy is
required to sustain crack propagation in the cleavage mode.
At a temperature of - 1 0 ~ some local cleavage is taking place before
fast fracture; the material is ductile enough to arrest these incipient fail-
ures locally, until the stress intensity becomes too high. This local cleavage
leads to a change in emission amplitude distribution to 85 to 90 per-
cent of the failure load.
At low temperatures, local cleavage tends to lead straight on to brittle
fracture since there is little chance that an incipient failure will be arrested
by local plasticity. Thus, the amplitude distribution does not change before
failure, but remains characteristic of the activity at the edge of the grow-
ing plastic zone that is the dominant source of emissions.
Fast fracture is predominantly by cleavage even when the preceding
slow growth zone shows ductile tearing. On the basis of our model, this
happens because ductile tearing takes time, and time is no longer available
once cleavage initiates at a high enough stress intensity. In the first stage
of fast crack growth, stored elastic energy is made available faster than
it is absorbed[7]. Ample energy will be available to sustain cleavage, the

crack will accelerate and ductile flow will become less and less likely.
Nevertheless, some traces of ductility were observed even in the fast
crack region at warm temperatures. This residual ductility is of some
importance. An area of slow ductile tearing may absorb nearly 20 times
more energy than an equal area of fast fracture[7]. Ductile tearing over
small regions of the fast fracture surface may make a disproportionately
large contribution to energy absorption. This could be one of the causes
of the large variation in dynamic surface energy with temperature (22,
52, and 117 • 105 ergs/cm 2 at - 197, - 135, and - 60~ respectively,
in UXW steel[2]). Our model also predicts that the amount of ductile
tearing would be greater at the beginning and end of the crack jumps,
where the crack was running more slowly; we do not have enough infor-
mation at present to say whether this effect would be readily observable.
The model is consistent with two other well-known effects in the frac-
ture behavior of steel. The lowering of critical stress intensity with increas-
ing specimen thickness is generally recognized to be a result of material
constraints; increasing the thickness reduces plastic flow and thereby in-
creases the likelihood of cleavage. Second, increases in strain rate lead to
reductions in K~c[2]; on our model, this would happen because plastic
flow takes time. At high strain rates, flow is delayed and the likelihood of
cleavage is higher. The persistence of emission for several minutes during
hold periods is an indication that this is a real effect at the strain rates
used for these tests.
Conclusions
In summary, a large number of experimental factors can be brought
into line on the basis of a very limited set of suppositions, namely
(a) Ductile rupture takes time and the process is strongly dependent
on temperature.
(b) Cleavage is rapid on the microscopic scale, and is averted by
ductile flow and rupture at the points of highest local stress.
Hopefully, in this model we have isolated some of the most salient
factors involved in the ductile-brittle transition. The model lends itself
naturally to making predictions and affords an outline picture whose
details are yet to be filled in.
Acknowledgments
Fractography for this project was conducted under the direction of
C. H. Jones. All DCB tests were performed by F. A. Johnson. The work
was carried out under contract to the Ministry of Defence (Navy).
University of Washington (University of Washington) pursuant to License Agreement. No further repro
References
[1] Turner, C. E. and Radon, J. C., "Fracture Toughness Measurements in Low
Strength Structural Steels," Paper 14, Proceedings, 2rid International Conference
on Fracture, Brighton, April 1969.
[2] Pollock, A. A. and Radon, J. C., "Acoustic Emission in the Fracture Toughness
Test of a Mild Steel," Document IIW-X-595-70, Lausanne, Switz., 1970.
[3] Nakamura, Y., Veach, C. L., and McCauley, B. O. in Acoustic Emission, ASTM
STP 505, American Society for Testing and Materials, 1972, pp. 164-186.
[4] Pollock, A. A., Nondestructive Testing, Vol. 6, No. 5, Oct. 1973, pp. 264-269.
[5] Harris, D. O., Dunegan, I4. L., and Tetelman, A. S., "Prediction of Fatigue
Lifetime by Combined Fracture Mechanics and Acoustic Emission Techniques,"
Technical Bulletin DRC-105, Dunegan/Endevco Corp., 1970.
[6] Pollock, A. A. and Smith, B., Nondestructive Testing, Vol. 5, No. 6, Dec. 1972.
[7] Radon, J. C. and Pollock, A. A., Engineering Fracture Mechanics, Vol. 4, 1972,
pp. 295-310.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions author
H. H. Chaskelis, 1 W . H. Cullen, 1 a n d J. M . K r a f f t 1
Acoustic Emission from 4340 Steel

During Stress Corrosion Cracking
REFERENCE: Chaskelis, H. H., Cullen, W. H., and Krafft, J. M. "Acoustic

Emission from 4340 Steel During Stress Corrosion Cracking," Fracture
Toughness and Slow-Stable Cracking, ASTM STP 559, American Society
for Testing and Materials, 1974, pp. 31-44.
ABSTRACT: A commercially available system is employed to detect and
count acoustic emissions emanating from the aqueous stress corrosion crack
propagation in 4340 steel. Standard ASTM E 399 compact tension specimens
with short notches, a/W=0.25 versus 0.50 standard, are prepared in four
tempering temperatures: 204, 316, 427, and 538~ (400, 600, 800, and
1000~ Crack length is monitored with a notch opening clip gage. Com-
parisons show the time rate of emission events to increase with stress inten-
sity in rough correspondence to, but much more rapidly than, the crack
velocity. Tempering back tends to suppress the count rate as a warning of
fast fracture instability. Some similarity is observed between the areal rate
of acoustic emission and factors which influence degree and intensity of
plastic flow instability in the fracture process zone.
KEY WORDS: acoustic properties, crack propagation, fracture properties,
fatigue (materials), acoustic detection
That fracturing is a source of sound should have been common knowl-

edge in any age. What is new, in view of Kaiser[I] 2 and others[2], is
that even extremely slow, ostensibly stable fracturing is not silent. Hart-
bower[3] has o b s e r v e d that the progress of a stable subcritical c r a c k is
signaled b y a s t e a d y succession of discrete emission events, o b s e r v a b l e by
"listening" t h r o u g h sensitive piezoelectric t r a n s d u c e r s a n d amplifiers.
P r i o r to his observation, there h a d been o u r own[4] that the event of
1 Physical science technician, metallurgist, and head, Ocean Materials Criteria

Branch, respectively, Ocean Technology Division, Naval Research Laboratory, Wash-
ington, D. C. 20375.
The italic numbers in brackets refer to the list of references appended to this
paper.
31
Copyright*1974byby ASTM lntcrnational
Downloaded/printed www.astm.org
criticality Kie eould often be associated with an audible "pop in." Brown
and Srawley's application of a phonograph type transducer in such tests
[5] did indeed show that the sound was emitted not as a single but as a
succession of events as the point of criticality K~e was reached and sur-
passed in the path toward general fracture instability K~..
An annunciator for K~, while useful in testing procedures, is rather
too late in the life of a structure to serve as a warning of destruction.
On the other hand, a warning of the early stages of subcritical cracking,
as proposed by Hartbower, could be of great value in proof testing, even
perhaps for in-service surveillance of fracture-sensitive structures[6]. It
would be helpful in such endeavors to understand the sources of the
sound in subcritical propagation; ideally a correct predictive model of the
sound producing event. But falling short of this, even a discriminating
characterization of the emissions to be expected in various materials and
corroding environments should be helpful. This paper reports an attempt
to supply such information for 4340 steel, quenched, then tempered back
to levels of hardness typical of its utilization in structural and machine
components.
Experimental Procedures
Briefly, the cracks, serving as sources of acoustic emission, are grown in
short notch compact-tension specimens (ASTM Test for Plane-Strain
Fracture Toughness of Metallic Materials (E 3 9 9 - 7 2 ) ) , immersed in
distilled water, and subjected to constant load. The opening of the notch
was used as a measure of compliance, thus providing a measure of crack
growth. The accumulation of emissions is recorded, along with the load,
versus the notch opening. Tensile stress-strain curves, needed to assign a
degree of fracturing instability within the crack tip region, were measured
on small tensile specimens. The crack growth data, compared here to the
stress wave emission, has been reported elsewhere[7], and some of the
procedural description is repeated here.
The 4340 specimens were cut from a 13-mm-thick plate, Interlake Co.
heat ~ B 0 - 5 3 5 - 5 0 0 . Compact tension specimens were cut, in full plate
thickness, so as to be subject to crack propagation in the long transverse
direction of the plate, TL. Of the 2 in. of width W between the back of
the specimen and the load line of the holes, about 1 in. is useful for
crack propagation a measurements, 0.25<_a/W~_0.75. Four batches of
about 20 each were austenitized at 843~ (1550~ oil quenched, then
tempered for 1 h at temperatures of 204~ (400~ 316~ (600~
427~ (800~ and 538~ (1000~ followed by air cooling to room
temperature. Small 4.3 by 12-mm tension-compression (T-C) specimens
were cut from the same material, axes normal to the crack plane (T-direc-
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions a
CHASKELIS ET AL ON ACOUSTIC EMISSION FROM 4340 STEEL 33
tion) and heat-treated along with the fracture specimens. All fracture
specimens were prepared with low-stress fatigue cracks.
The compact tension specimens were tensioned in an N R L dynamic
loader[8] (Fig. 1) set up to maintain constant load through a selective
gas pressurization. Some tests were run in stages in which the head was
locked, allowing a "load shedding" which held KI to a rate of increase
about one third that for the constant load condition. The crack growth was
detected with a displacement clip gage inserted in the notch, with out-
put, time modulated, plotted versus the load on a Hewlett Packard Model 2
F R A X-Y1-Yz recorder.
A standard PZT transducer was positioned atop one arm of the speci-
ment (Fig. 1) using an interface of viscous petroleum grease and a dead
weight of some 89 to 1 kg. The counting instrument, Dunegan 301 total-
izer, was set to provide 94 dB gain, 40 dB from the 801 P preamplifier and
54 dB from 301 totalizer to the received electrical signal, then cutting off
all signals below 1.0 V. Of the several filters available in this instrument,
that passing the Band 0.3 to 1.0 MHz was employed. The natural fre-
quency of the PZT transducer was about 0.16 MHz, with a second peak in
the 0.3 MHz range, 10 dB down from the primary peak. This secondary
peak falls at the lower end of the band pass window which was used, a
FIG. 1--Specimen and grips with load cells, A E transducer and notch opening
clip gage. Immersion of the back half of the specimen in water reduces by one fourth
the A E count rate.
selection which removed some of the extraneous noise. However, use of a

band more nearly centered on the transducer center frequency, designated
0.1 to 0.3 MHz, at the same gain setting, increases the A E count by a
factor of about four. It should also be noted that the immersion of the lower
half of the specimen in the water decreased the emission rate by a factor
of about four. This is attributable to the greater loss of the vibrations by
leaking the sound pulse out of the specimen across the steel to water inter-
face.
In a typical record (Fig. 2), time marks are superposed by a dotting of
the recorder pen at selected intervals, using a cam driven microswitch.
In addition to 15 and 1 min marks by this method, 5 and 1 s amplitude
modulations were available from a Tektronix time mark generator (Mod
180A), capacitively coupled to the load channel. Full scale of the record
ordinate (Fig. 2) corresponds to 105 counts.
FIG. 2--Tracing of typical record and data plots, 4340, 316~ (600~ temper,
upper graph with 1 min time modulations, lower graph with derived crack length a
and stress intensity factor K.
CHASKELIS ET AI. ON ACOUSTIC EMISSION FROM 4340 STEEL 35
Data Reduction
The data processing involves converting the load versus crack-opening-
displacement record to crack length, a, and then to stress intensity factor,
K. At selected time stations on the record (Fig. 2), the load P and dis-
placement v, relative to the origin, values are read and recorded. The
compliance at each station, v / P is then referred to a compliance to crack
length calibration. The calibration was derived empirically by measuring
compliance for a series of fatigue propagation crack lengths, marked with
bands of reduced AK propagation[8].
In a graphical method, the K / P ratio is read from a calibration curve,
after Wilson[9], plotted versus crack length, a, on the same paper as v / P .
A parametric curve combining both calibrations (Fig. 3) may be used to
monitor crack growth and stress intensity level, or even to read them
from the record. Values of a and K are then plotted against time, t, or
alternatively, all three against notch opening displacement v directly on a
copy of the record (Fig. 2). Vertical lines corresponding to a convenient
set of K values and which intercept these various curves are then drawn.
At such intercepts, slopes are measured corresponding to dv/da, d N / d v ,
and d v / d t where N is the cumulative acoustic emission. These are suitably
combined to yield tabulations of d N / d t , da/dt, and d N / d a versus K. If
regions of extraneous noise appear to have exaggerated the stress wave
count, these can be eliminated by a judicious setting of slopes.
FIG. 3--Parametric a and K calibration curves for standard CTS specimen of

steel 0.508-in. thick.
As an alternative method of data reduction, a computer program was

set up to obviate the graphical interpolations, substituting, instead, analytical
forms for the calibration curve and stress intensity factors. Two curve
fitting routines are used to generate functional forms of crack length and
cumulative stress wave emission versus time. For each increment of time,
computed values of a are used in Wilson's polynomic formulas for K, and
both fitted curves are differentiated to provide values of dN/dt and da/dt.
CALCOMP plots of cumulative emission and crack length versus time,
plus da/dt versus K are drawn.
Comparison of Acoustic Emission to Crack Growth

Drawing back the hardness of this steel has pronounced effects on crack
propagation: the velocity of stress corrosion cracking da/dt is markedly
decreased; the velocity becomes more insensitive to the stress intensity
factor in the midrange of K values; and the K thresholds of crack propa-
gation increase. Quantifying these behaviors, Fig. 4 shows crack velocity as
a function of K. It is fitted with curves derived from the plastic flow prop-
erties, using the tensile ligament instability model[7]. Some aspects of
this are discussed in the next section. The time rate of stress wave
emission dN/dt (Fig. 4) appears to be much more sensitive to the stress
intensity level. An important practical note is that the time rate of acoustic
emission signaling the approach of fast fracture instability is markedly
reduced by drawing back. A change of some three orders of magnitude is
produced by tempering at 538~ (1000~ versus 204~ (400~
Another way of displaying this is with a logarithmic plot (Fig. 5) of crack
velocity da/dt versus emission rate dN/dt. The slope indicates an emis-
sion rate varying as about the cube of the crack velocity but with the pro-
portionality factors decreasing markedly with tempering.
Comparison of Acoustic Emission to the Fracturing Process

If a bar-shaped tensile element is stressed in tension, a stress wave
compression will accompany its unloading if this occurs in a time period
short compared to the time for an elastic wave to traverse its length. To
do this, a crack velocity comparable to the bar wave velocity, about
5 • 10 ~ cm/s would be required. The average velocity of these stress cor-
rosion cracks, 5 • 10 -4 cm/s, is some nine orders of magnitude slower.
Thus, it seems unlikely that the occurrence of the emission event could be
uniquely related to the average velocity of the crack. It would seem more
reasonable, after Hartbower[3], to normalize the emission count, at given
K level, to the area opened up.
It is of interest to estimate the largest possible emission rate per unit
FIG. 4--Comparison of crack velocity d a / d T with rate of acoustic emission

d N / d T shows a rough ordering with respect to effects of tempering temperature.
area. The tensile ligament instability TLI model provides a process zone
dr which could be considered as a unit area for breakdown. The value of
dr may be estimated from the observed proportionality between K~e and
the strain for (triaxial) tensile instability c~, K~e= V'2rdr E~c where E is
Young's modulus for the material. This same proportionality factor is
used for relating the lower points on the stress-strain relationship to values
FIG. 5--Cross plot shows A E rate quite sensitive to crack velocity. The A E rate
at a velocity verging on fast ]racture instability is greatly reduced by tempering.
of K below (or subcritical of) K~c. The value of d~, for this particular
steel, 2.7 ~m (105 ~in.) has been determined from flow to fracture cor-
relations on a large set of corrosion fatigue as well as these stress corrosion
cracking results[7]. The areal rate of acoustic emission is normalized by
multiplying dN/da by dr2/B, where B is specimen thickness. The result,
dN/daT, where aT=dT 2, should reach a value of unity if each area large
enough to contain a dr-ligament were to emit a signal, and each signal were
to be counted only once, that is, no ringing of the transducer. The plot of
dN/dar versus K (Fig. 6 left) shows values approaching unity as Kie is
approached, particularly for the lower tempering temperatures. This is far
FIG. 6 - - A c o u s t i c emission count per dT process zone area correspond roughly to

degree of process zone instability as derived #orn stress-strain curves of Fig. 8 using
T L I model.
short of the maximum areal count rate which could be expected. However,
the total count is depressed by the offset of the filter band and the immer-
sion of the specimen, as noted earlier.
dN/dar drops off rapidly at lower levels of crack loading K. We would
tend to associate this with the fractographic observation of a greatly
reduced population of ductile tear dimples for propagation at low K levels.
This is illustrated (Fig. 7) with SEM fractographs of the 316~ (600~
tempered specimen of the test record shown in Fig. 2. It is possible that a
dense population of dr-ligaments do form at the throat of the growing
subcritical crack, as implied by the TLI model, but that only a small frac-
tion of these survive to rupture abruptly, as by sudden void coalescence,
to emit a sound pulse.
FIG. 7 - - T h e rapid increase in dN/daT with K (Fig. 6) is associated with increasing

area of ductile tear dimple residue on the fracture sur[ace, typified with the 4340/
316~ temper specimen of Fig. 2.
In looking for factors which might reasonably be taken to influence the

number and intensity of emissions per unit area, it is possible that the
condition of the dr-ligament would be important. The degree of mechanical
plastic flow instability varies as the flow stress and inversely as the tangent
modulus 0z. For a triaxial case this may be taken as
(0_K~+ 0 5 y 1
o. r " ]
If the stabilizing influence of the crack tip strain gradient, increasing with K,
is considered, the entire growth rate factor (GRF) is
eF+e~ +0.5
where e~, is the strain of the plastic flow tensile stress-strain curve, and the
triaxially corrected yield point strain
where v is the Poisson ratio and era is the ordinary yield point strain.
The form of the growth rate factor is quite sensitive to the form of the
stress strain curve. Illustrating this, Fig. 8 shows both curves for this set
of 4340 tempering temperatures. For simplicity here, the e~ term has not
been included. When the alloy is drawn back far enough to restore the
lower yield point effect, 1000~ the GRF becomes large at strains cor-
responding to the yield point as well as at those for maximum load point.
This double peak is normally obscured by other effects which raise the
stress corrosion cracking threshold well above its lower limit associated
with the elastic limit in the TLI model. In aggressive environments, pro-
ducing high cracking speeds, this effect is often observed. For a discussion
of this, as well as a comprehensive treatment of the TLI model, inappro-
priate here, study of Refs 7 and 8 is suggested.
A second factor which might be included is the elastic energy which can
be released in the instant severing of a tensioned ligament. This should be
proportional to ~r2/E. Combining these two as multiplicative effects, Fig. 6
right shows the GRFxea, Z / E ( - G R E F ) matched to the dN/daw data.
Both are positioned relative to the corrected strain cF+ eR( = or), consistent
with the predetermined value of dr, where as noted earlier K = ~/2rrdTEer.
As a whole, the correspondence of emission rate to the GREF is close
enough to be suggestive of an underlying cause worth considering in frac-
ture data collections.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions auth
42 FRACTURETOUGHNESSAND SLOW-STABLECRACKING
I0 2
y AISI 4340
TEMPER
T
0+ I0 204"G~
316"
538 ~
2u CTT
t
E F ~ E~IO%
i0 -t I I I I L I llll
10-3 10-2 iO-I
s
FIG. 8 - - T e n s i l e stress strain c u r v e s f o r 4340 steel become more angular with

tempering leading to a corresponding [orm o[ the V(K) relationship indicated [rom
the basic growth-rate-[actor GRF agrument.
Discussion
The consistency with which stress wave emissions accumulate in response

to crack extension is truly remarkable. Yet it must be recognized that it is
not a direct measure of crack propagation. Plastic deformation alone can
induce acoustic disturbances. Except in the most isolated systems, spurious
mechanical and electrical noises may add to the stress wave count. From
crack extension, the signals are short and pulse-like, shocking the detector
into a damped oscillatory decay [6]. Even before reaching the detector,
the waves are subject to transformations such that it becomes difficult to
assign cause and effect relationships between fracture dynamics and acoustic
emissions.
In the system employed, the count is a measure of each oscillatory peak
in the decay envelope above a preset threshold voltage. Thus, the count will
generally exceed the total number of events, reflecting too the amplitude or,
energy level of each exciting pulse. On the other hand, too high an accept-
ance threshold will tend to blind the system to the weaker pulses. The
acuity of the sonic pulse, and thus its tendency to shock the detecting trans-
ducer, will diminish by dispersion and attenuation during its propagation
through the materials. The natural frequency and damping of the trans-
ducer, as well as the instrument filtering characteristics, will have an
ameliorating effect.
Despite all of these qualifications, the detection system is commercially
reproducible and has demonstrated a capability of providing an indication
of the impending fast-fracture instability. Results of this paper suggest that
the character of this indication is affected by the metallurgical condition
of the alloy, which therefore should be considered in utilizations of the
method.
Conclusions
The following are based upon, and thus possibly limited to, various
tempers of the low alloy quenched and tempered 4340 steel subjected to
aqueous stress corrosion cracking and monitored with a certain commer-
cially available system.
1. The rate of acoustic emission seems to vary roughly with the cube
of the rate of crack propagation, but the proportionality factor there-
between decreases as the hardness is reduced by tempering.
2. The time rate of emission in approach to fast fracture instability
Kie decreases markedly with tempering. Thus, the danger of fast fracture
instability would be more difficult to detect by acoustic emission in the
more stable materials.
3. There appears to be a correlation between the areal emission rate
and factors characterizing the degree and energy level of instability at the
crack tip, as determined from ordinary tensile stress-strain characteristics
of the material.
Acknowledgments
This work was carried out in the Ocean Technology Division of the
Naval Research Laboratory, J. P. Walsh, Superintendent, under sponsor-
ship of the Naval Ship Systems Command and the Office of Naval
Research. Authors are also indebted to several colleagues for help and
advice, particularly to S. D. Hart, L. C. Cardinal, and C. L. Lamb.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductio
References
[1] Kaiser, J., "Untersuchungen uber das Auftreten Gerauschen Bein Zugversuch",
Arch [ur das Eisenhuttenvesen, Vol. 24, pp. 43-45.
[2] Schofield, B. H., Bareiss, R. A., and Kyrala, A. A., "Acoustic Emission Under
Applied Stress," WADC Technical Report 58-194, ASTIA Document No.
AD155674, 30 April 1958.
[3] Hartbower, C. E., Gerberich, W. W., and Crimmins, P. P., The WeMing Journal,
Vol. 47, No. 1, Jan. 1968, pp. 15-18S.
[4] Boyle, R. W., Sullivan, A. M., and Krafft, J. M., Welding Journal (Research
Supplement), Vol. 41, No. 9, Sept. 1962, pp. 428S--432S.
[5] Brown, W. F., Jr. and Srawley, J. E. in Symposium on Fracture Toughness Test-
ing and Its Applications, A S T M STP 381, American Society for Testing of
Materials, 1964, pp. 186-188.
[6J Harris, D. O., Tetdman, A. S., and Darwish, F. A. in Acoustic Emission, ASTM
STP 505, American Society for Testing of Materials, 1972, pp. 238-249.
[7] Krafft, J. M. and Smith, H. L., "Ligament Instability Model for Stress Corrosion
and Fatigue Crack Propagation in 4340 Steel," Paper D-3 to be presented at the
International Conference on Stress Corrosion Cracking and Hydrogen Embrittle-
ment of Iron Base Alloys, Unieux-Firminy, France, 10-16 June 1973.
[8] Krafft, J. M. in Techniques o[ Metal Research, Vol. 5, Part 2, Wiley, New York,
1971, pp. 1-102.
[9] Wilson, W. K., Engineering Fracture Mechanics, Vol. 2, 1970, pp. 169-171.
Wolf Elber 1
Effects of Shot-Peening Residual Stresses

on the Fracture and Crack-Growth
Properties of D6AC Steel
REFERENCE: Elber, Wolf, "Effects of Shot-Peening Residual Stresses on

the Fracture and Crack.Growth Properties of D6AC Steel," Fracture Tough-
ness and Slow-Stable Cracking, ASTM STP 559, American Society for Test-
ing and Materials, 1974, pp. 45-58.
ABSTRACT: The fracture strength and cyclic crack-growth properties of
surface-flawed, shot-peened D6AC steel plate were investigated. For short
crack lengths (up to 1.5 ram) simple linear elastic fracture mechanics--
based only on applied loading---did not predict the fracture strengths. Also,
Paris' Law for cyclic crack growth did not correlate the crack-growth be-
havior. To investigate the effect of shot-peening, additional fracture and
crack-growth tests were performed on material which was precompressed to
remove the residual stresses left by the shot-peening. Both tests and analysis
show that the shot-peening residual stresses influence the fracture and crack-
growth properties of the material. This report presents the analytical method
of compensating for residual stresses and the fracture and cyclic crack-
growth test results and predictions.
KEY WORDS: fracture strength, fatigue (materials), steels, shot peening,
fracture strength, mechanical properties, crack propagation, steels
Nomenclature
A,B C o m p l i a n c e gage c a l i b r a t i o n constants
a D e p t h of surface c r a c k ( m )
b D i s t a n c e of c o n c e n t r a t e d force f r o m free surface ( m )
C Crack-growth constant
COD Crack-opening displacement (m)
c O n e half of surface length of surface c r a c k ( m )
](b/a) G e o m e t r i c function in stress-intensity solution
1 Aerospace technologist, NASA Langley Research Center, Hampton, Va. 23365.
45
Copyright9 1974 by
Downloaded/printed byASTM lntcrnational www.astm.org
KEc Stress intensity for an edge crack ( N / m 3/s)

Keff Effective stress-intensity range (N/m 3/z)
KIc Fracture toughness value (N/m 3/s)
KRES Stress intensity due to residual stresses (N/m ~/~)
Ks Stress intensity due to stress S (N/m 3/s)
Ksc Stress intensity for a surface crack (N/m ~/s)
n Crack-growth exponent
P Applied load or concentrated force (N)
a Surface crack shape factor
S Externally applied stress (N/m z)
Depth of shot-peening parameter (m)
AKef~ Range of stress intensity (N/m 3/s)
(7C Compressive residual stress ( N / m s)
O"T Tensile residual stress (N/m s)
da
Crack growth per cycle (m)
dN
When a material is shot peened, the residual compressive stresses at

the surface prolong the fatigue life. A recent investigation of the fracture
and cyclic crack-growth properties of shot-peened, surface-flawed D6AC
steel revealed some anomalies in these properties. For short crack lengths
(up to 1.5 ram) linear elastic fracture mechanics did not predict the
fracture strength. Also, Paris' Law for cyclic crack growth did not
correlate the crack-growth behavior. Separate relations resulted between
stress-intensity range and growth rate for different values of cyclic stress.
This behavior was previously reported in titanium alloys [1],s and has
become known as the stress-level effect. D6AC steel, whose basic mechan-
ical properties are given in Table 1, was chosen for the investigation be-
cause the material and crack growth and fracture data were available.
The study was designed to show that the residual stresses at the surface
caused these anomalies. The study was both experimental and analytical.
A simple mathematical model was constructed to evaluate the contribu-
tion of the residual stresses to the stress intensity. For an assumed residual
stress distribution, the effect of the residual stresses explained the dis-
crepancies between the experimental results and the linear elastic fracture
mechanics analysis. In a series of tests, specimens were precompressed to
a permanent compressive strain of 0.1 percent to remove the residual
stresses left by shot peening. Fracture and cyclic crack-growth tests
2The italic numbers in brackets refer to the list of references appended to this
paper.
ELBER ON SHOT-PEENING RESIDUAL STRESSES 47
TABLE 1--Chem&al composition and mechanical properties of D6AC steel.
Chemical Composition
Element Percent Mechanical Properties
C 0.48 Yield strength 1450 MN/m ~
Mn 0.83 Ultimate strength 1600 MN/m ~
P 0.01 Elongation 14%
S 0.005
Si 0.28
Ni 0.58
Cr 1.06
Mo 1.01
V 0.1
Cu 0.15
showed that the properties were in good agreement with linear elastic
fracture mechanics analysis.
Residual Stress Model

General
Most machining operations and surface treatments leave residual stresses
in the material. The distribution and depth of these residual stresses depend
on the particular process. The D6AC material used in this investigation
probably contained residual stresses caused by rolling, heat-treating, and
shot-peening processes. These residual stresses were not measured for
this study. Rather, their distribution was estimated from available data
of residual stresses caused by the machining and shot peening [1].
When a crack is growing through a shot-peened surface, the stress
intensity at the crack front is influenced by the residual stresses. The effect
of these stresses is to cause the crack to remain closed until the externally
applied stress can overcome the action of the residual stresses. The effec-
tive stress intensity of the crack tip is obtained by superimposing the
solution for the stress intensity due to external loading and the solution
for the stress intensity due to the residual stresses. The crack will open
when the effective stress intensity is positive.
Methods exist for the analysis of the stress intensity at the tip of a
crack due to these residual stresses. The distribution of stresses and the
stress-intensity calculation are described in the next two sections.
The Residual Stress Distribution

The residual stress distribution caused by shot-peening is compressive
near the surface and tensile to some depth below the surface. To simplify
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
the stress-intensity calculation a piecewise linear residual stress distribution

was assumed. The shape of this distribution is given in Fig. 1.
Three parameters define the assumed residual stress distribution. The
depth of the constant compressive stress is 3, the remaining stress distribu-
tion defined in multiples of 8. The compressive stress is ~c, and the peak
value of the residual tensile stress is ~rr. When a flaw exists in a shot-peened
surface, a surface crack originating at the flaw must grow through this
residual stress field. The model derived here is designed to calculate the
influence of the residual stress field on the fracture and cyclic crack-growth
behavior of such a surface crack.
Determination o] Stress Intensity

For a linear distribution of stress on the surface of an edge crack, a
closed-form solution for the stress intensity was obtained by Benthem
and Koiter [2]. This solution is shown in Fig. 2a. It was used for the
analysis of stress intensities for crack lengths shallower than 3.
To obtain a solution for cracks deeper than 3, the piecewise linear
distribution was divided into a series of concentrated force pairs. A solu-
tion for a single force pair is given in Ref. 3, and shown in Fig. 2b.
The residual stress intensity Ka~s was obtained by the summation
KltES= Ki ( 1)
where Ki is given in Fig. 2, and the Pi are the equivalent concentrated

forces substituted for the residual stress distribution.
The analytical approximation shown in Fig. 2 for the function f(b/a)
was obtained by fitting a curve to the numerical results presented in Ref 3.
m
t~
DISTANCE FROM
E
SURFACE
ac
0 6 i 6 _ 46 _
F I G . 1 - - E s t i m a t e d residual stress distribution.
2 Pi [;+r(b/~)]
Ki
KEC - ( 1.122 p+ 0.439q ) ~,/~a-- ,~- a~-b 2
f ( b l .:a)~ 0.35 ( l - b / a ) - 0.655 ( l - b / a) 7
F I G . 2--Stress-intensity solutions for edge cracks; ( a ) distributed stress and ( b )

concentrated Jorce.
The Surface Flaw

The solutions described for the stress intensity due to residual stresses
apply to an edge crack or to a surface crack of a/2c=O. To apply the
solutions to a surface crack of semicircular shape, the stress intensity for
the edge crack KEe is divided by the square root of the surface flaw shape
factor Q.
Ksc= K~c/ ~/ Q (2)

Fracture Strength
The fracture strength of a cracked component is determined by the
effective crack front stress intensity and the fracture toughness of the
material. The effective stress intensity is the sum of the residual stress
intensity K~Es, and the stress intensity from the applied stress S
K = Kr~Es+ Ks (3)
At fracture the effective stress intensity is equal to the material's fracture
toughness, so that
K ~ s + Ks= KIc (4)

The residual stress intensity K~Es can be obtained from the calculation de-
scribed by Eq 1. The stress intensity caused by the applied loading S is
given by
K s : 1.12 S ~ra (5)
Equations 1, 4, and 5 can be combined to obtain the fracture strength
S=(KI~--KREs)/1.12 V ~ - (6)
Cyclic Crack Growth

The rate of cyclic crack growth is assumed to be a function of the
effective stress intensity of the form
da
dN = C( AKaf)" (7)
For zero-to-tension loading, and a residual stress intensity K~ES, Eq 5

becomes
da
~-~=C (Ks +K~Es)" (8)
for (Ks+KREs) > 0 .
Experiments
General
As part of the F-111 Recovery Program, fracture strength tests and
cyclic crack-growth tests were conducted on specimens cut from several
plates of "low-toughness" D6AC steel. The general mechanical properties
of this material have been summarized in Ref 4. The specific mechanical
properties of the material tested are listed in Table 1. That report presents
results of studies from several laboratories including Langley Research
Center.
The tests reported herein were carried out on the same stock of material
as the tests reported in Ref 4. The plates from which specimens were cut
had been shot peened and cadmium plated. As part of the present study,
a small number of specimens were loaded to a permanent compressive
strain of 0.1 percent to eradicate the residual stresses caused by shot
peening. Figure 3 shows the distribution of the original residual stresses
through the thickness of the specimen. If the specimen is subjected to an
axial compressive load to a total strain of 0.15 percent, the stress dis-
tribution across the thickness will be almost uniform at the material's
yield stress. On unloading, the permanent strain will be approximately
0.1 percent with a final residual stress distribution as shown in Fig. 3.
For an elastic-perfectly plastic material, the final residual stresses would
be exactly zero. For the D6AC steel final residual stresses will be an order
FIG. 3--Effect of precompression on residual stresses.
of magnitude less than the initial residual stresses. These precompressed

specimens were then used to generate a new set of fracture strength and
crack-growth data for comparison with the data from the shot-peened
material.
Specimens
Hour-glass shaped specimens (Fig. 4) were cut from the plate stock
with the center line parallel to the rolling direction. Semicircular notches
were electromachined into the test section to start the fatigue cracks.
Fracture strength test specimens were precracked to the desired crack
depth at (275 M N / m 2) stress range by applying zero-to-tension cyclic
loads.
Instrumentation
Crack depth was monitored by the compliance technique using the
NASA COD-gage (Fig. 5). The gage length of this gage is 1.2 mm. The
electrical signals representing load and crack-opening displacement (COD)
FIG. 4--Specimen configuration (dimensions in mm ).
FIG. 5 - - N A S A COD--gage. Gage length 1.2 mm.
were displayed on an X - Y oscilloscope; compliances--and hence crack

lengths--were computed from photographic records of the display using
an equation
d (COD) = A +Ba
The constants A and B had previously been determined for the gages
used.
Loading
The specimens were loaded uniaxially in a 1.8-MN-capacity, servo-
hydraulic testing machine. The test frequency for cyclic crack-growth
tests was 3 Hz. The load rate for fracture tests was 15 KN/s.
Environment
Cyclic crack-growth tests were conducted in laboratory air at 283 K and
70 percent relative humidity. Fracture strength tests were conducted in a
dry gaseous nitrogen atmosphere at 233 K.

Fracture Strength
Figure 6 shows the fracture strengths for eight shot-peened specimens
tested under cryogenic conditions. Under these conditions, as defined in
the section dealing with environment, the material has toughness values
which result in critical crack depths comparable to the depth of the
compressive stress layer and which therefore are more influenced by these
residual stresses. The solid line is a line of constant stress intensity fitted
to the fracture strength data for the four longest crack lengths. The data
for short crack lengths deviate markedly from the constant stress-intensity
line.
Four additional fracture specimens were tested. These specimens were
precompressed to a permanent residual strain of 0.1 percent to remove
the residual stresses caused by the shot-peening process. The fracture
strengths for these specimens are shown in Fig. 7. The solid line in that
figure is the same as the constant stress-intensity line shown in Fig. 6.
The fracture strengths obtained for the shorter crack lengths in precom-
pressed specimens do not deviate significantly from the line of constant
stress intensity. This leads to the conclusion that the apparent higher
toughness for short crack lengths in the shot-peened material (Fig. 6) is
caused by residual stresses from the shot peening.
2oo0
% 1600 D
o o
1200
800
400
I I I
1 2 3
CRACK DEPTH , mm
FIG. 6--Fracture strength of D6AC (shot peened).
2OOO
12oo
8oo
.Ks c= 48 MN/m3/2
Q..
400
I I I I
0 1 2 3 4
CRACK DEPTH ,mm
FIG. 7--Fracture strength of D6AC (shot peened and precompressed).
To test this conclusion, fracture strength calculations were made with

the residual stress model. Model parameters were selected to simulate the
residual stress distribution assumed to exist in the shot-peened material.
The parameters were 8 = 0.5 mm, -c = 315 M N / m z, ~r = 240 M N / m 2, and
K~c=48 M N / m 3/2. Figure 8 shows the predicted fracture strength as a
function of crack length for a plate containing a semicircular surface flaw
and having a residual stress distribution shown in Fig. 1. The dashed line
is a line of constant stress intensity based only on external loading
( K s = 4 8 MN/m2). For short crack lengths, the model predicts that
fracture strengths are higher than predicted by considering the external
loading only. This result is consistent with the test results from the
shot-peened specimens (Fig. 6).
Cyclic Crack Growth

Figure 9 shows the crack-growth rate for zero-to-tension loading as a
function of the stress-intensity range for the shot-peened material. The
data from tests at three stress levels (275 M N / m 2, 414 M N F , and 690
M N / m 2) fall along distinctly separate curves. The dashed line represents
cyclic crack-growth data obtained from compact tension specimens in the
investigation in Ref 4. In that investigation cracks were grown from the
edge of the plate and, therefore, were less affected by the shot-peening
residual stresses.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions author
2000
~E
z
~E
-
1600
800
~ ODEL PREDICTION
.-~. K. C = 48 MN/m 3/2
400
I I I I
1 2 3 4 5
CRACK DEPTH , mm
FIG. 8--Fracture strength prediction.
10-7
I
j./.S, 1~.
f
coMP
ACTTENSON
RESULTS ( REFERENCE 4 )
/"
10-8
<.
10-9
9 S--275
10-10 I I I I I
10 20 30 40 50 60
STRESS INTENSITY RANGE , MN/ m3/2
FIG. 9--Crack-growth rates for D6AC (shot peened).
Figure 10 shows cyclic crack-growth data from three specimens after

a precompression cycle. The stress levels in this investigation were identical
to those for the data of Fig. 9. The dashed line in Fig. 10 is the same
reference line as shown in Fig. 9.
The data from the precompressed specimens agree with the data from
the compact tension tests, and can be described by Paris' Law, where
~-~=cda (AK)'~ (9)
and
C = 1.87x 10-I~ n=2.72

To show that the stress-level effect shown in Fig. 9 for shot-peened
specimens is caused by the residual stresses, the effect of the residual
stresses on the cyclic crack-growth rate was calculated with Eq 8. The
model parameters were the same as those used for the fracture strength
calculations.
Figure 11 shows the calculated cyclic crack-growth rates as a function
of the stress-intensity range Ks, the stress intensity due to the external
loading. The three curves labeled with stress levels represent the computed
crack-growth behavior for those three stress levels used in the experiment.
The curve labeled "No Residual Stress" represents the basic Eq 9.
FIG. lO---Crack-growth rates for D6AC (shot peened and precompressed).
ELBERON SHOT-PEENING RESIDUALSTRESSES 57
10-7
S=690 ~ RESIDUAL
S=41~ STRESSES
10-8
t.u
-n-
10-9
c.)
10-100 / I
I0
I
20
I
30
t
40
I
50 60
STRESS INTENSITY RANGE , MN/m3/2
FIG. 11--Crack-growth rate prediction.
The model calculations showed that the residual stress intensity was
negative for crack lengths smaller than 1.3 mm, and was positive for
crack lengths larger than 1.3 mm.
For crack lengths shorter than 1.3 mm the crack-growth rates were
lower than predicted by Paris' Law; for crack lengths larger than 1.3 mm,
crack-growth rates were faster than predicted by Paris' Law.
The results in Fig. 11 show that for lower values of applied stress the
crack-growth rate is higher for the same stress-intensity range. This result
is qualitatively the same as that obtained experimentally from the shot-
peened material.
Conclusion
1. The fracture strength and cyclic crack-growth properties of D6AC
steel were affected by the residual stresses left by the shot-peening.
2. Compression residual stress near the surface caused shallow cracks
to grow more slowly than observed for cases without residual stress.
Tension residual stresses below the shot-peened layer caused deeper
cracks to grow more rapidly than observed for cases without residual
stress.
3. Compression residual stresses near the surface gave shallow cracks
an apparent fracture toughness higher than the fracture toughness of the
stress-free material. Tension residual stresses below the shot-peened layer
gave deeper cracks an apparent fracture toughness lower than the fracture
toughness of the stress-free material.
4. A simple model based on the contribution of the residual stress
to the effective stress intensity explains the trends in both fracture strength
and cyclic crack growth.
References
[1] Koster, W. B., Field, M., Fritz, L J., Gatto, L. R., and Kahles, J. F., "Surface
Integrity of Machining Structural Components," Technical Report AFML-TR-
70-11, March 1970.
[2] Bentham, J. P. and Koiter, W. T., Asymptotic Approximations to Crack Prob-
lems, Mechanics o[ Fracture, Noordhoff International Publishing, Leyden, 1972.
[3] Hartranft, R. J. and Sih, G. C., "Alternating Method Applied to Edge and Surface
Crack Problems," Lehigh University, Technical Report IFSM-72-13, April 1972.
[4] Fedderson, C. E., Moon, D. P., and Hyler, W. S., "Crack Behavior in D6AC
Steel," Metals and Ceramics Information Center, MCIC-72-04, Jan. 1972.
E . J. R i p l i n g 1
Fracture Properties of a
Cold-Worked Mild Steel
REFERENCE: Ripling, E. J., "Fracture Properties of a Cold-Worked Mild

Steel," Fracture Toughness and Slow-Stable Cracking, ASTM STP 559,
American Society for Testing and Materials, 1974, pp. 59-73.
ABSTRACT: An AISI 1018 steel was rolled various amounts between 0 and
50 percent, after which it was partially stress relieved, and the following
properties measured: smooth tensile, Charpy V-notch versus test tempera-
ture, fracture toughness, stress corrosion cracking susceptibility, and fatigue
crack growth rate.
The yield strength was increased almost three-fold by rolling, and the
upper-shelf Charpy V-notch toughness was reduced to approximately one-
third its unrolled value. The fracture toughness was only moderately
changed, and the crack growth rate in fatigue (da/dN) unchanged by roll-
ing. There was no subcritical crack growth in distilled water on sustained
loading to K~ values just less than K~c for periods up to 64 h; however,
da/dN in distilled water was higher than its dry value if fatigue loading was
slow.
KEY WORDS: fracture properties, steels, cold working, fatigue (materials),
toughness, yield strength
By far the most widely used structural metals are non-heat treated
mild steels. One reason for their popularity is that they are readily cold
worked; and since this process is so c o m m o n , it raises the question of
h o w working the steel influences the performance of structures m a d e
f r o m it. W o r k i n g increases the yield strength while reducing the ductility,
and this increased yield strength increases the tendency toward brittle
behavior for two reasons. ( 1 ) Fracturing is more likely to occur u n d e r
a condition of plane strain than plane stress, and the related change in
crack tip stress state decreases the inherent toughness of the steel. ( 2 )
Higher stresses are likely to be applied to members having higher yield
1 Director of Research, Materials Research Laboratory, Inc., Glenwood, Ill. 60425.
59
Copyright*1974byby ASTM lntcrnational
strengths, due to both accidental overloads and higher design stresses.

The largest stress that the structure might experience due to accidental
overloading is frequently limited to the yield stress. For example, if a
member is impacted, energy is generally dissipated by plastic flow; hence,
increased yield strengths increase the stress that can be applied accidentally.
Increased design stresses are a result of the fact that the allowable working
stress is frequently selected as some fraction of the yield strength. On
this basis, cold work increases the allowable working stress, raising the
question of whether or not such a procedure is sound design.
Mild steel structures are commonly used out of doors and are subjected
to alternating loads. Defining the performance of structures built from
such steels requires that their fracturing characteristics be described by
laboratory tests that consider these service conditions. Hence, the follow-
ing properties were measured: fracture toughness, stress corrosion crack-
ing resistance in the presence of water, crack growth rates under alternat-
ing loads in a reasonably dry environment (50 percent relative humidity),
and crack growth rates under alternating loads in the presence of water.
Cold working in this program was all by unidirectional rolling so
that the mechanical properties of the finished plate would be expected
to be anisotropic. The change in properties produced by working, there-
fore, will depend on the relationship between the test specimen direction
and the rolling direction. For the product of interest in this program, the
long transverse direction was the critical one. Hence, although this is
a direction of essentially zero strain in rolling, all test specimens were
taken in this direction.
Materials and Procedure

Two bars, 12 ft long by 2 by 3 in. cross section, of cold-drawn AISI
1018 steel, which were identified as belonging to a single heat, were
purchased from a local warehouse. One of these bars was used for de-
termining rolling procedures and stress relief treatments; having estab-
lished procedures on the first bar, the second one was long enough so
that all the specimens were cut from it. The composition of the steel is
given in Table 1.
The bars were normalized by heating at 1700~ for 1 h followed by
a still air cool. Sections were then rolled 12.5, 25, 37.5, and 50 percent
TABLE 1--Composition of steel used.
C Si Mn P S
0.21 <0.05 0.79 0.007 0.024
RIPLING O N COLD-WORKED MILD STEEL 61
TABLE 2 Hardness values after stress relief.
Average Hardness (R~)

Testing
Temperature, One-fourth
~ Midthickness thickness
500 98 94
600 99 96
700 99 96.5
800 98 95
900 96 92
1000 94 91
in a two-high, 8-in.-diameter rolling mill. Portions of the 50 percent

reduced bar were stress relieved for 1 h at temperatures from 500 to
1000~ The hardness values at the midthickness and one-fourth thickness
positions after stress relief are shown in Table 2. It was apparent that
there was no significant hardness loss on stress relieving at temperatures
up to 800~ Beyond this, however, the hardness decreases appreciably.
A short section of the bar, rolled 50 percent with no stress relief
(NSR), was saw cut on its center plane for approximately 8 in. in order
to make a coarse measurement of residual stress. Another rolled section
was stress relieved at 800~ and also slit. The amount of opening in the
latter, although still appreciable, was about one-half that of the "as-
rolled" bar. Hence, since it appeared that the residual stresses were re-
duced by about one-half by stress relieving at 800~ while the hardness
was not appreciably changed, this temperature was selected for stress
relieving. The microstructures of the bars, after working and stress
relief, are shown in Fig. 1.
After rolling and stress relieving the bars, threaded end, 0.252-in.
diameter tensile, Charpy V-notch (CVN) and 1-in.-thick ( 1 T ) compact
toughness specimens (CTS) were all machined in the WR direction, as
shown in Fig. 2. The CTS were used for fracture toughness testing, crack
growth under fatigue loading, stress corrosion cracking, and corrosion
fatigue. One set of specimens taken from the bar rolled 50 percent were
not stress relieved. By comparing the data collected on these with the
rolled and stress relieved specimens the effect of stress relief was ascer-
tained. All testing was done at room temperature, with the exception of the
CVN specimens which were tested over a range of temperatures to define
the transition temperature.
FIG. lmMicrostructure of steel rolled as indicated and stress relieved ( X 250).

Tensile Properties
The effect of cold work plus stress relief on the long transverse tensile
properties of A I S I 1018 steel is shown in Fig. 3. After an initial fast in-
crease in tensile and yield strength on rolling the first 12.5 percent, the
rate of increase was considerably reduced.
The effect of stress relieving on the tensile properties can be evaluated
RIPLING ON COLD-WORKED MILD STEEL 63
FIG. 2--Direction in which specimens were taken.

by comparing the two sets of data collected after 50 percent rolling since
one set of specimens was stress relieved, and the other not. It is apparent
that this treatment had only a minor effect on tensile properties.
Charpy V-Notch
The Charpy toughness of the steel in each of the six worked condi-
tions is shown in Fig. 4 as a function of testing temperature.
The largest loss in toughness occurred between 0 an 12.5 percent
rolling. At the super-transition (that is, upper shelf) temperatures, the
CVN toughness dropped approximately 30, 5, 5, and 0 ft.lb for each
12.5 percent increment of rolling. In the transition range, the major
toughness drop again occurred with the first increment of rolling after
which the toughness was reasonably constant.
Fracture Toughness
The measured values of fracture toughness of the steel as a function of
amount of rolling is shown in Fig. 5, and a photograph of typical fracture
surfaces in Fig. 6. With the exception of the two materials rolled 50 per-
cent, all of the specimens showed very small shear lips and a typical brittle,
cleavage fracture. The 50 percent rolled specimens, however, showed a
considerable amount of lamination in the direction of crack extension.
This structure was far more pronounced for the specimens that were not
stress relieved than for those that were.
The curve in Fig. 5 was drawn through the average toughness value at
each reduction: 0, 12.5, 15, 35, and 50 percent. The shape of the curve
suggests that intermediate values of cold work (25 and 37.5 percent)
produced a slight minimum in toughness although not enough tests were
conducted to be sure that this is the case. In any event, the average
toughness values after rolling from 0 to 50 percent only vary from 36 to
50 ksiVin, while the yield strength was raised from 33 to 93 ksi. Rolling
has far less effect on toughness than on yield strength.
In describing the data in Fig. 5, it must be pointed out that all of it
cannot be considered valid according to ASTM Test for Plane-Strain
Fracture Toughness of Metallic Materials (E 399-72). Fatigue pre-
110
100
9O
8O
-.; 70
6O
!:,0
40
30
~o~ I l I I
50 ~ ~ n R e d u c t i o n in A r e a
= 9 I ] 4
1 ]
0 !0 20 30 40 30
Reduction - percent
F I G . 3--EfJect of cold work on tensile properties for AISI 1018 steel.
50
!
/.J.
40
30 J O"
20 /
"7
/
0
I0
i ,i,-/ ,
0% (asnormalized) .0" 1Z89 (stress relieved
0 I I
o
t>
20 ' 9
25%(stress relieved)~ 37~(stress relieved)
al
tD 10
o
i 0~.. 9 I
20
'50% (St/elSeSved)
0/ 50% (not stress J
r e I ~
I0
.Y yc~ -'o
I
100 Z00 300 0 I00 200 300
Test Temperature - OF
FIG. 4---Effect of cold work on Charpy V-notch energy [or A I S I 1018 steel.
cracking was done in a manner consistent with this specification although

the crack front for some of the 50 percent NSR specimens was excessively
bowed. It can be seen from Fig. 6 that the fatigue cracks, other than for
the latter specimens were quite straight.
The load was applied by means of ASTM tension testing clevises, and
University of Washington (University of Washington) pursuant to License Agreement. No further reproduction
66 FRACTURE TOUGHNESS AND SLOW-STABLE CRACKING
100 9 Popped J
OA No pop I ~ 1
a No~ ~t . . . . . . . I ~ I
lieved ~- J
Yield Strengti-,
~; //t =z.5
0 10 20 30 40 50
Keduction - percent
FIG. 5--Effect of cold work on toughness o] AISI 1018 steel.
FIG. 6--Fracture appearance of typical toughness specimens ( B = I in.).
displacement measured by means of a clip gage that fit into the machined
integral knife edges. Loading time was of the order of 1 min. Distinct
"pops" occurred on many of the load, P, displacement, ~x, curves for
reductions of 0, 12.5, 25, a n d 37.5 percent, but not for either of the
50 percent reduced specimens. For those tests where pop-in did not
occur, the 5 percent secant off-set method was used to select the critical
load. Again the curve shapes were consistent with the requirements of
ASTM E 399-72.
Crack length was measured on the fracture surface at the completion
of the test. No ry corrections were made on crack length. Had such
corrections been made, the indicated values of toughness would have
increased by the largest amount for the least worked specimens because
they had the lowest yield strengths. The fracture appearances gave no
indication of stable cracking prior to the onset of rapid fracture with the
possible exception of the 50 percent worked specimens. The lesser worked
specimens that "popped" did show evidence of arrest markings.
The specimens were not always thick enough for the data to be con-
sidered valid according to the criterion:
B - ~( Kic y (1)
- - \ (Tys /
where
B = specimen thickness = 1 in., if
t~= 2.5.
Only the three largest reductions, 25, 37.5, and 50 percent, produced a
sufficient elevation of the yield strength (from 33.4 to < 7 0 ksi) for the
critical values to be identified as K~c (see ~--2.5 curve in Fig. 5).
However, other programs at the Materials Research Laboratory ( M R L )
and U. S. Steel Research Laboratories suggest that for mild steels ~ = 1
rather than 2.5 may be sufficient to approximate a value of plane-strain
toughness2, 3 That is, the value of toughness only need exceed that of
the yield strength for the 1-in.-thick specimen used in this program. On
the basis of this criterion, essentially all of the data in Fig. 5 is useful.
In spite of the fact that the toughness is only moderately changed by
rolling, the critical crack size for accidental overloads would be decreased
with increased amount of cold work. As just discussed, the maximum stress
that can be developed during accidental overloads is frequently limited to
Crosley, P. B. and Ripling, E. L, Journal of Basic Engineering, Sept. 1969, p. 525.

3This conclusion by MRL was based on the use of contoured double cantilever
beam specimens. Whether or not the use of a less severe value of ~ can be used with
CTS is not known.
the yield stress. Assuming that a typical flaw would be a semielliptical

one at the surface of the member, and having a ratio of depth to length,
that is, a/2c of 0.25, the critical crack depth on accidentally loading to
the yield strength would depend on the amount of cold reduction as
shown in Fig. 7.
Structural design is frequently based on the use of a "factor-of-safety"
(FS) where the latter is defined as the ratio of yield stress, aYS, to work-
ing (or design) stress, ~ ; that is FS=~Ys/~w. Since cold working raises
~YS, it also raises ~w if FS is held constant. Increasing the aYS by a process
that does not produce a corresponding increase in toughness, however,
can lead to a severe loss in crack tolerance of structures designed by a
criterion involving a constant FS, as shown in Fig. 7.
Stress Corrosion Cracking

Stress corrosion crack (SCC) susceptibility was also evaluated by using
fatigue precracked IT compact tension specimens. The corrosion medium
was distilled water containing a wetting agent (0.5 weight percent Kodak
Photoflo). (The wetting agent was used because other studies at MRL
1. O0
0.8
!
0.6
r of Safety = 1.5
0.4
O.Z Stress Loading

~ f Safety = 1.0)
I
0 10 20 30 40 50
Reduction - percent
F I G . 7--Effect of cold work on critical crack size for yield stress loading and 2/3
yield-stress loading.
suggested that it may reduce incubation time.) Since no SCC was ex-
pected, only two specimens were tested, one rolled 25 percent, and the
other 50 percent, both stress relieved. Exposures were conducted for
times up to 64 h over a range of applied K's from 30 to 40 ksiVTff. No
measurable crack extension occurred, even at 40 ksi~i-m, which was just
less than the critical values of KI.
Fatigue Crack Growth Rate

Crack growth was measured for a range of a K ' s from approximately
25 to 40 ksiVTm. The ratio of alternating to mean K was 0.95; the fre-
quency, 20 Hz; the load-time curve, sinusoidal; and the relative humidity,
approximately 50 percent.
Cracks were extended 20 to 50 mils at a constant load, and AK cal-
culated for the interval assuming the crack length, a, was equal to the
value it had at the middle of the interval. These values of AK were
plotted with the corresponding values of crack growth per cycle, da/dN,
for each interval. Two specimens were used for evaluating each step of
rolling, and the data collected on each specimen is shown as open or
closed circles in Fig. 8. It is apparent that over the complete range of
tests, the data can be represented by the power law:
The data for each pair of specimens representing one rolling condition
was used in a computer program to find the values of C and n that gave
the best least-square's-fit to Eq 2, and the appropriate expressions are also
given in Fig. 8. A comparison of the values of C and n show that there is
no effect of rolling on fatigue crack growth rate although the specimens
that were not stress relieved showed a slightly higher cracking rate than
those that were relieved. One value of C and n appeared to describe all
of the data collected with a K ' s over the range from less than 20 to more
than 40 ksiVTm. Since the data could all be fitted to a single value of n
it must have all been collected below the transition zone ' in spite of the
fact that fatiguing was conducted at K values close to its critical one.
Corrosion Fatigue
The crack growth rate was also measured in stagnant water (plus
wetting agent) on specimens from each rolling condition. To compare
cyclic crack growth rate in the presence and absence of water, the " d r y "
4The value of Kx at which fatigue crack propagation is accelerated as the stress-

intensity factor approaches its critical value.
e
i0 -4
9 Spec. A
o Spec. B
I
!
I
/I /
II
I III
I III
Iit
J
q
r
b!
lz.5% Z5%
b
o
~o
i
g eP I
o9
U I
@
e o
,da/dN o da/dN = da/dN =
1 3 . 3 x 10-12~K 4" 5 ~) 2 . 4 x 10-12~K 4" I9 7.0 x 10-12&K 4"2
9 , , , , ,
I i
10 20 30 50 80 10 20 30 50 80 10
; 20 30
IIIIIL
50 80 I00
Stress-lntens~ty-Factor Range, A H I - ksl - "V'[-ff.
FIG. 8--Crack growth rate as a function of stress-intensity-factor range (20 Hz,

50 percent relative humidity).
d a / d N curves in Fig. 8 were replotted in Fig. 9 for each step of cold

working. The data points in the figure were collected using three different
loading profiles which are identified by the key shown in the bottom of
Fig. 9.
In the first test series, the load-time curves were sinusoidal, as was the
case for the "dry" tests, but the frequency was reduced from 20 to 1 Hz.
As shown in Fig. 9 by the open and dosed circles, the water environment,
even with the reduced cyclic frequency only had a minor effect on
cracking rate, the data tending to scatter upward from the "dry" values.
A second series of tests were run using a much slower cyclic rate,
that is, 2 cpm, consisting of 5 s at the low load, 10-s rise time, 5 s at the
high load, and 10-s decay. The dwell at the low and high loads were
RIPLING ON COLD-WORKED MItD STEEL 7]
~ 10-~
/
I
o o
(
+,e
o 4D
~ 1~ ~ 0
Qc I I
,P
IA, I I
D +!
3789 s, %
0 I
O
I ~
V
9 da/dN = i J, Oda~
Uda/dl',I =
eI 3.0 x 1 0 - 1 2 A K 4"-~ / 3 . 4 x 10-12A i,(4"6
10 -(
/ + iiiiii
10 20 30 50 80 10 20 30 50 80 10 20 30 50 80 100
S t r e s s - l n t e n s i t y - F a c t o r Range ",K I - kst--~q~-.
FIG. 8--Continued.
necessary because of the type of equipment used. As seen in Fig. 9 by the

X's, this slower cycling increased the crack rate. The fact that an aggres-
sive environment may accelerate fatigue crack growth even below the
threshold value for stress corrosion cracking, Kiscc, is not surprising
since it has been reported by Barsom ~ and by Gallagher. 6
Barsom has also shown that for some systems, at least, it is the loading
portion of the cycle that causes cracking so that frequency is important
only as it influences loading time. 5 To evaluate the effect of rise time, one
~Barsom, J. M., "Effect of Cyclic-Stress Form on Corrosion-Fatigue Crack

Propagation Below Kxscc in a High-Yield-Strength Steel," International Conference
"Corrosion Fatigue," 1971.
o Gallagher, J. R., lournal of Materials, Vol. 6, No. 4, 1971, p. 941.
E~
10
-4
7
u
z5%
~ 1 0 -5
III 7
I
5~0 15 i 1 0 ~
10-6
Time ~ (seconds) I ~ I I
I
10 Z0 30 50 80 l0 Z0 30 50 80 t0 Z0 30 50 80 100
Stress-Intensity-Factor Range, AK I - ksi-~/~.
FIG. 9--Corrosion fatigue rates in distilled water (plus wetting agent) at 1 Hz and
30 Hz.
specimen, rolled 25 percent, was tested at 1 Hz in water using a positive

saw tooth loading profile having the shape shown in Fig. 9. Changing
from sinusoidal to the saw tooth had almost no effect on cracking rate.
Of course, even with the positive saw tooth, the rise time for these tests
(1 s) was only one tenth as long at it was for the slower tests.
Conclusions
1. Cold rolling a mild steel bar (AISI 1018) caused large changes in
yield strength and super-transition (that is, upper shelf) Charpy V-notch
toughness for specimens selected in the long transverse direction. Fracture
toughness, on the other hand, appears to be only moderately affected,
although not all toughness values were valid according to ASTM E 399-72.
2. This mild steel does not appear to exhibit subcritical crack growth
/
10 -4
/ ?
m_ ++~1 i
+, 9
.
I0-5
3789 50y0
/ I ,
50% nsx
/
li
I
/
10-6 ~ /
10 20 30 50 80 I0
m
/
20 30 50 80 10
/ 20 30 50 80 100
Stress-Intensity-Factor Range, AK I - ksi-~-ff.
FIG. 9~Continued.
in distilled water containing a wetting agent under sustained load even at

applied values of K~ just less than KTc for periods up to 64 h.
3. The fatigue crack growth rate at 50 percent relative humidity and
20 Hz is described by the power law d a / d N = C A K n at cracking rates
from 10-8 to 10 -4 in./cycle. The values of C and n are essentially inde-
pendent of the amount of rolling for material stress relieved 1 h at 800 ~ F.
4. Fatigue crack growth rate is increased in a stagnant distilled water
(plus wetting agent) environment if the cracking rate is slow (10-s rise
time) but is essentially unaffected if the rate is high (1 H z ) .
Acknowledgments
This program was partially funded by the Frankford Arsenal. The
many helpful suggestions of H. Markus, J. Corrie, J. Mulherin, and C.
Carman are much appreciated.
J. G. K a u [ m a n 1 a n d F. G . N e l s o n ~
More on Specimen Size Effects

in Fracture Toughness Testing
REFERENCE: Kaufman, J. G. and Nelson, F. G., "More on Specimen

Size Effects in Fracture Toughness Testing," Fracture Toughness and Slow-
Stable Cracking, ASTM STP 559, American Society for Testing and Mate-
rials, 1974, pp. 74-85.
ABSTRACT: A new study of the effect of specimen size on the results of
plane-strain fracture toughness tests of a relatively tough aluminum alloy,
2219-T851, suggests that an increase in specimen size requirements may be
necessary to assure size-independent test results. Specifically, it appears that
the crack length limit should be increased to 5(K~c/~rYS)2, which has the ef-
fect of keeping the maximum nominal net-section stress below two thirds of
the yield strength. The current limit on thickness could be maintained at
2.5 (Kxo/eys)2, although there is some evidence that it might be relaxed
further. Additional work is needed to check the generality of these sugges-
tions for other materials, as well as the possibility that if W/B=2 and
a=B ~ 5(Kic/~ys) 2, K ~ x may be useful as an engineering estimate of KI~.
KEY WORDS: fracture properties, toughness, aluminum alloys, thickness,
yield strength, fatigue (materials)
It is the p u r p o s e of this p a p e r to p r e s e n t a d d i t i o n a l i n f o r m a t i o n o n the

effect of s p e c i m e n size o n the results of p l a n e - s t r a i n toughness tests of
relatively high toughness a l u m i n u m alloys. Previous d a t a p r e s e n t e d b y
Nelson, Schilling, a n d K a u f m a n , 2 in 1970, s h o w e d that when s p e c i m e n
thickness a n d c r a c k length were less t h a n the limits r e q u i r e d b y A S T M
T e s t for P l a i n - S t r a i n F r a c t u r e Toughness of M e t a l l i c M a t e r i a l s (E 3 9 9 - 7 2 ) ,
the test results c o u l d b e h i g h e r o r l o w e r t h a n the true value of K~c, d e p e n d -
1 Manager and senior engineer, respectively, Engineering Properties and Testing

Division, Alcoa Laboratories, Aluminum Company of America, Alcoa Center, Pa.
15069.
2Nelson, F. G., Schilling, P. E., and Kaufman, J. G., Engineering Fracture
Mechanics, Vol. 4, 1972, pp. 33-50.
74
Copyright*1974byby ASTM International
KAUFMAN AND NELSON ON SPECIMEN SIZE EFFECTS 75
ing upon the particular combination of thickness and crack length. The
general conclusions, however, were that so long as specimens above the
current limiting thickness were utilized, relatively consistent values of K~e
should be obtained.
Since that time, additional data have suggested that there is some
specimen size effect for relatively tough materials for sizes substantially
larger than the ASTM limit, and therefore the study of 2219 presented in
the original paper has been substantially expanded. It is the purpose of
this paper to present the results of the further study.
Material
One lot of 3-in.-thick 2219-T851 plate produced commercially at
Alcoa's Davenport Works was used for this investigation. The composi-
tion and tensile properties are shown in Table 1. Both meet the require-
ments of applicable specifications for this alloy and temper, and indicate
that the sample is representative of good commercial quality.
Procedure
Compact specimens of the sizes listed in Table 2 were taken from the
single lot of 3-in.-thick 2219-T851 plate. All were of the ASTM E 399-72
plan proportions in Fig. 1 with W / a = 2 . 0 , but of various thicknesses to
provide W / B ratios from 2 to 12.
Initially, a series of V2-in.-thick specimens, with 1.5-in. crack lengths
were taken from several locations through the thickness of the plate;
when these data, though invalid as K~e values, showed that the toughness
was relatively uniform through the thickness, as shown in Table 3, the
remaining specimens were taken without any consistent regard to location
through the thickness, but with the following general pattern: all speci-
TABLE 1--Tensile properties o/ some 3-in.-thick 2219-T851 aluminum alloy plate.
Location Tensile Yield Elongation

Test in Strength, Strength,~ in 2 in.,
Direction Thickness ksi ksi %
Longitudinal T/2 64.1 50.1 10.0
T/4 65.8 50.7 10.0
Transverse T/2 64.8 49.3 9.8
T/4 65.9 49.2 9.8
Specified Minimums b 62 45 6
(2.001--3.000)
Offset equals 0.2 percent; gage length 2 in.

Aluminum Standards and Data, 1972-73 Edition, The Aluminum Association.
T A B L E 2--Results oJ some compact fracture toughness tests o/ various sizes o/specimen from 3-in. 2219-T851 plate "q
( L-T orientation), o,
Plane Strain
Factor, Po
Nomi- Stress Intensity a B Net
Nominal Nominal hal Load, kips ksiVin. Ratio Section
Thickness, Width, Ratio, Length, (KQ/ (KQ/ Pm~/ Stress,
NSS/r C
B, in. W, in. W/B a, in. P~ Pmax Ko Kmax o-xs) 2 o-~rs)'z P~ ksi o-r
0.50 1.00 2 0.52 1.34 1.58 27.5 32.5 1.77 1.70 1.18 51.2 1.01 z
0.52 1.28 1.60 26.0 32.5 1.91 1.98 1.25 49.1 0.97
26.8 32.5 >
Z
2.00 4 1.00 2.12 2.86 28.9 38.9 3.08 1.54 1.35 42.5 0.84
1.00 2.28 2.93 31.0 39.8 2.68 1.34 1.29 45.8 0.91 ~,
30.0 39.4
3.00 6 1.50 2.95 3.88 32.7 42.9 3.60 1.20 1.31 39.4 0.78
1.50 3.05 3.98 33.6 43.7 3.43 1.11 1.30 40.8 0.81
m
33.2 43.3 c~
5.00 10 2.54 4.02 5.70 35.7 50.6 5.11 1.01 1.42 31.9 0.63
2.60 3.92 5.62 35.8 51.3 5.21 1.00 1.43 30.2 0.60 _~
z
35.8 51.0 O
6.00 12 3.01 4.80 6.72 38.0 53.2 5.37 0.89 1.40 32.1 0.64
3.00 4.75 6.88 37.4 54.1 5.52 0.92 1.45 31.7 0.63
37.7 53.6
1.00 2.00 2 1.01 4.40 5.12 30.2 35.2 2.79 2.76 1.16 44.0 0.87
1.01 4.40 5.10 30.4 35.2 2.75 2.72 1.16 44.0 0.87
30.3 35.2
3.00 3 1.53 5.75 6.62 32.7 37.7 3.59 2.35 1.15 37.7 0.75
1.52 5.90 6.72 33.3 38.0 3.45 2.26 1.14 38.7 0.77
33.0 37.8
4.00 4 2.07 7.10 8.10 35.9 41.0 4.03 1.94 1.14 34.2 0.69
2.03 7.25 8.50 35.5 41.7 4.04 1.99 1.17 35.7 0.71
35.7 41.4
5.00 5 2.58 7.85 9.60 35.2 43.1 5.22 2.02 1.22 30.5 0.60
2.61 7.92 9.25 36.4 42.5 4.93 1.89 1.17 30.4 0.60
35.8 42.8
6.00 6 3.01 9.45 11.75 37.4 46.5 5.40 1.80 1.24 31.5 0.62
3.07 9.40 11.40 38.2 46.3 5.28 1,72 1.21 30.6 0,61
37.8 46.4
1.50 3.00 2 1.55 8.40 9.48 32.6 36,8 3.66 3,53 1,13 37.2 0.74
1.55 8.50 9.45 32.9 36.6 3.59 3.44 1.11 36.7 0.73
32.8 36.7 C
5.00 3 89 2.62 11.25 12.70 34.7 39.1 5.50 3.23 1.13 28.6 0.57
z
2.62 11.90 13.00 36.5 a 39.9 4.94 2.83 1.09 30.2 0.60
35.6 39.5 z
io
6.00 4 3.17 12.35 14.50 35.2 41.3 6.41 3.04 1.17 26.1 0.52 z
3.05 13.35 15.60 35.8 41.8 5.98 2.94 1.17 29.2 0.58
35.5 41.6 o
z
2.00 4.00 2 2.09 14.00 15.10 35.9 ~ 38.7 4.07 3.90 1.08 33.5 0.66 o
2.09 13.90 15.00 35.8 a 38.7 4.10 z
3.92 1.08 33.2 0.66
35.8 * 38.7
5.00 21/2 2.68 15.00
N
16.20 35.9 ~ 38.8 5.22 3.90 1.08 27.9 0.55
2.70 14.70 15.95 35.9 * 39.0 5.26 3.87 1.09 27.3 0.54 z
35.9 ~ 38.9
6.00 3 3.14 18.15 19.50 38.1 ~ 40.9 5.42 3.46 1.07 29.0 0.57
3.07 18.30 20.15 37.1 " 40.6 5.59 3.65 1.10 29.9 0.59
37.6 ~ 40.8
"Fully valid values of Kxc per ASTM E 399-72.

.,q
.,q
2H
:t IF
.
'~E-
L
C
o
III1[
i S E C T I O N CC
.m
A .,,--J_J ~ 1o,A
76176 SECTION A A
PROPORTIONS
a = NOMINAL CRACK LENGTH.

INCLUDING FATIGUE CRACK
W = 2a
W 1 = 2.5a
S
S = 0.1a
H = 1.2a
D = 0.5a
~.)~ L. ~/',~-'" F = 2E = 1.1a
B = THICKNESS
NOTCH
ENLARGED VIEW
FIG. 1--Compact [racture toughness specimen.
T A B L E 3--Results o[ some compact [racture toughness tests ~ through the thickness

o[ 3-in.-thick 2219-T851 plate (L-T orientation).
Location Crack
in Length, ao, KQ, Test Ratio
Thickness in. ksiV~. Valid Pmax/Pt~
Surface 1.50 34.4 no 1.39
1.51 34.4 no 1.34
Avg. 34.4
T/4 1.52 34.7 no 1.28
T/2 1.52 35.2 no 1.32
1.48 32.8 no 1.41
Avg. 34.0
The following dimensions constant for all specimens: W (width) i n . : 3 . 0 0 , B

(thickness) i n . : 0 . 5 0 .
University of Washington (University of Washington) pursuant to License Agreement. No further rep
mens equal to or less than 11/~ in. in thickness were taken midway
center-to-surface each side of the center, while all of the 2-in.-thick
specimens were taken from the center of the thickness.
All specimens were fatigue cracked and all tests were carried out in
strict accordance with the ASTM E 399-72, aside from the specimen size
variations.
Results
The results of the tests are shown in Table 2. Included in the table, in
addition to the regular data from plane-strain toughness tests, are values of
Kmax, that is, the maximum value of stress intensity factor developed
during a test, calculated using the maximum load from the test in place
of the 5 percent secant load for the calculation, together with the original
crack length. Values of KQ and Kmax are plotted in Figs. 2 through 7 in a
way designed to demonstrate the points made in the following.
Discussion
The average results of the tests of various sizes of specimens of the

"standard" geometry in ASTM E 399-72, that is, with W / B = 2 , are
plotted in Fig. 2, to illustrate three points. (1) There is a definite varia-
tion in test results, with KQ increasing continuously with both crack length
and thickness, well beyond the current thickness and crack length limit.
Thus, the thickness limitations themselves are not sufficient to ensure
"meaningful" K~,. values, that is, values essentially independent of size
effects. (2) Only the last (highest) values are completely valid Kic
values, as only they meet the P .... /PQ ~ 1.1 criteria; the latter is appar-
ently a necessary criterion with the current standard geometry ( W / B = 2 )
40 i i i ! i
36
32
i 28
E
K,//
~ 24
O'J" I I I I I 1"
0 0.5 1.O 1.5 2.0 2.5 3.0
THICKNESS AND CRACK LENGTH, in.
FIG. 2--Results of plane-strain fracture toughness tests of 3-in. 2219-T851 plate

with four sizes of "'standard" compact specimen (W/B=2).
TN,CNN,SS" "k~;/-.-~,~I ,-~Ig',d,

4G s Kk
,-, ,-...... Z,--.-~ ....
"~"I'"-~/// a= 891 _ I ~TII S~IANOAIIOGEOMETRY
2st COMPACTSPECIMEN
0 0.5 1.0 1.5 2.0 2.5

SPECIMENTHICKNESS, B, In,
F I G . 3--Influence of specimen thickness in plane-strain [racture toughness tests

of 3-in. 2219-T851 plate.
44, i i i i i i
LIMITING a = 2.5
CONOITIONS
(~]~ ,(~)'
3.4 ~,_Y
4a,s c,_,Y
/ 5oys /
40 %'= O.S3Oys 0.8%s 0.74ays O.SOoy~
36
I~ 32
~ 28
24
:////
2O
1" I l i l i I
hr~
l
0.5 I.O 1.5 2.0 2.5 3.0

CRACKLENGTH,a, In.
F I G . d---Influence o] crack length independent of thickness in plane-strain fracture

toughness tests of 3-in. 2219-T851 plate (L-T).
4O i i i i I I
.~ o
36 O:2" ~P~ ' ~ , , , t / o ' ~ f ' e = l ' '
32
28 o o
o COMPACT SPECIMEN
o
24
PRESENT PRESENTMAX.
STANDARD ALTERNATE
I I I I ' J ~r~
2 4 6 8 10 12
W/B RATIO
FIG. 5mln[tuence of width-thickness ratio on results of plane-strain fracture tough-

ness test DE3-in. 2219-7"851 plate (L-T).
to avoid accepting values well below the true K~c of the material. (A
troublesome point is that there is little to suggest that even the last group
are on any plateau of size independence.) (3) Even the Kmax values for
undersized standard specimens [namely, a = B < 2 . 5 ( K o A r ~ : s ) 2] may
underestimate the true Kit value of the material.
All of the Ko data are plotted in Fig. 3 as a function of thickness and
Fig. 4 as a function of crack length. The former indicates that the various
44
i i
4O
V DO o
o
36 ~ o
oo
_'32
o
w/s:2 \ o
28 ASTM
STANDARD ~
24
THICKNESS
o~h
ol
20 a vh
?2
COMPACT SPECIMEN
1.0 1.1 1.2 1.3 i.4 1.5

RATIOPmax/Po
FIG. 6--P,~~ versus Ko 3-in. 2219-T851 plate (LT).
University of Washington (University of Washington) pursuant to License Agreement. No further reproduct
44 I I , , I
40 84
36
.~a 32
~28
24
2
20 5 4 3.4 25 Kk
0 0.2 0.4 0.6 0.8 1.0 1.2

ON~/O'yS
FIG. 7--Influence of net-section stress on results o/plane-strain [racture toughness

tests of 3-in. 2219-T851 plate (L-T).
levels of KQ are associated more with crack length than with thickness,
and this is confirmed in Fig. 4, where with only a few exceptions the KQ
values increase consistently with crack length independent of thickness
above 2.5 (KQ/~ys)2. There is an indication that the values reach a plateau
at crack lengths of 4 or perhaps 5(KQ/ays) ~, and this is also the level at
which values of Pmax/PQ for specimens with W / B - - 2 reach levels at or
below 1.1, where they meet the current ASTM E 399-72 criterion for
validity. The P .... /PQ values for the thinner specimens ( W / B > 2) still
exceed 1.1, but yield the same values of KQ.
In an effort to identify the reasons for the apparent crack length de-
pendence of KQ, the variation in test results with several other factors
was investigated. The Kq values are plotted as a function of W / B
ratio in Fig. 5; while there are trends, there is no helpful consistency.
It is noteworthy that with all thicknesses, the upper level of KQ (near
36 ksi!/in.) is achieved with sufficiently high W / B ratios. In Fig. 6, KQ
is plotted against P .... /PQ; there is no singular relationship, though it does
illustrate that for the standard geometry ( W / B = 2), the current limit of
Pma~/Po= 1.1 provides the most meaningful value of KIe. As first indi-
cated, however, the same KQ value is achieved with other W / B ratios at
much higher Pmax/Pq ratios.
Harrigan and Somer ~ suggested that there was a relationship between
Minutes of 29 Sept. 1971 meeting of ASTM Subcommittee E24.01 on Fracture
Mechanics Test Methods.
the normal net-section stress and KQ value, and that it was necessary to
keep that stress below 0.8 ~Ys. This criteria was tested in Fig. 7, and al-
though the limits are not as suggested, the apparent interrelationship is
confirmed. The KQ value increased with decrease in ratio of net-section
stress (~NpQ/crys), seemingly reaching a plateau value when the ratio is
less than 0.7. Regrettably, the range of sizes studied does n o t tax this
plateau, as no ratios below 0.5 were obtained; so the limit of the plateau
cannot be established with certainty. Nevertheless, it appears that KQ
values are relatively consistent at a value near the crack length plateau
(Fig. 4) if the net-section stress is kept below about 0.7 of the yield
strength.
While this type of criteria may seem offhandedly novel, it is simply a
restatement in other terms of the current type of limit of specimen size
based upon plastic zone size (according to ASTM E 3 9 9 - 7 2 ) . The present
thickness (B) and crack length (a) criteria state that a and B must
equal or exceed approximately 50 times the radius of the plane-strain
plastic zone size, which may be approximated
1 (KI~'~2 1 (Krc'~ 2
Thus, 50 times the plastic zone size is about 2.5(Ki~/~ys)L Analysis of

the maximum nominal P/A + M c / I stresses (crm,Q) on a compact tension
specimen when a~ W = 0.5 leads to
~rNPo=l.478Kie ora=2.18{ KI~ ~2

\/ a \ crNpQ]
When a=2.5(Kie/gys) z, crNpo=0.95 ~rys. Thus, when we limit plastic
zone size we also limit maximum nominal net-section yield stress. Some
other convenient conversions are as follows:
~NPQ= 0.80 ~rs a=3.4( K~c~2

\ ~YS /
~pQ = 0.66 gvs a=5.0( K'e~ 2

\ O'YS/
Implications in Size Requirements

The results of the current study suggest that it may be important, even
within the general plane-strain region, to keep the net-section stresses
below 0.7 ~Ys to obtain size independent results. This can be accomplished
by controlling crack length as before, but to a higher limit, say 5(Ki~/~ys) 2,
so long as thickness is maintained at 2.5(K~Ja~s) ~. There is some indi-
cation from Fig. 3 that the thickness limit might be dropped below
! I I I
60 cltm~
l LIENGIXI it
/ ' = 3 v,,
50 L ' : 2~0~',,
1'=2 ~,'~..
] 40L.
_"~1 -~" ,~-
..... ,,
-7-
GEOMETRY
2O W/B = 2
10
0 I I I I
0.5 1.0 1.5 2.0 2.5
SPECIMEN THICKNESS, B, in.
60 I I I I r I
50
40
-~ 3 0
ASTMSTANDARD
G~TIty
W/B -~ W/a = 2 ~ Dr IL
20 0 89
D 1
~ 189
10 v 2
I I I I I I
0.5 1.0 1.5 2.0 2.5 3.0
CRACK LENGTH, a, in.
60
I I I I I I
50
40
o 89
~ 20 - 0 i -
A 189
V2
10
I I I I I I
2 4 6 8 I0 12
W/B RATIO
FIG. 8--(a) Influence of specimen thickness on K~a~ for compact tension tests of
3-in. 2219.-I"851 plate. ( b ) Influence of crack length on K ~ for compact specimens
of 3-in. 2219-T851 plate. ( c ) Influence of W / B ratio on K~a~ for compact specimens
of 3-in. 2219-I'851.
84
University of Washington (University of Washington) pursuant to License Agreement. No further reprod
2.5(Kic/ays) 2 if the crack length is ~5(K1c/gx-s) 2, but this should be

checked further to establish what the limits would be.
The Pm,x/PQ=1.1 criteria does not hold down net-section stresses
except with W/B=2. Since the current study suggests that W/B=4
could be accepted as standard, if a~5(KiJ,rys) ~, it appears that the
Pnlax/PQ limitation could be dropped.
Value of K .....
If K .... is plotted as a function of specimen thickness, crack length, or

W/B ratio as in Figs. 8a, b, and c, respectively, several trends of interest
are evident. From Figs. 8a and b, for a given thickness, Km,x tends to
increase with increase in crack length or W/B ratio, but as thickness
increases the effect decreases; when thickness is greater than 5(Ki/~u 2,
K~ax is almost independent of crack length, and is relatively close to K~c,
This suggests that if we could keep both thickness and crack length greater
than 5(KiJ~Ys) 2, we could eliminate the need for curve plotting and use
maximum load to obtain a good estimate of KIc. From Fig. 8c, however,
K~,~ is seen to increase consistently with W/B ratio, with little difference
dependent upon specimen thickness; there is a slight tendency for greater
thicknesses to give greater Knla.~ values for a given W/B ratio, but the
differences are small. This seems to contradict the first suggestion, but
simply says that if we are to use K,n,~ as an indication of KI~, the W/B
ratio must be kept small, probably 2.
Summary
A new study of the effect of specimen size on the results of plane-
strain fracture toughness tests of a relatively tough aluminum alloy,
2219-T851, suggests that an increase in specimen size requirements will
assure size-independent test results. Specifically, it appears that the crack
length limit should be increased to 5 ( K i J ~ - s ) 2, which has the effect of
keeping the maximum nominal net-section stress below two thirds of the
yield strength. The current limit on thickness could be maintained at
2.5(K~c/~ys) ~, although there is some evidence that it might be relaxed
further. Additional work is needed to check the generality of these sug-
gestions for other materials, as well as the possibility that if W/B= 2 and
a=B-~5 (K,~/~ys)2, Km~ may be useful as an engineering estimate of Kx~.
P. C. Paris, ~ R. J. Bucci, ~ and L. L. Loushin
Dynamic Compact Tension Testing

for Fracture Toughness
REFERENCE: Paris, P. C., Bucci, R. J., and Loushin, L. L., "Dynamic

Compact Tension Testing for Fracture Toughness," Fracture Toughness and
Slow-Stable Cracking, ASTM STP 559, American Society for Testing and
Materials, 1974, pp. 86-98.
ABSTRACT: Within this paper it is demonstrated that dynamic fracture
toughness evaluation may be made by adherence to the ASTM Test for
Plane-Strain Fracture Toughness of Metallic Materials (E 399-72) proce-
dure with the single exception of test speed. The test procedure, specimen,
fixturing, and instrumentation are otherwise identical to ASTM E 399-72.
This makes possible direct comparisons of static Kxc values with the results
of these dynamic tests.
The additional equipment required for these tests, over the normal servo-
controlled hydraulic test machine typically used for K~c testing, is only the
provision of a high flow and high response servo-valve with hydraulic ac-
cumulators and the capability of dynamic data storage (digital) for con-
venient playback into the usual X-Y recording unit.
Employing the test technique developed, dynamic fracture toughness
evaluation was demonstrated on two steel alloys (302 B and A516-70)
tested at room and --50 ~ F temperatures.
KEY WORDS: fracture properties, dynamic tests, mechanical properties,
crack propagation, evaluation
Methods for the evaluation of static fracture toughness have been well
enough developed that an American Society for Testing and Materials
standard practice currently exists for the establishment of plane-strain
fracture toughness, K~e, ASTM Test for Plane-Strain Fracture Toughness
1Chairman and chief executive officer, Del Research Corporation, Hellertown, Pa.
18055, and Visiting Professor of Engineering, Brown University, Providence, R. I.
02912.
Formerly, Del Research Corporation; currently, senior engineer, Engineering De-
sign Division, Alcoa Laboratories, Alcoa Center, Pa. 15069.
s Formerly, Esso Research and Engineering Company, Florham Park, N. J. 07932.
86
Copyright*1974by
Downloaded/printed by ASTM lntcrnational www.astm.org
PARIS ET AL ON DYNAMIC COMPACT TENSION TESTING 87
of Metallic Materials (E 399-72). Further, it would seem to be advan-

tageous to have a common ground for comparison of static and dynamic
fracture toughness evaluation. To this end similar test procedures and
philosophies, which differ only in the rate of test, would be highly desirable.
Parallel investigation which demonstrates the suitability of dynamic fracture
toughness evaluation via compact tension testing was provided by Shab-
bits[l] 4 and Bush[2]. Hence, it is the intention of this work to further
develop the test method as well as demonstrate the feasibility for compact
tension fracture toughness evaluation at high loading rates by strict adher-
ence to ASTM E 399-72 in every way but the exception of testing rate.
The advantage of such methodology, if successful, would be rather straight
forward extension of the ASTM E 399-72 procedure to cover dynamic
testing rates.
Previous dynamic fracture toughness testing has involved impact tests
(sometimes instrumented) and crack arrest tests, for example, Charpy or
the drop weight tests. In both cases, an essential difficulty in test precision
and interpretation has centered around evaluation of load at the instant
of measurement (the instant of flaw instability or arrest). These types of
tests have been developed so that agreement in results is generally con-
ceded. It has been generally recognized for structural steels[I-4] that
fracture toughness at constant temperature will decrease with increasing
strain rate, ~ (or crack tip stress rate, ~ or/~) to some minimum level. For
steels, a typical relationship between Kxc and the reciprocal of time to the
critical measurement point, 1/tcr, is schematically depicted in Fig. 1. The
range in strain (or loading) rates associated with static loading are gen-
erally four or more orders of magnitude slower than those rates associated
with impact loading. Since the minimum band of crack toughness data
extends over a relatively broad range of strain (or loading) rates, test data
obtained at rates slightly less (say one order of magnitude less) than impact
values (Point B, Fig. 1) might be expected to exhibit much the same tough-
ness values as those associated with impact rates (Point A, Fig. 1).
Therefore, consistency between the more rigorous static test instru-
mentation methods (ASTM E 399-72) and the dynamic tests is sug-
gested. By running tests fully instrumented to obtain load-displacement
(crack opening) records as rapidly as possible, loading times ranging from
10 to 50 ms to maximum load were found possible using the usual static
fracture toughness instrumentation techniques (ASTM E 399-72) on stan-
dard compact tension type specimens up to 4 in. thick. Similar success
has also been reported in Refs 1 and 2. This loading rate, in contrast to
paper.
TEMPERATURE - CONSTANT
K!c ,ANGE OF RATES

ASSOCIAT O W,TH
..... STATIC LOADING 0 " RATES
ASSOCIATED WITH
IMPACT LOADING
i i i i i i I I I
I0 "~ I IO g 10 4 lO G
I/tr r - SEC-i
OR
FIG. l--Schematic representation of fracture toughness variation with strain rate

(stress rate).
the 1 to 5 ms loading rates encountered in impact testing, is known to

give fracture toughness behavior approaching minimum dynamic values
(compare data from Refs 2 - 6 ) established from impact or arrest tests.
The advantage of having a full load-displacement record for detailed
comparison to static (ASTM E 399-72) test results is regarded as out-
weighing small differences with producing minimum dynamic toughness
conditions. In addition, at loading duration which approaches minimum
dynamic toughness conditions (that is, approximately 1 ms) the com-
plexities introduced by test method susceptibility to inadequate test sys-
tem response, mechanical vibration, and shock effects can become sig-
nificant. Hence, the prior approach which utilizes only slightly slower
testing speeds forms a link between the more rigorous static test practice
and previous dynamic testing. Therefore, a recognized potential exists for
the attainment of dynamic Kie (in the same sense as recommended by
ASTM E 399-72), and also for the establishment of dynamic crack
growth resistance (R) curves, and dynamic J integral interpretation.
In the discussion to follow, a test procedure for the determination of
dynamic fracture toughness is outlined and demonstrated for two medium
strength steels (A302-B and A516-70).
Alloys
A 3 0 2 - B Steel (Normalized)
The material was supplied in a nominal 2-in. section thickness by U. S.
Steel. The material tensile properties were reported as being similar to
those listed in Ref. 7. For convenience, the room temperature tensile

properties are listed in the following[6].
~r~ ( 0 . 2 % ) ~t~ % Elongation (1 in.) % Reduction
56.0 ksi 88.0 ksi 26 67
.4516-70 Steel
The A 5 1 6 - 7 0 steel was heat treated at 1 1 7 5 _ 25 ~ F, 1 h/in. and cool
at < 5 0 0 ~ F. The material was supplied in a 2-in. nominal section thick-
ness. The material tensile properties were found to be:
~r~ ~t~ % Elongation (2 in.)
40.5 ksi 70.5 ksi 32
Experimentation
Specimen Configuration and Preparation
Standard compact tension specimens ( H / W = 0 . 6 ) of the type shown
in Fig. 2 were used for dynamic fracture toughness evaluation. All speci-
mens were notched so as to be tested in the L T orientation. Actual test
specimen dimensions are given on the appropriate accompanying load dis-
placement test record, Figs. 3, 5, 7, and 9.
All specimens were fatigue precracked at stress intensity levels less than
25 ksi~/in. In this manner, the initial machined notch was extended to
fatigue crack severity as recommended by ASTM E 399-72 to a final
length of 0.45 <_a/W<0.55.
P
B
f
/
I.
~.---~
Q
J:
-I
J J= W ='
mW
NOMINAL DIMENSIONS B = 2", W= 5"
FIG. 2--Compact tension configuration.
FIG. 3--Dynamic load versus displacement test record jor A302-B steel at --50~
Each alloy was tested to failure at room temperature and at - 5 0 ~ F.

The cold temperature was provided by initially soaking the specimen in
dry ice ( - 9 0 ~ F) for approximately 1 h. The specimen was then removed
from the initial soak, mounted in the test frame and allowed to gradually
warm up to the desired test temperature ( - 5 0 ~ F ) , whereupon loading
was triggered. Insulation was provided to the specimen surfaces to further
decrease the thermal gradient through the thickness. As best as could be
determined by strategic placement of thermocouples during the warm up
FIG. 4 - - D y n a m i c load versus time and displacement versus time records for
A302-B at --50~
phase, the internal and surface temperature never varied from one
another by more than 10 ~ F.
Equipment and Instrumentation

The loading equipment used was a closed loop, hydraulically actuated,
servo-controlled mechanical test system with 100 kip load capacity. In
order to achieve the rapid displacement rates required for dynamic test-
FIG. 5--Dynamic load versus displacement test record for A302-B at room tem-
perature.
ing, some equipment modification was necessary. Essentially, this modi-

fication employed two high response 15 gpm servo-valves and accom-
panying accumulators with a storage volume greater than the ram
displacement volume.
Test records were obtained for each specimen by plotting load versus
crack opening displacement curves to the point of failure. Loads were
measured by a standard strain gage load cell while crack displacements
FIG. 6 - - D y n a m i c load versus time and displacement versus time record for A302-B
at room temperature.
were measured using a clip-on (strain gage) device as per standard

ASTM K~c testing. The natural frequencies of both the load cell and
clip-on gage were found to be more than one order of magnitude greater
than those required for the dynamic load application and, hence, were
not considered to contribute any detrimental influence to the load-displace-
ment resolution. For dynamic tests at high loading rates, load and dis-
placement signal inputs were stored in a transient recorder for subsequent
playback at convenient X-Y recorder pen speeds. A digital transient
recorder as well as a digital computer with test machine interface have
been used with equal success for this purpose.
Test Records
Figures 3, 5, 7, and 9 show dynamic load versus displacement test
records generated, employing the dynamic test procedure previously
described. For each of the aforementioned dynamic load displacement
records, the related load versus time and displacement versus time record
FIG. 7--Dynamic load versus displacement test record for A516-70 at --50~
for the same test specimen is given in subsequent figures (namely, Figs.
4, 6, 8, and 10). From these curves, one can quickly assess the order of
magnitude loading (or displacement) rates encountered during the dynamic
tests. For the more brittle fractures (for example, Figs. 4 and 8), the "linear
load rise time" to failure was of the order of 10 ms. Whereas for some
of the more ductile failures (for example, Figs. 6 and 10), the "load
rise time" to maximum load or instability is of the order of 40 ms or less.
Past experience employing this technique for similar medium-strength
FIG. 8 - - D y n a m i c load versus time and displacement versus time record for A 5 1 6 -
70 at --50~
steels on compact tension specimens up to 4 in. thick has shown that

dynamic loading to failure can generally be accomplished within a range
of 10 to 40 ms load rise time (that is, within an approximate one order of
magnitude deviation from rates associated with the usual impact test
conditions).
FIG. 9 - - D y n a m i c load versus displacement test record tor A 5 1 6 - 7 0 at room tem-

perature.
Analysis of Test Records

A summary of test results for each material test condition is given in
Table 1.
The KQ value for the A302-B test at - 5 0 ~ F satisfies all the require-
ments, with exception of loading rate, for a valid Kxe determination by
ASTM E 399-72, and hence represents an unambiguous evaluation of
dynamic Kxe.
The test record for the A 5 1 6 - 7 0 material tested at - 5 0 ~ F, Fig. 7,
gives the appearance that the test result is very close to being valid plane
strain. An estimation of the yield strength[8] for - 5 0 ~ F and t - - 1 0 ms
shows 2.5 (KQ/gys)2=2.26 in. That is, the ASTM size requirement
exceeds the specimen thickness by a quarter inch.
When general yielding occurs prior to maximum load, such as for the
room temperature dynamic tests of Figs. 5 and 9, the determination of an
effective fracture toughness or Kr~, is not a simple matter. Moreover, no
standard procedure for static or dynamic K~ measurement or comparison
currently exists. However, since this dynamic test procedure is similar to
conventional static test procedures, one then has available for use in
FIG. lO----Dynamic load versus time and displacement versus time record for
A 5 1 6 - 7 0 at room temperature.
TABLE 1 - - S u m m a r y o[ test results.
2.5
Tempera- Kq, K~Ax, ( KQ ~
~'ys, a IYys,b ture, ksi__ ksi__ PMAX ~ qys b /
Material ksi ksi ~ Vin. Vin. pQ m.
A302-B
( B = 2 in.) 56 79.8 --50 49.8 49.8 1.0 0.97
56 65.5 75 83.5 136 1.63 4.06
A516-70
( B = 2 in.) 40.5 64.3 --50 61.1 61.1 1.0 2.26
40.5 50.0 75 74.6 109 1.46 5.57
r is the uniaxial static yield strength at room temperature.

b r is the estimate of uniaxial yield strength for the test condiitons as given by
Madison and Irwin [8]. The Madison Irwin equation for yield strength estimates as a
function of temperature and test speed (r t) ) is as follows:
174 000 (ksi) 27.4 (ksi)
~rys (T, t) =~rs (To, to) + ( T + 4 5 9 ) l o g l o ( 2 X 10~~
where units of ~ys in ksi, T is testing temperature in ~ and t is seconds of time
from zero to maximum load.
dynamic Kc interpretation current "state of the art" practice employed in

static Kc interpretation. Therefore, summarized in Table 1 are the fol-
lowing parameters as determined from the dynamic test records:
KQ established in accordance with A S T M E 399-72, and
KMAx a calculation of K based on the maximum load and the fatigue
precrack size, that is, KMAx=PMAx/PQ (KQ).
Having full load displacement records permits other possibilities for
the interpretation of results within available methodology (for example,
ry corrected values for K~, development of R curves or J integral or both,
equivalent energy, etc.).
Summary
The feasibility of dynamic plane-strain fracture toughness evaluation by
strict adherence to the A S T M E 399-72 procedure, with the single
exception of test speed, has been demonstrated. The procedure outlined
in this paper makes possible direct comparison of static and dynamic K~
without ambiguity.
A cknowledgm ents
The Del Research authors gratefully acknowledge the sponsorship of
Esso Research and Engineering Company, within which J. E. Guthrie
and V. Anderson assisted significantly with their stimulation and interest
in the overall program development. Moreover, a special acknowledg-
ment is due to the laboratory staff of Del Research whose assistance was
invaluable in the test and program development.
References
[1] Shabbits, W. O., "Dynamic Fracture Toughness Properties of Heavy Section
A533 Grade B, Class 1 Steel Plate," Heavy Section Steel Technology Program
Technical Report No. 13.
[2] Bush, A. J. in Impact Testing of Metals, ASTM STP 466, American Society for
Testing and Materials, 1970, p. 259.
[3] Shoemaker, A. K. and Rolfe, S. T., "Static and Dynamic Low-Temperature KIo
Behavior of Steels," Journal of Basic Engineering, Sept. 1969.
[4] Krafft, J. M. and Irwin, G. R., "Crack Velocity Considerations," Symposium on
Fracture Toughness Testing and its Applications, ASTM STP 381, American So-
ciety for Testing and Materials, 1965.
[5] Eftis, J. E. and Krafft, J. M., "A Comparison of the Initiation with Rapid Crack
Propagation of a Crack in a Mild Steel Plate," Journal of Basic Engineering,
Transactions, American Society of Mechanical Engineers, Series D, Vol. 87, 1965.
[6] Shoemaker, A. K. and Rolfe, S. T., "The Static and Dynamic Low-Temperature
Crack Toughness Performance of Seven Structural Steels," Engineering Fracture
Mechanics, Vol. 2, No. 4, 1970.
[7] Barsom, J. M. and Rolfe, S. T., "Correlations Between Kxc and Charpy V-Notch
Test Results in the Transition Temperature Range," Impact Testing of Metals,
ASTM STP 466, American Society for Testing and Materials, 1970.
18] Madison, R. B. and Irwin, G. R., Journal o[ Structural Division, Proceedings,
American Society of Chemical Engineers, Sept. 1969, p. 2235.
A . M . S u l l i v a n 1 a n d J. S t o o p 1
Further Aspects of Fracture Resistance

Measurement on Thin Sheet Material:
Yield Stress and Crack Length
REFERENCE: Sullivan, A. M. and Stoop, J., "Further Aspects of Fracture

Resistance Measurement on Thin Sheet Material: Yield Stress and Crack
Length," Fracture Toughness and Slow-Stable Cracking, ASTM STP 559,
ABSTRACT: The fracture resistance value, Kc, has been determined for a
number of alloys of aluminum, titanium, and steel and found to depend in-
versely upon the material yield stress. During this investigation various ini-
tial crack lengths, 2ao, were employed.
Analysis of the data has disclosed that the amount of crack extension,
that is, final crack length, 2ao, appears to be influenced by the initial crack
F 1 [K~ \ 2 l
length. Plotted on a log-log scale of (ao/a,) versus #--LB~r .J' statisti-
cal analysis justifies a relationship conforming to the equation
In 10 (ao/a~) = A - - B In 10
which can be transformed to ~ = ~ -

The development of these relationships can be of real assistance in the de-
sign of a standard initial screening test for Kc.
KEY WORDS: mechanical properties, fracture properties, fracture strength,
crack propagation
C o n t i n u e d e s c a l a t i o n of the strength to weight ratio in m o d e r n struc-

tures using s h e e t m a t e r i a l s increases the difficulties of fail-safe design
since i n c r e a s e d strength is always a c c o m p a n i e d b y l o w e r e d fracture
resistance. This p r o p e r t y is c o n s i d e r e d as the ability of a m a t e r i a l to with-
s t a n d the deleterious effect of cracks. O n e m e a s u r e m e n t of f r a c t u r e
1 Consultant and metallurgist, respectively, Fracture Machanics, Metallurgy Divi-

sion, Naval Research Laboratory, Washington, D. C. 20375.
99
Copyright9 1974 by
resistance is the p l a n e stress fracture m e c h a n i c s p a r a m e t e r , Kc. Deter-

m i n a t i o n of this value is h a n d i c a p p e d by the fact that it is influenced by
geometrical dependencies: specimen width, W, crack length to width
ratio, 2 a / W , a n d material thickness, B. These dependencies preclude the
d e v e l o p m e n t of a single s t a n d a r d test specimen for its absolute deter-
mination.
However, since guidelines have now b e e n established to assist in the
estimation of allowable specimen width a n d crack length to width ratio,
it is appropriate to consider two other aspects of Kc d e t e r m i n a t i o n : the
degradation of fracture resistance with increasing yield strength, ~+s, a n d
TABLE 1--Fracture resistance data for aluminum alloys.
B, W, 2ao, 2ac, ~ys, Kc,

Alloy in. in. in. in. ksi ksi x/in. /~ b
7178-T6 0.040 12 2.130 2.40 79 49 9.7
0.063 12 2.124 2.40 78 50 6.4
0.091 12 2.138 2.40 78 44 3.6
0.127 12 2.140 2.54 77 45 2.7
7079-T6 0.037 12 2.138 3.66 68 62 21.9
0.063 12 2.138 3.42 71 71 15.7
0.101 12 2.128 3.90 62 98 24.9
0.140 12 2.144 3.12 72 68 6.5
0.250 12 2.086 3.34 66 81 6.0
2014-T6 0.040 12 2.156 3.18 62 75 37.4
0.063 12 2.120 3.06 61 72 22.0
0.091 12 2.129 3.10 57 75 19.1
0.125 12 2.100 3.80 60 92 19.1
0.250 12 2.118 3.20 59 66 4.9
2219-T87 0.032 12 2.150 3.81 52 86 84.4
0.062 12 2.155 3.44 54 77 32.9
0.091 12 2.137 3.76 51 92 35.6
0.125 12 2.139 4.59 52 88 22.8
0.250 12 2.120 3.60 54 76 8.0
7178-T6 a 0.063 12 2.132 2.576 78 36 3.5
0.091 12 2.190 2.784 76 45 3.9
0.125 12 2.145 3.28 78 48 2.9
0.25 12 2.16 2.16 78 36 0.84
0.30 12 2.135 2.88 79 36 0.71
7075-T6" 0.063 12 2.132 2.64 72 70 14.8
0.092 12 2.150 3.136 76 77 11.5
0.124 12 2.132 2.96 74 61 5.4
0.25 12 2.168 2.168 76 42 1.2
0.30 12 2.132 2.85 74 46 1.3
Rolled down from 0.30-in. sheet.

+~ = ~ (K~/~) ~.
CopyrightbyASTMInt'l(allrightsreserved);MonDec 713:10:55EST2015
Downloaded/printedby
UniversityofWashington(UniversityofWashington)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
SULLIVAN AND STOOP ON THIN SHEET MATERIAL 101
TABLE 2--Fracture resistance data [or titanium alloys.
B, W, 2ao, 2ac, o-y~, Kc,

Alloy in. in. in. in. ksi ksi x/in. ~~
6A1-4V 0.032 12 2.138 2.82 151 54 4.00
0.062 12 2.132 2.70 158 64 2.86
0.090 12 2.132 2.40 146 77 2.66
4A1-3Mo-IV 0.042 12 2.144 2.144 162 50 2.30
0.058 12 2.234 2.234 153 54 2.30
0.087 12 2.290 2.290 152 62 1.90
0.124 12 2.132 2.132 160 35 0.38
16V-2.5A1 0.041 12 2.128 2.52 182 52 2.00
0.059 12 2.137 2.137 176 46 1.16
0.118 12 2.138 2.64 182 44 0.51
13V-I1A 0.040 12 2.152 2.152 198 34 0.74
3A1 0.063 12 2.113 2.113 207 38 0.56
0.091 12 2,116 2.116 201 30 0.24
~= -~ ( Kr ) 2.
the a m o u n t of crack growth preceding failure. Some measurement of the

former is necessary to utilize estimates of minimum width which are
related to the Ke/try s ratio; a method for estimating the latter would
simplify testing procedures by eliminating the necessity for crack exten-
sion measurement.
Experimental Parameters
Materials
Fracture resistance tests contributing to this study were largely on
aluminum, titanium, and steel alloy sheet specimens of varied thickness.
Relevant data and fracture resistance values are presented in Tables
1, 2, 3, 4. In all specimens the fracture path was parallel to the rolling
direction of the sheet ( T L ) .
T e s t Procedure
The center-cracked tension ( C C T ) specimen was utilized[/].'-' The
central slit is produced by an electric discharge method (Elox) to give a
0.063-in. wide slit with a 0.003 to 0.006-in, tip radius. This is extended
with a fine electrode to sharpen the tip radius to 0.001 in. W h e n the
specimen is loaded in, tension, load, and central crack opening displace-
ment, C O D , are simultaneously graphed by an X - Y recorder until failure
paper.
TABLE 3--Fracture resistance data [or steels.
B, W, 2ao, 2a~, ~ys, Kc,

Steel in. in. in. in. ksi ksi Vin. ~'"
4130 0.032 12 2.18 4.00 a 168 151 21.80
0.032 12 4.51 6.84 168 146
0.050 12 2.21 4.56 172 172 20.10
0.050 12 4.17 6.72 172 174
0.063 12 2.26 3.78 170 172 18.20
0.063 12 4.21 6.54 170 191
0.087 12 2.20 3.26 183 124 4.85
0.087 12 4.18 6.07 183 144 4.85
0.125 12 2.21 3.54 176 148 6.80
0.125 12 4.18 5.86 176 158
4130 0.032 12 2.13 4.98 185 154 22.6
0.032 12 4.15 6.66 185 161
0.051 12 2.15 3.16 185 146 13.60
0.051 12 4.10 6.60 185 158
0.063 12 2.16 3.72 178 129 9,40
0.063 12 4.25 6.12 178 146
0.087 12 2.10 3.26 200 123 4.50
0.087 12 4.08 6.46 200 127
0.125 12 2.10 3.72 191 155 4.60
0.125 12 4.10 4.92 191 138
0.250 12 2.20 4.22 185 163 2.30
0.250 12 4.16 5.40 185 121
D6A 0.098 12 2.10 2.10 228 54 0.66
0.098 12 4.09 4.09 228 62
0.190 12 2.12 2.12 220 48 0.23
0.190 12 4.12 4.12 220 45
0.25 12 2.16 2.16 230 68 0.45
0.25 12 4.14 4.14 230 86
RSM-250 0.063 12 2.162 2.68 244 180 10.0
0.063 12 4.154 5.94 244 207
0.090 12 2.226 2.94 246 204 7.6
0.090 12 4.120 4.75 246 204
0.140 12 2.126 3.04 248 194 4.2
0.140 12 4.154 5.04 248 186
Used average values of ao/ac at each thickness for statistics.

b Used average value of Ko at each thickness for/~ calculation.
1
/~=~- (K~/o~,,) 2.
occurs. C r a c k o p e n i n g is m e a s u r e d by a strain gage i n s t r u m e n t e d dis-

p l a c e m e n t p r o b e p o s i t i o n e d in a circular hole in the center of t h e initial
slit. T h e C O D m e a s u r e m e n t is referred to a n o r m a l i z e d e x p e r i m e n t a l
c a l i b r at i o n c u r v e which relates the a m o u n t of c r a c k o p e n i n g to the
i ns t an t an eo u s c r a c k length of the specimen. This p r o c e d u r e has b e e n dis-
cussed previously in d e t a i l [ / ] . U s in g this m e t h o d of c r a c k length deter-
University of Washington (University of Washington) pursuant to License Agreement. No further re
TABLE 4--Fracture resistance data for Aluminum 7475-T61.
Percent
B, W, 2ao, 2a~, ays, K,., 2ac ~ Ke a Differ-
in. in. in. in. ksi ksi Vin. Ob in. ksi Vin. eIlCe
0.063 12 2.13 3.96 68 123 ~ 52.8 3.62 116 --5.8

0.063 12 3.13 5.04 68 124 ~ 52.8 5.32 128 +4.2
0.107 12 2.12 3.07 72 103 19.1 3.26 106 +3.6
0.107 12 3.10 4.14 72 105 19.9 4.78 117 +11.0
0.190 12 2.16 2.94 72 102 9.0 3.06 105 +2.7
0.190 12 3.09 4.14 72 87 7.6 4.30 90 +2.8
0.25 12 2.13 3.26 72 96 7.0 2.94 91 --5.7
0.25 12 3.07 4.24 72 102 8.0 4.28 104 +1.0
0.063 20 3.61 5.56 68 130 60.6 5.20 120 --8.0
0.063 20 5.10 6.60 68 111 42.4 8.48 132 +19.6
0.109 20 3.62 4.70 72 107 20.4 5.60 118 +10.6
0.109 20 5.10 6.90 72 108 20.8 7.88 118 +9.6
0.190 20 3.60 4.20 72 80 6.4 4.92 88 + 10.0
0.190 20 5.13 5.60 72 76 5.8 6.96 88 + 15.4
0.25 20 3.62 4.50 72 88 5.9 4.92 93 + 5.2
0.25 20 5.11 6.64 72 99 7.4 7.10 104 +5.0
~r net ~ ~ .
1
fl= ~ (Kc/a"s) 2"
~ Values calculated from Eq I.
a Values calculated from calculated value of 2ac.
mination an "effective" crack length is measured so that no plastic zone

correction factor is needed.
Data Analysis.
Fracture resistance, K~, is calculated according to the equation[2]
Kc=~o Vac f ( 2 a / W ) (1)

where
f ( 2 a / W ) = 1.77( 1 - . 1 ( 2 a / W ) + ( 2 a / W ) ~- (2)
The values of gross stress, ,re,, and 2a determined in the preceding section
are utilized.
Regression lines, correlation coefficients, and t and F-statistic calcula-
tions were computed from conventional statistical definitions and equa-
tions[3,4].
Degradation of Kc with Increased Yield Strength

The fact that materials show a degradation in resistance to fracture,
K~, with increasing yield stress has long been recognized for thick plate
material. As can be seen in Fig. 1, the inverse relationship just discussed

is also obtained for thin sheet alloys. Upper and lower limit lines define
the highest and lowest Kc values measured with respect to yield strength
for the alloy systems of aluminum, titanium, and steel. (Thicknesses
from 0.032 to 0.25 in. are included.) The ratio lines KJcr~.~ are accom-
panied by estimates of the minimum width requirement for the CCT
specimen. Normalizing the data shown in Fig. 1 by the yield strength and
elastic modulus gives the curve shown in Fig. 2. A similar curve could be
plotted using yield stress and alloy density as normalizing factors.
These diagrams indicate the range of values observed. From them it is
possible, knowing yield stress and modulus, to estimate an upper limit for
the K~ of an unknown material so that an appropriate test specimen width
can be selected. It can be noted that a "standard" width of 12 in. is
sufficient for values of K~/(r~.s < 1.5.
Relationship Between Initial and Final Crack Length

When a sheet specimen containing a notch is loaded in tension, the
stress must reach a certain value before a crack will form at the notch
tips. Once such a crack has initiated, it will grow under a rising load
until the instability value is reached. Since computation of Ke (the frac-
ture resistance value) requires knowledge of both stress and crack length
320
W (IN.) (20)
K/o'y$ 2.0
240
2OO
160
v z 0.33
;",20
80 ~ ~ 0.20
40
16-2.5J~'-~--v~13-II-3
1 ~ I
~ I t I ~ I l 1 I 1 I I ~ I l
O0 40 80 120 160 200 240 280 320 360 400 440
YIELDSTRESSO-ys(KSI)
FIG. l--Fracture resistance, K~ versus yield stress, ~ . Straight lines denote K/a~,
ratio. An estimate ol minimum specimen width, W accompanies each ratio value.
FIG. 2 - - F r a c t u r e resistance versus yield stress n o r m a l i z e d as Kc/au8 versus ~r~s/E.
at failure, crack length must be measured during the course of the test.
A method for estimating the amount of crack extension from the initial
notch length would be of practical value since it would eliminate the
necessity for crack extension measurement. The question to be resolved is
whether or not the final crack length, 2ac, is a function of the initial crack
length, 2ao. Opinions differ as to whether no influence should exist or
whether, in fact, an influence does exist[5-10]. The results of this investi-
gation suggest a form for a relationship between 2ac and 2ao.
When final crack length is plotted against initial crack length, a straight
line relationship can be observed. This is illustrated in Figs. 3a and b for
Aluminum 7075-T6; other materials indicate a similar trend. Attempts
to correlate this slope value with other parameters showed a relationship
with the dimensionless value of
KoV7
To explore this, data taken from K~ tests on thickness series from various
sheet alloys were analyzed.
If the data are plotted on linear coordinate paper as in Fig. 4, the
scatter diagram suggests the possibility of a linear-logarithmic relation-
ship. Such a plot is seen in Fig. 5 together with the calculated regression
curve and confidence limits.
F I G . 3 - - H a l f critical crack length, ac versus initial crack length, ao.
SULLIVAN AND STOOP ON THIN SHEET MATERIAL ] 07
1.0
o "8
~ o~176
9 ~
o o
0.8
~0 ~ ~ . ~ . , . ~ 0
0
0.6 0 "~-0
0
o
0.4
o A L U M I N U M ALLOYS
[] T I T A N I U M ALLOYS
0.2 m
A STEELS
0 , I , I i I , I
0 I0 20 30 40
t 2
k O"ys /
FIG. 4---Ratio oJ initial to final crack length, ao/a~, versus [3; [3_ ~ \I(KoY
-~) .
2.0
1.0
~ ~ e' ~ ~e,~ ~ o ~ .
o ALUMINUM ALLOYS
TITANIUM ALLOYS
~, STEELS
In I O ( ~ - ) = 2.4134-0.1026 In IO
Oo
E = 0 . 8 8 2 /8-0.i026
0.I I I I I [ I I I i
I0 O0
I Kc 2
"=~ (o-,s)
FIG. 5--Log ao/a~ versus log 3. Regression curve flanked by 95 percent confidence
limit lines; closed symbols Aluminum 7475-T61.
For purposes of arithmetical expedience, the data were analyzed in

the form
In Ioa~ In 10/3 (3)

ae
The regression curve calculated for 59 datum points is
In 10a~ =2.4367-0.1142 In 10/3 (4)

ac
with a standard deviation for y = 0.21 and a correlation coefficient r = 0.75.
A t-statistic computed from the correlation coefficient equals -8.56,
which is outside the 99 percent confidence interval of t = • this
indicates that a relationship does exist between at~at and /3.
Since the data comprised values obtained from testing aluminum, titan-
ium, and steel alloys, these data were also analyzed separately and inter-
compared. An analysis of variance ratio (F-statistic) indicates that:
l. One regression line may be used for all three alloy systems
F = 3 . 1 6 < F 99% =3.65
2. The slopes of the three systems are equal
F = l . l 1 8 < F 99% =4.98
3. Regression of the means is linear
F = 6 . 5 3 < F 99% =7.08
4. Slope within groups are equal to the slope among groups
F = 4 . 4 4 < F 99% =7.08
Finally, another set of aluminum alloy data was added bringing the
number of datum points to 75. The regression curve computed is
In 10 (a~ =2.4134-0.1026 in 10/3 (5)

ae
with a standard deviation for y = 0.20 and a correlation coefficient r = 0.71.
This equation was used to compute ao/a, /or the last set of aluminum
data (Alloy 7475-T61) and K~ values from the estimates of 2a~. These
estimates are included in Table 4. The fact that these are slightly high is
rationalized by the fact that certain values of ao/a~=1.00 were included
in the analysis. When no crack growth occurs with the sharp Elox notch
tip, a fatigued crack might well have indicated slight growth and thus
altered the constants in the regression curve. Nonetheless, it is believed
that this analysis indicates the dependence of crack extension upon the
initial crack and the feasibility of using such a relationship for estimation.
SULLIVAN AND STOOP ON THIN SHEETMATERIAL | 09
Because the final crack length 2a~ employed for these analyses is the
"effective"crack length determined from a COD calibration and includ-
ing a zone factor, one might envision this relationship between initial and
final cracklength as indicating that the plastic zone size is influenced by
27r ru
some constraint factor in the specimen since ~ = B
Screening Specimen
The relationship between K~ and ~.s shown in Fig. 1 indicates the
minimum specimen width required for various ratios of KJcry~[1]. How-
ever, it is suggested that a width of 12 in. is adequate for the majority of
high strength materials.
The steps of a possible screening procedure are outlined:
1. estimate K~, from the relationship between K~ and ~.~, Figs. 1 or 2,
2. calculate/3 = 1 [ K e ~ ,
B\ ~ys]
3. determine at~at from Fig. 5 or Eq 5,
4. select an initial crack length, 2at, such that 2a~ will be less than
2a/W=0.5 (a generally acceptable crack length to width ratio is
0.3),
5. perform the test, that is, load the specimen to fracture recording
the maximum load, and
6. compute K~ from the following equation
Ke=~a,,~ ]/ao~-a~f 2a/W (6)

I' at
Conclusions
1. There is an inverse relationship between Kr and ays.
2. The final crack length at instability failure is influenced by the
initial crack length.
3. This relationship is statistically acceptable in the In 10 (ao/ae)=
at A
A - B I n 10/3 and can be transformed to --=a~/3--~"
4. The relationship between ao/a~ and/3 suggests that the plastic zone
size varies with crack length.
5. Estimation of the final crack length permits a simplification of Ke
determination for an initial screening test.
Acknowledgment
The assistance of L. Wiener, Mathematics and Information Sciences
Division, Naval Research Laboratory, in providing consultation for and
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions aut
c o r r o b o r a t i o n of the statistical analysis is gratefully a c k n o w l e d g e d b y the

authors.
References
[1] Sullivan, A. M. and Freed, C. N., "The Influence of Geometric Variables on
Ko Values for Two Thin Sheet Aluminum Alloys," NRL Report 7270, 17 June
1971.
[2] Brown, W. F., Jr. and Srawley, J. E., Plain Strain Crack Toughness Testing of
High Strength Metallic Materials, A S T M STP 410, American Society for Test-
ing and Materials, 1966.
[3] Fretwell, C. C., "Regression Analysis," Wang Program Library, (300 Series),
Vol. 1, 1968.
[4] Ostle, B., Statistics in Research, Iowa State University Press, 1963.
[5] Carman, C. M. and Irwin, G. R., "Plane Stress Fracture Toughness Testing,"
Note for ASTM Committee E-24 Meeting, Cleveland, Ohio, 1969.
[6] Wang, D. Y. in Progress in Flaw Growth and Fracture Toughness Testing,
A S T M STP 536, American Society for Testing and Materials, 1973, pp. 334-349.
[7] Allen, F. C., "Stress Analysis of Centrally Cracked Plates," Douglas Paper 5513,
presented to ASTM Committee E-24 on Fracture Testing, Philadelphia, Pa.,
March 1969.
[8] Broek, D. in Aerospace Proceedings, 1966, pp. 811-835.
[9] Forman, R. G., "Experimental Program to Determine Effect of Crack Buckling
and Specimen Dimensions on Fracture Toughness of Thin Sheet Materials," A F
Flight Dynamics Laboratory Technical Report 65-146, Jan. 1966.
[10] Newman, J. C., Jr., Journal of Engineering Fracture Mechanics, VoL 1, 1968,
pp. 137-154.
R. H. Weitzmann 1 a n d I. F i n n i e ~
Further Studies of Crack Propagation

Using the Controlled Crack
Propagation Approach
REFERENCE: Weitzmann, R. H. and Finnie, I., "Further Studies of Crack

Propagation Using the Controlled Crack Propagation Approach," Fracture
Toughness and Slow-Stable Cracking, ASTM STP 559, American Society for
Testing and Materials, 1974, pp. 111-126.
ABSTRACT: In a previous paper, a new fracture-toughness-measuring tech-
nique was presented which made it possible to evaluate the functional rela-
tionship between load and cracklength at the moment of crack initiation or
arrest. It was demonstrated that this relationship is directly relatable to the
strain energy release rate and experimental results were presented for a 7075-
T6 aluminum alloy.
This paper presents the results of additional exploratory work as the con-
cept is applied to more 7075-T6 aluminum, a 250 grade maraging steel, and
AZ31B magnesium plate.
KEY WORDS. fracture properties, crack propagation, fracture strength,
fracture tests, crack stability, fatigue (materials)
An approach to fracture toughness testing will be described which

differs from conventional fracture toughness test methods in the manner
by which the crack-opening-mode load is applied to the specimen. In the
usual fracture toughness test the load is applied at fixed points of the
specimen geometry and crack tip stress intensity varies with load only.
In the present approach, by contrast, the point of load application is a
variable also and the stress intensity at the crack tip is therefore a func-
tion of both load and load position. A schematic representation of this
technique is given in Fig. 1. The I-beam specimen of arbitrary length is
securely held by upper and lower tracks as shown. A cluster of three
1 Senior development engineer, Space Sciences Laboratory, and professor of Me-

chanical Engineering, respectively, University of California, Berkeley, Calif. 94720.
111
Copyright9 1974
Downloaded/printed byby ASTM lntcrnational Www.astIII.OI'g
i... ' ... /
FIG. 1--Schematic of controlled displacement ]racture test apparatus.

rollers on each side of the specimen is atfixed to a common load frame
which allows the roller assembly to contact the tracks holding the speci-
men in the manner indicated. The spacing between the top roller and the
lower rollers is such that it is identical to the separation between the
roller-contacting track surfaces when the displacement bar has thickness
y---0. Thus, for y > 0 the resulting interference imposes a crack-opening-
mode force P. Crack initiation is brought about with the aid of a milled
starting cut when force F moves the roller assembly past the sloping
starting ramp of the upper track, and the crack propagation direction
is controlled by means of shallow side grooves of depth 1/2 (B-B,) on both
sides of the specimen. Once a natural crackfront has been established
the crack advances in discrete, small-scale steps whenever the advancing
roller assembly causes opening-mode force, P, and cracklength, a, to
assume values which raise the crack tip stress intensity to a critical level.
As we have shown[/], ~- the strain energy release rate may be written as
G- y(dP) (1)
where the negative sign indicates cracklength reduction as loading proceeds,
z The italic numbers in brackets refer to the list of references appended to this paper.
WEITZMANN AND FINNIE ON CONTROLLED CRACK PROPAGATION 1 13
Evaluation of Eq 1 at the point of crack instability yields the fracture

toughness, Go, associated with that point. Since values of Bn and y are
known, determination of fracture toughness requires merely the record
of the slope d P / d a at the point of instability, that is, the slope of the
record of P as a function of roller travel. From the data which we will
present it is seen that this test technique allows individual crack advance
steps to be kept to very small values resulting in a large number of data
points for specimens of modest length.
The test apparatus used in this investigation is shown in Figs. 2 through
4. The 3-in. by 1-in. I-beam shaped specimen of overall length 21 in. and
nominal web-thickness of 3~ in. is shown clamped to the hardened 250
grade maraging steel tracks which have been treated to 50-52 HRC.
Hardened steel back-up bars transmit the bolt-loads to the specimen and
assure intimate contact between tracks and specimen along the full length.
The roller assembly was machined from 4340 steel and has commercial
grade needle roller bearings, resulting in an opening-mode load capacity
of 12 000 lb and compliance across the rollers of 1.3 x 10-~ in./lb. Strain
gages placed on both sides of the assembly, directly above the top roller,
generate the summed P-force signal which, along with the crosshead travel
FIG. 2--1-beam shaped specimen clamped in load tracks.
1 14 FRACTURETOUGHNESS AND SLOW-STABLE CRACKING
FIG. 3--The roller assembly.
of the installed configuration in the Instron testing machine, yields the

required P versus a relationship which is recorded on a two-pen strip
chart recorder along with the pulling force F.
In summary, the present approach is characterized by the following
primary features.
1. Stress intensity at the crackfront is dependent upon two test varia-
bles: load and point of load application with respect to the crackfront.
2. Fracture toughness measurements are independent of compliance
calibration, knowledge of stress intensity, or cracklength measurements.
3. Since only two easily-recorded test values are needed for fracture
FIG. 4--General view of apparatus in tensile tester. For convenience in viewing

the crack, the roller assembly is held stationary and the specimen is moved.
toughness measurements (load and displacement) the complexity of the

method is comparable to that of a conventional tension test.
4. A continuous natural crackfront is propagated in a controlled way
with small individual crack advance steps, resulting in a relatively large
number of data points per inch of specimen length. While this feature
eliminates the need for fatigue precracking required in the conventional
test, the possibility that fracture toughness values measured from fatigued
preeracks and arrested cracks may differ should be recognized.
5. Relatively shallow side grooves are sufficient to control the crack
direction. Thus, the effect of side grooves can be minimized compared to
say a double cantilever beam specimen.
21.000-+'032 _1
I~.1AI.O05 TOTALI FX
1 5oo -L-
OOl max
r ---I -~%7~ ~
- 3 E - 1
I?;i,.oo,, r - - ~
2.000
.062+-.017~__ ~(SEE NOTE)
Ill
(4)PLCS ~ "-I--
/ p ,ioo
O~
~1 --~,~''~176 I-)x FACE GROOVE
SECTION X-X
NOTE: THE RANGE OFTHIS DIMENSION (2.000) DIMENSIONS IN INCHES

MAY BE FROM 1.998 TO 2.006 INCLUSIVE.
WIDTH OFCUT WITH RESPECT TO THE
OTHER SIDE MUST BE WITHIN .0005
FIG. 5--Detailed dimensions of 1-beam shaped specimen.
WEITZMANN AND FINNIE O N CONTROLLED CRACK PROPAGATION 1 17

7075-T6 Aluminum
A total of 305 fracture toughness values were obtained from three
21-in.-long specimens. Except for side groove details, all specimens in
this investigation were prepared according to dimensions given in Fig. 5.
In order to assure minimum deviations between the interference displace-
ment value of one side of the beam with respect to the opposite side when
clamped to the tracks of the mechanism, the tolerance on the 2-in. nominal
width of cut on either side of the beam was held to a value which assured
the two cuts to be within 0.0005-in. This was conveniently achieved by
using the same 2-in.-diameter milling cutter for both sides.
Specimens AL-B and AL-G were prepared from adjoining sections of
1-in.-thick commercial grade 7075-T6 aluminum alloy plate such that
crack propagation occurs in the RW (transverse) and WR (longitudinal)
directions, respectivel~r The nominal tensile yield strength (0.2 percent
offset) of this material is 73.0 ksi. The specimens were side-grooved with
0.006-in.-wide (square bottom) by 0.020-in.-deep cuts which resulted in
net section thicknesses of 0.335-in. and 0.334-in. for AL-B and AL-G,
respectively. For the 3/8 in. nominal thickness specimen, a ratio of net-to-
I I I I I I I
70--
!
SPECIMEN A L - G
7 0 7 5 T - 6 A L U M I N U M PLATE
03
NET SECTION THICKNESS: 0.335 IN.
60-- D I R E C T I O N : WR
bJ
50--
Z
"T-
<.9
o 40--
.. ~ ;..... ... ...... ... ......%.,. . .,,,. .. .'.-" "." .. . --"";'. .a
Ld
t~ *_.~ .%-."
owo ~ 9 9
OoS " ~ ~ o~ ~ 9 "*~176
30--
0
t~
h
20 I I I I I I I
0 2 4 6 8 I0 12 14 16
TOP ROLLER POSITION IN SPECIMEN, D, IN
FIG. 6--Fracture toughness values [or 7075-'1"6 aluminum with crack propagating
in the longitudinal (weak) direction.
gross-section thickness of 0.9 is obtained. A starting slot of 3-in. in length

was cut to facilitate natural crackfront initiation.
Figure 6 shows the plot of fracture toughness as a function of top
roller position, D, for Specimen AL-G. Because of fracture toughness
variations, the cracklength varies and for this reason the values have been
shown as a function of top roller position rather than crackfront position.
This is merely a matter of convenience since top roller position is accu-
rately known, whereas to plot fracture toughness as a function of crack-
front position requires a transformation through cracklength measure-
ments at each crack advance step. The data points shown were calculated
by using Eq 1 and assuming plane stress conditions so that Ke= (GEE) ~,
where E is Young's modulus, although the uniform fracture surface
appearance of 35 percent oblique fracture indicated a mixed mode of frac-
ture. The constant 0.020-in. interference displacement used resulted in a
cracklength range of 0.4 to 1.0 in. with an opening-mode load range of
3930 to 6600 lb. Over the total crack propagation distance of approxi-
mately 141/~ in. some 114 Kc values were obtained, resulting in an
average fracture toughness of 35.5 ksiVin, and upper and lower bounds
of 38.3 ksiVin, and 31.9 ksi~/in., respectively. Because of the large
I I I I I I I
:# 70-
_ 6C --~,,-~ o en e~176 9 9 ~149

y' U e9
9o 9 0.. 9149 9 ; "9 o~
9 ~149 "
; 9149149149
9
7-" ." .o~
9
":
9176 "
9176149 o
; "~
50-- 9" "~ 9 "- 9 9
LtJ
z
-'r-
40--
I.- SPECIMEN A L - 8
t..l..I 7 0 7 5 T-6 ALUMINUM PLATE
n.-
NET SECTION THICKNESS: 0.334 IN.
t.-.- 3 C - - DIRECTION: RW
t'~
a,
20 I I I I I I I
0 2 4 6 8 I0 12 14 i'6
TOP ROLLER P O S I T I O N ON S P E C I M E N , D, IN.
FIG. 7--Fracture toughness values/or 7075-T6 aluminum with crack propagating

in transverse (strong) direction.
number of data points obtained with this single specimen, statistical infor-
mation is meaningful and a standard deviation of 1.5 ksiVin, was
calculated.
Fracture toughness values obtained with Specimen AL-B are shown
in Fig. 7. A trial run for this transverse crack propagation direction indi-
cated significantly higher Kc values which led to the selection of an inter-
ference displacement of 0.039-in. The resulting cracklength range was
0.9 to 1.5 in. with opening-mode load extremes of 6610 and 10870 lb.
The average for the 111 Kc values plotted is 53.8 ksiv~in, which is
approximately 50 percent higher than that obtained for the longitudinal
direction specimen. Fracture toughness values ranged from 46.8 to
61.0 ksiVin, with a standard deviation of 3.3 ksiVin, indicating con-
siderably greater data scatter due to larger fracture toughness gradients.
The increased variation in K~ was also reflected in the fracture surface
appearance which ranged in amount of oblique fracture from 40 to 60
percent, with high toughness values being associated with maximum
oblique fracture. In all cases when we refer to oblique fracture, the frac-
ture surface is symmetrical with the 45 deg slant fracture portion starting
at the root of each side groove. In the limit with 100 percent oblique
fracture, the fracture surface is V-shaped.
FIG. 8--Variation of fracture toughness values with thickness obtained on a single

specimen o[ 7075-T6 aluminum.
One attractive feature of the present concept is its ability to readily

obtain fracture toughness as a function of net-section thickness since all
that is required is a tapered specimen with varying net section thickness.
Figure 8 shows K,. values for tapered Specimen AL-T3 as a function of
net-section thickness. For this tapered specimen all cracklength dimensions
were recorded so that K,. values could be associated with values of B,.
The specimen was machined from 7075-T6 l~/~-in, by 4-in. rectangular
bar material, whose average K~ for a constant 88 net-section thickness
had previously been found by the present method to be 29.5 ksiVin.
which is significantly lower than the comparable fracture toughness of
the plate material. Tapering of the specimen was accomplished by first
removing the bulk of the material in the fracture path on both sides of
the 3/8-in. web with a 1/~-in.-diameter ball end-mill, using a taper of 0.160
in./ft. Subsequent face-grooving in-line with the ball end-mill pass and
tapered the same amount was done with a 0.006-in.-wide slitting saw
with the depth of cut held constant at 0.020-in. This combination cut
served to reduce the uniform change in the B/Bn ratio from a value of
0.3 at the 0.020-in. net-section thickness start to 1.0 at the taper run-out.
In order to minimize cracklength variations for this tapered specimen, a
tapered displacement bar was used which caused y to vary from 0.005 to
0.023 in. with an opening-mode load range from 520 to 3350 lb. Crack-
lengths varied from 1.3 to 3.3 in. A photograph of the tapered specimen is
shown in Fig. 9.
The peak value of K~ in Fig. 8 is approximately 115 ksiVin, and
occurred at a net-section thickness of approximately 0.042 in. Super-
imposed on the data are Kc values reported by Irwin[2] and Kaufman[3]
for 7075-T6 aluminum plates without face grooves, fractured in tension
in the W R direction. While good agreement is obtained for small net-
section thicknesses, there is definite indication that a more rapid fall-off in
toughness with increasing thickness is obtained for the tapered, face-
FIG. 9--Photograph of fracture sur[ace of tapered specimen.
WEITZMANN AND FINNIE ON CONTROLLED CRACK PROPAGATION 121
grooved specimen. Probably, the presence of the groove increases the

constraint of the material surrounding the crackfront and thus aids in
establishing the beginning of partial plane-strain fracture sooner than
would be expected with an ungrooved sheet of the thickness B,. Also
plotted in Fig. 8 is percent oblique fracture. Despite the constraining influ-
ence of the face-groove, a full 45 deg oblique fracture surface is main-
tained well beyond the point of peak toughness. The beginning of a plane
portion on the fracture surface is not discernible until B, has increased to
approximately 0.09 in. where Kc has fallen off to approximately 45 ksiVin.
It has been suggested by Tetelman and McEvily[4] that the maximum
toughness is achieved at a thickness where plane-strain fracture just
begins to develop. Figure 8 and additional results to be presented, indi-
cate that the peak toughness is reached without regard to onset of a
plane fracture surface.
250 Grade Maraging Steel

Specimen MS-1 was machined from 6-in.-diameter bar stock in such a
way that the crack propagation plane was approximately in the center of
the bar. The specimen was machined with the material in the annealed
condition and subsequently aged to 52 HRC, with a yield strength (0.2
percent offset) of 250 ksi. The face-groove geometry is identical to that
of the tapered aluminum Specimen AL-T3 with the exception of the
0.020-in.-deep crack-direction-control groove which, because of the
hardness of the material in the maraged condition, had to be machined
with a 45 deg V-milling cutter, leaving a groove root radius of 0.005 in.
Figure 10 shows fracture toughness and percent oblique fracture as a
function of net-section thickness. The peak toughness of 237 ksi~/in.
occurs at B,~=0.038 in. while a plane portion on the fracture surface was
observed to develop at B. ~ 0.090 in. The Kc points plotted show again a
very sharp peak toughness curve which, after rapidly falling off to about
100 ksiVin., then shows a gradual toughness reduction as the net-section
thickness increases.
The ASTM recommended plane-strain requirement yields, for the 250
ksi yield strength material, and an assumed plane-strain fracture tough-
ness, KIe of 80 ksi~/i-~.,
B,a> 2.5 (KI~/S~) 2__ 2.5 (80/250) ~= 0.256 in.
This requirement is met for those points in Fig. 10 where B~>0.30-in.,
and the average value of Kie, based on 12 data points, has been cal-
culated to be 82.4 ksi!/in. This value is within 10 percent of the results of
Novak and Rolfe[5], giving a range of 86 to 87 ksiVin, for a somewhat
lower yield strength (246 ksi) maraging steel, and reports of a range of
..300/ i I I I I I
S P E C I M E N MS-I I00
2 5 0 GRADE MARAGING
STEEL BAR
DIRECTION: WR
~ K c 80
~ 200 I - W
OC
60
(J
~ 150-
40 ,,,
I:E
h
I00-
""..,..-.:.....,...'.;'"':..,,<.,'-..,,,.,..._
m
..J
I.- m
U 20 o
r,-
LI_
50--
0" I I I I I I 0
0 0.05 0.10 0.15 0.20 0.25 0.50 0.35
NET SECTION THICKNESS, B n , IN.
FIG. lO--Variation of fracture toughness values with thickness obtained on a

single specimen of 250 grade maraging steel.
87 to 88 ksi~/in, for a 232 ksi yield strength maraging steel by Srawley[6].

As was done in the case of Specimen AL-T3, tapered displacement bars
were used for Specimen MS-1 such that y varied linearly from 0.006 to
0.035 in. However, at a net-section thickness of 0.234 in. the interference
displacement of 0.024 in. proved to be too small to provide sufficient
cracklength and keep the opening-mode load within capacity of the test
apparatus. The test was interrupted at this time and the displacement
bars changed to provide a constant y-value of 0.061 in. for the remainder
of the test. The cracklengths for this specimen ranged from 0.5 to 3.5 in.
with load extremes of 1400 and 8940 lb.
AZ-31B Magnesium
This low yield strength ( 19 ksi, nominal) metal had been chosen as one
of the three materials investigated to determine whether the concept would
yield useful fracture toughness information even though considerable
plastic deformation might be associated with the fracture process. The
specimen was machined from 1-in. tooling plate in the fully-annealed
condition such that crack propagation occurred in the RW direction. A
tapered specimen geometry was again used with taper and side-groove
details identical to Specimen AL-T3. The interference displacement was
WEITZMANN AND FINNIE ON CONTROLLEDCRACK PROPAGATION 123
initially set at 0.040 in. but had to be increased to 0.070 in. at B,=0.230
in. when the opening-mode force signal indicated large-scale yielding.
Several data points had to be discarded because of this behavior.
When the crackfront had reached a net-section thickness in excess of
about 0.27 in., the opening mode load occasionally rose to values just
prior to crack-advance which again showed evidence of large-scale yield-
ing. Under these conditions, the slope dP/da following crack propagation
was measured. Thus, the reported fracture toughness values are associated
with crack arrest rather than crack advance for this material.
The plot of Ke as a function of net section thickness is given in Fig. 11.
A peak fracture toughness of 84 ksi~/in, occurs at B,=0.040 in. It is
also noted that the absence of any plane fracture surface suggests more
strongly than the results obtained with the 7075-T6 aluminum or the
250 grade maraglng steel that some criterion other than the onset of a
plane fracture surface is involved in bringing about a maximum in frac-
ture toughness. Load and cracklength range for this specimen were 1000
to 9810 lb, and 0.5 to 3.0 in., respectively.
In general, the Ke values obtained for the materials investigated do not
appear to be randomly distributed in the fracture path of a given speci-
men. As is particularly noticeable for Specimens AL-B and MS-I, the
I I I I I I
- z IOO -- 0 0.,-~
I00
SPEC,MEN MAG-C
__ i~ AZ 31 B MAGNESIUMPLATE
80 J ~ DIRECTION: RW - -
80
z 9 Kc
/ ~ o OBLIQUEFRACTURE, %
60
i
~<2.~'. . 9
9 eeee 9 "e
...:..--
oe
"..2.~
9 . ~
. "." ..;%.~,..
qW~
-%.
~
~.
4o 40 ,,
o
W
n~ ._J
rn
20 20 o
0
<I
b_
0 I I I I I I 0
0 0.05 O. IO 0.15 0.20 0.25 0.50 0.35
NET SECTION THICKNESS, Bn, IN.
FIG. 11--Variation of fracture toughness values with thickness obtained on a

single specimen of .4Z31B magnesium.
"continuous" type of fracture toughness variation suggests that the smooth

toughness gradient may be due to metallurgical variations which extend
over distances of the order of inches. No correlation could be found
between the periodicity of the fracture toughness variation and features of
the apparatus.
General Test Aspects

Several unique aspects are associated with the present fracture test.
The choice of an appropriate interference displacement may be made if
an estimate of fracture toughness is available for a given material. As was
reported earlier[/], the stress intensity at the crackfront is given by
C
KI=~-~-, ea 1 (3)
where the stress intensity constant, C, had been found to be approxi-

mately 2.0 for the present test configuration by comparison with experi-
mental results. Thus, for a desired cracklength, a value of P may be
computed. The interference displacement associated with the computed
P-force can then be obtained through the spring constant of the mecha-
nism/specimen combination at the desired cracklength. Since the stored
elastic energy available for crack propagation is reduced as the crack-
length is reduced, it is generally desirable to select the shortest crack-
length compatible with elastic behavior in order to reduce the individual
crack advance steps and maximize the number of data points.
As loading proceeds between contiguous crack advance steps, the
cracklength is reduced and the loading rate is affected by the change in
cracklength. It may be shown that the loading rate under these condi-
tions can be written as
dKx C da[- P ~- ~P -]
dt - B , d-tL2~a + Va a a J (4)
For the uniform pull speed (da/dt) of 0.5in./min used in this investi-
gation, a typical loading rate of 0.1 ksiVin./sec was obtained for the
specimens of constant net thickness.
In conventional fracture toughness tests, a natural crackfront is estab-
lished through fatigue loading of a machined starter slot and crack
extension is normally limited to some small fraction of the specimen
thickness. While the determination of plane-strain fracture toughness is
not dependent on the distance of natural crackfront propagation, this is
not the case for specimens whose thickness is insufficient for plane-strain
conditions. For specimen thicknesses which preclude plane-strain condi-
WEITZMANN AND FINNIE ON CONTROLLED CRACK PROPAGATION 125
tions there is, in general, a transition region between the point of natural
crackfront initiation until the crackfront has progressed sufficiently far
into the material to have attained its final shape and amount of oblique
fracture. Kc for a constant thickness specimen can thus vary in this
transition region as a result of crack front-geometry changes. The data
presented for the constant Bn Specimens AL-B and AL-G are fully-
developed crackfront data based on a constant amount of oblique fracture.
The transition region for Specimen AL-G, for example, was nearly 1, in.
with initial Kc values in the transition region being as much as 20 percent
lower than the average fracture toughness for that specimen under fully-
developed crackfront conditions.
The advantage of knowing K~c in a design application is obvious in
those cases where the crackfront is subjected to plane-strain condi-
tions. There are, however, important cases where it is more meaningful to
know K~ for a given material as a function of thickness, as for example in
predicting safe-venting or catastrophic failure in pressure vessels. For many
important medium-strength materials the mixed-mode fracture toughness at
moderate thickness is sufficiently high to allow cracklengths of the order of
inches before the critical stress intensity is reached. It is, therefore, im-
portant to know the value of K~ under conditions of a fully-established
crackfront shape. For this type of toughness testing the present approach
appears to have advantages.
Conclusions
The controlled displacement fracture test concept, which makes possible
the determination of fracture toughness independent of compliance cali-
bration, knowledge of stress intensity, or cracklength measurements, has
been applied to three different metallic materials, and test results obtained
indicate the following.
1. For the case of a 250 grade maraging steel, where the specimen
thickness met ASTM recommended plane-strain conditions, a K~ value
of 82.4 ksi\/in, was determined which compares well with published data
obtained with conventional test methods.
2. The concept is particularly useful in obtaining fracture toughness
as a function of material thickness.
3. The occurrence of maximum fracture toughness as a function of
material thickness is not, in general, associated with the onset of a plane
fracture surface.
4. Fracture toughness variations in the path of fracture for a single
specimen are not, in general, randomly distributed but may vary in a
continuous fashion over distances of the order of inches.
Acknowledgrnents
The L a w r e n c e Radiation Laboratory, University of California, Liver-
more, constructed the test mechanism and the specimens.
References
[1] Weitzmann, R. H. and Finnie, Iain, Journal ot Materials, American Society for
Testing and Materials, Vol. 7, No. 3, Sept. 1972, pp. 294-298.
[2] Irwin, G. R., Transactions, American Society of Mechanical Engineers, Journal
of Basic Engineering, Vol. 82, 1960, pp. 417-425.
[3] Kaufman, J. G. in Review of Developments in Plane Strain Fracture Toughness
Testing, ASTM STP 463, American Society for Testing and Materials, 1970,
pp. 3-21.
[4] Tetelman, A. S. and MeEvily, A. J., Fracture of Structural Materials, Wiley, New
York, 1967, pp. 132-139.
[5] Novak, S. R. and Rolfe, S. T. in Plane Strain Crack Toughness Testing of High
Strength Materials, ASTM STP 410, American Society for Testing and Materials,
1966, pp. 126-129.
[6] Srawley, J. R. in Fracture 1969, P. L. Pratt, Ed., Chapman and Hall Ltd., Lon-
don, 1967, p. 131.
J. O. O u t w a t e r , 1 M . C. M u r p h y , ~ R . G. K u m b l e , 3 a n d J. T. B erry
Double Torsion Technique as a Universal

Fracture Toughness Test Method
REFERENCE: Outwater, J. O., Murphy, M. C., Kumble, R. G., and Berry,

J. T., 'Double Torsion Technique as a Universal Fracture Toughness Test
Method," Fracture Toughness and Slow-Stable Cracking, ASTM STP 559,
ABSTRACT: A method of measuring the fracture toughness of materials is
described that is both simple and universal. A specimen is failed in double
torsion so that a crack is driven down the center of a rectangular specimen.
The energetics of failure are such that the load causing the crack remains
constant as the crack deepens and this load is directly related to the fracture
toughness. This enables a simple specimen to be used and obviates the diffi-
culties associated with crack depth or compliance measurements or both. Re-
sults are shown for aluminum alloys to supplement earlier work on glass and
polymers.
K E Y WORDS: stresses, fatigue (materials), fracture properties, double tor-
sion, tests, aluminum
T h e c o n c e p t of fracture energy is well a c c e p t e d a n d its q u a n t i t a t i v e

use as an i m p o r t a n t design m e a s u r e m e n t is gaining acceptance. O n e of
the p r o b l e m s a s s o c i a t e d with its use is the difficulty i n v o l v e d in m a k i n g
tests to d e t e r m i n e its actual v a l u e w h e n t h i c k plates are u s e d as the
structural m e m b e r s . Such plates w o u l d r e q u i r e expensive a n d h e a v y
m e a s u r i n g devices if c o n v e n t i o n a l m e t h o d s are used. Difficulties in m e a -
s u r e m e n t are n o t u n i q u e to thick plates either. N o t c h e d specimens u s e d
in a b e n d i n g t y p e test b e c o m e m o r e c o m p l i a n t in a n o n l i n e a r fashion as
1 Professor, Department of Mechanical Engineering, University of Vermont, Bur-

lington, Vt. 05401.
2 Assistant professor, Department of Engineering and Technology, Norwich Uni-
versity, Norwich, Vt. 05663.
3Assistant professor (visiting), Department of Industrial Engineering, University
of Rhode Island, Kingston, R. I. 02881.
Professor, School of Mechanical Engineering, The Georgia Institute of Technol-
ogy, Atlanta, Ga. 30332.
127
Copyright*1974byby ASTM International
the crack depth increases so the crack tends to be unstable and, conse-
quently, demands precision and care when such specimens are used as
the actual depth of the crack is important and difficult to measure.
The double torsion method avoids many of these difficulties. The com-
pliance of the specimen is linearly proportional to the crack depth. The
critical forces involved in propagating the crack are independent of the
crack length so the critical force applied by the testing machine does not
vary and can be measured in a quasi-static fashion. The specimens are
relatively small even for long crack lengths so small forces are involved.
The most important combination of factors is the convenient size and
shape of the specimens requiring only a rectangular piece with grooves
milled on each side and no tensile loading attachments--coupled with
the fact that the actual depth of the crack is not germane to the compu-
tation of the fracture energy.
Fundamental Concepts of Fracture Mechanics

The fracture of any structure demands the creation of new surface.
The energy required to create this surface is the fracture energy. It is
made up of two components: the energy required to create the surface
and the energy dissipated in the material as the crack penetrates.
Griffiths[Ip originated this concept and applied it only to brittle materials.
In this instance the plastic flow was negligible and the fracture energy
was the surface energy of the substance. Irwin[2] modified this con-
cept by accounting for the energy dissipated at the crack tip. In many
engineering materials this energy is substantially greater than the surface
energy, but, as the volume of material deformed plastically is essentially
small and uniform as the crack deepens, the concept of a fracture energy
still holds and is an important p r 0 p e r t y - - a measure of the material's
resistance to fracture.
Irwin[3] determined the stress state in an isotropic elastic medium
near the crack tip and showed it to be related to a function of the posi-
tion of the point and a quantity that depended on the nature of the crack
and the loads on the body. This quantity he referred to as the Stress
Intensity Factor, K. When the load on a body containing a crack reaches a
critical value, the crack will propagate. Corresponding to this load is a
critical stress intensity factor, Ke. Hence, the structure's ability to resist
crack propagation is measured either by the critical stress intensity factor,
K~e, or by its fracture energy G~.
paper.
OUTWATER ET AL ON DOUBLE TORSION TECHNIQUE 129
Theoretical and Experimental Limitations

Both Griffith's and Irwin's theories determine the necessary energetic
conditions for fracture; accordingly, flaws or cracks must be present in the
materials before fracture. The fracture toughness is a correlation of the
effects of load and crack length, hence a typical test would involve a
measurement of the load and the crack depth during the test. Most test
methods have been developed for metallic specimens of moderate strength
and use configurations as in Fig. 1. Brittle materials can be adapted only
with difficulty to such test methods. Furthermore, the minimum thick-
ness of conventional plane-strain fracture toughness specimens increases
T r12 PI2
I L
i J I
T
r12 elz
p
P
I I
p
F I G . 1--Schematic of specimens commonly used [or fracture toughness testing o/

metallic materials.
with decreasing yield strength[4--6] implying that a valid test would

consume substantial quantities of material and large testing machines
would be needed for low strength materials. It is then impossible to use a
single experimental technique over the whole material spectrum.
The double torsion method eliminates the handicaps of conventional
methods and leads to much simpler specimens, computation methods, and
experimental techniques.
Theoretical Basis for the Double Torsion Technique

The double torsion specimen is essentially an elastic plate containing a
crack with a curved profile, Fig. 2. When this crack propagates an incre-
ment of crack surface area, dAc, is created. Applying this principle of
conservation of energy to the process, we obtain
dW~ - dU~ = G~cdA~ ( 1)
where W. is the work done on the boundary of the specimen, U. is the
strain energy in the double torsion specimen, and Gxe is the Mode I
fracture energy.
But,
d W~ = MtdO
and
dUs: V2d(MtO)
where M t is the torque and 0 is the angular displacement.
Substituting in Eq 1 gives
L ~ -f
F -i
FIG. 2--A sketch of a double torsion specimen under load.
Gied/4 c = 1//2Mt2d(~--7-) (2)
If we assume that the curved profile of the crack front propagates along
the specimen in such a way that the profile itself remains unchanged,
Fig. 3, then the relationship between the crack surface area, the crack
thickness to, and the crack depth is
dA o= t~dc (3)
The torsional compliance, Lt, is defined as
0
gt~ --
Mt
and
(4)
Substituting Eqs 3 and 4 into Eq 2 and solving for G~c yields
GI~ = Mt~ dL~ (5)

2to dc
This equation is analogous to the Irwin[2] equation derived for G~e
of an elastic plate containing a through thickness crack (straight crack
front) in terms of load and the derivative of compliance with respect to
crack length.
Each half of the double torsion specimen can be regarded as a prism
with rectangular cross-section rigidly held at one end and subjected to a
torque, Mt, at the other. The length of the prism is the effective crack
length, ce, which differs'from the true overall crack length, c, by a con-
stant to allow for the end effect.
+t
7-
o
FIG. 3--A sketch of a double torsion specimen crack]ront.
c=k +ce
The torsional compliance of one half of the double torsion specimen is
0 ce
Lt-- Mt I~G
where I~ is he polar moment of inertia of one half of a double torsion
specimen, and G is the shear modulus, and
dLt_ I (6)
dce - I~G
which is a constant dependent only on the cross-section and modulus of
the specimen.
When the double torsion specimen is loaded as shown in Fig. 2,
M~= P~Wm/4
where Pc and Wm are the critical loads and the distance between the sup-
ports, respectively.
Substituting in Eq 5 for Ms and dLt/dc~ yields
(P~ (7)
G~c= 16tJ~G
now lp for a rectangular cross-section is given by Roark[7] as
Wt~/16 t 4 t 4
~,=-~-~-ff- -6.72 ~ ( 1 - ~ (-~) ))~Wt3/6
and W and t are the width and thickness of the double torsion specimen,
respectively.
Paris et all8] give the relationship between K~c and G~c, in the case of
plane strain, as
K [ EGIr \+ (8>
where E and v are the modulus of elasticity and Poisson's ratio, respec-
tively. Substituting for Gic and E = 2 ( 1 + v) G gives
KIr = P~(
8 t JW'~2
~ - - v ) )+ (9)
These equations giving G~c and K~c do not contain any value demanding
knowledge of the depth of the crack. The fracture energy and the critical
stress intensity factor can then be determined from knowledge of the
geometry, elastic modulus, and the load needed to propagate a crack.
We do not have to measure the crack depth nor expect the critical load
to vary as the crack depth increases. The load needed to propagate a

crack can be applied and the values checked over and over again on the
same specimen. This implies unusual accuracy and simplicity.
Experimental Procedure and Results

The application of the double torsion method to metallic alloys of
moderate strength and ductility was the purpose of this investigation.
Consequently, a wrought aluminum alloy (2124-T85l) of known tough-
ness was selected for testing. The fracture toughness of this material had
been determined previously from fatigue precracked compact tension
specimens (thickness= 1.5 in.). The recommended ASTM thickness[9] is
given by the empirical formula
where cr~ is the material's yield strength. Using the properties for 2124-
T851, Table 1, the minimum specimen thickness for this alloy is ap-
proximately 0.45 in. (1.14 cm). However, the thickness of the double
torsion specimens ranged from 0.25 to 0.50 in. (0.634 to 1.27 cm) while
the length and width remained approximately 8.0 in. (20.3 cm) and
2.8 in. (7.11 cm), respectively.
The double torsion specimens were loaded as shown in Fig. 2 until a
crack approximately 2 in. (5.08 cm) long had formed. In this manner
the critical load, Pc, was obtained for a nonfatigue crack. Before the
critical load to propagate a fatigue crack could be determined a fatigue
crack had to be created. This was done by cycling the load over a range
0.1 to 0.6 of the nonfatigue crack critical load when the existing crack
grew into a fatigue crack. About 10 h elapsed to extend the crack 0.125
in. (0.318 cm). This was partly a result of the low cycle (about 0.5 hz)
capability of the testing machine. The reloaded specimen cracked at the
critical load for a fatigue crack; the corresponding fracture toughness was
consistently 0.81 to 0.89 of the nonfatigue value in this alloy, Table 2.
When a load is applied by a constant crosshead velocity, as with an
Instron testing machine, then the movement of the crack tends to reduce
the load until the moving crosshead increases it again to the critical value.
Thus, it is possible to obtain a large number of readings on each specimen,
TABLE 1--Properties o[ 2124-1"851 aluminum alloy.
Yields Strength (~r~) Fracture Toughness (K~c)

67.0 ksi (46.1 MN/m2) 28.4 ksi x/in. (31.1 MN/m~/~)
CopyrightbyASTMInt'l(allrightsreserved);MonDec 713:10:55EST2015
Downloaded/printedby
UniversityofWashington(UniversityofWashington)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
TABLE 2--Ratio of the mean fracture toughness (K~) for fatigue precracks
to the mean Jracture toughness (K~o) for nonfatigue precracks.
Specimen Thickness
Specimen in. cm Ratio
1 0.25 0.635 0.807
2 0.35 0.889 0.867
3 0.35 0.889 0.855
4 0.35 0.889 0.888
5 0.52 1.321 0.873
TABLE 3--Material: 2124-T851 aluminum alloy.
Data for nonfatigue precracks

Specimen Thickness Kxe
Mean of
in. cm ksi x/in. MN/m a/~ (x) values
0.25 0.635 33.8 37.1 8
0.35 0.889 35.8 39.3 55
0.52 1.320 34.9 38.3 19
Data for fatigue precraeks
Specimen Thickness K~c
Mean of
in. cm ksi Vin. MN/m a/2 (x) values
0.25 0.635 27.3 30.0 4
0.35 0.889 29.4 32.3 8
0.52 1.320 27.7 30.4 2
Mean KI c o[ nine bl~eeimen$ with thickness of 0.25 to 0.$2 in. (0.635 to 1.320 cm).
Type of Precrack
Fatigue Nonfatigue
28.3 ksi x/in. (31.1 MN/m 8/~) 34.9 ksi Vin. (38.3 MN/m 8/~)
Fig. 4. Approximately 100 values including both fatigue and nonfatigue

fracture toughnesses were computed from the nine specimens tested.
The gross fracture was square (perpendicular to the direction of ap-
plied stress) and showed no evidence of shear lips, Fig. 5, over the full
range of specimen thicknesses. So, it was not surprising that the effect
of specimen thickness was not observed in either the fatigue or non-
fatigue crack fracture toughness data, Table 3. Although the minimum
specimen thickness was one half the specimen thickness recommended
for compact tension specimens, these tests did not establish a lower limit
for double torsion specimen thickness. Thicknesses as low as 0.25 of the
recommended value have been used in recent double torsion tests on a
cast aluminum aUoy[10].
Distribution: A. A specimen along which KIC varied s i ~ i f i c a n t ~

over a 6.0 i n . ( 1 5 . 2 4 cm.} l e n g t h , however a
r e l a t i v e l y c o n t i n u i t y i n the d i s t r i b u t i o n was
e v i d e n t ( n o n - f a t i g u e precraok).
B. A specimen along which KTC was essentially c o ~
40 s~u~t over a 6.0 in.(15.2~ ore.) length for .44
both the non-~atigue preorack(B 1) and the
fatigue preorack(B2).
~9
37 S
.40
_ __ specimen mean I. . . . . . . .
,38
r
I I_ s ~ c i ~ n man N
3Z "35~
B1
,33
N
J
28 specimen mean
.31
311
27
.29
26
S PFd2~d~N Lg]ti~tl
FIG. 4---Fracture toughness distributions along 2124-T851 double torsion specimens.
Discussion
Initial work with the double torsion method was conducted by Out-
water et a l [ l l ] investigating slow crack propagation in glasses. Stable
crack growth is possible in materials such as glass which exhibit slow
crack growth under static load (static fatigue). The fracture energy is a
FIG. 5 - - A photograph of fractured 2124-T851 double torsion specimen.
function of both load and crack length for most test specimens; conse-
quently, loads must be continually adjusted during the test to produce a
constant crack driving energy to maintain a constant velocity crack. Since
the fracture energy of the double torsion specimen depends only on the
magnitude of the load during the test, static loading of glass specimens
produced constant velocity cracks. These results agreed with those of
other investigators[12, 13] confirming Eq 7 which is the mathematical
essence of the double torsion technique. There has also been good agree-
ment between the results obtained by other methods[14-16] and by
double torsion techniques[17, 18] for polymethylmethacrylate, epoxy, and
polyester resins.
The effect of specimen thickness is insignificant with brittle materials
as the size of the plastic zone at the crack tip is negligible. This is not
true, however, in low strength metallic materials; hence the results of
testing the aluminum alloy 2124-T851 are particularly significant: a good
correlation was obtained between the recognized fracture toughness test
and the double torsion test indicating the validity of this test. The results
of these tests re-emphasize the importance of basing design criteria on
plane-strain fracture toughness even in thin sections.
Conclusions
The double torsion technique for the measurement of fracture energy
appears to have many advantages over the present methods of determining
the opening mode fracture energy.
1. The specimen is of a simple rectangular shape.
2. The mechanics of the technique is such that it is not important to

measure the depth of the crack.
3. The crack is stable as it penetrates the material, permitting many
measurements of the value of the fracture energy on one specimen.
4. The crack length can be greater than the thickness of the specimen.
5. The loads required to measure the fracture energy are substantially
smaller than with the usual techniques as a result of the curved Mode I
crack.
Accordingly, the method just described is a valid and simple method
of determining the fracture toughness of a material.
Acknowledgment
The material used in this investigation was kindly supplied by D. J.
Brownhill of Alcoa Corporation.
Author's Note Added in Proof

It is gratifying to see that both Evans and co-workers[19, 20, 21] and
McKinney and Smith[22] have recently realized the potential of our test
and have been able to press it into use with alumina-water[21] and a
glass-ceramic[22] as well as with other more common materials. The
authors also wish to point out that the test has been under continual
development at U.V.M. since its inception there by Outwater and Gerry
in 1965, see, for example, the Doctoral Dissertations of R, G. Kumbl6
(1973) and M. C. Murphy (1972) and earlier related publications[23-26].
References
[1] Griflith, A. A., Philosophical Transactions of Royal Society, Series A, Vol. 221,
1920, pp. 163-198.
[2] Irwin, G. R., Encyclopedia o/Physics, Springer, Berlin, Vol. VI, 1958, pp. 551-
590.
[3] Irwin, G. R. in Structural Mechanics, Pergamon Press, New York, 1960, pp. 557-
594.
[4] Srawley, J. E. and Brown, W. F., Jr. in Fracture Toughness Testing and Its Ap-
plications, ASTM STP 381, American Society for Testing and Materials, 1965,
pp. 133-198.
[5] Knott, J. F., Materials Science and Engineering, Vol. 7, 1971, pp. 1-70.
[6] Osgood, C. C., Machine Design, Vol. 11, No. 9, 1971, pp. 421-428.
[7] Roark, R. J., Formulas/or Stress and Strain, McGraw-Hill, New York, 1959.
[8] Paris, R. C. and Sih, G. C., in Fracture Toughness Testing and Its Application,
ASTM STP 381, American Society for Testing and Materials, 1965, pp. 59-61.
[9] ASTM Committee E24, Tenative Method of Test for Plane Strain Fracture
Toughness in Metallic Materials, (E 399-70T), American Society for Testing
and Materials, 1970, pp. 919-935.
[10] Murphy, M. C. et al, "Fracture Toughness Determination in Cast Materials,"
Submitted to American Foundrymen's Society, 1973.
[11] Outwater, J. O. and D. J. Gerry, "On the Fracture Energy of Glass," U.S.N.R.L.
Report Contract NONR 3219 (01)(x), 1966, AD640848.
[12] Shand, E. B., Journal of the American Ceramic Society, Vol. 44, No. 1, 1961,
pp. 21-26.
[13] Roesler, F. G. in Proceedings, Royal Society, LXIX, 10, 1956.
[14] Mukherjee, B. and Burns, D. J., Experimental Mechanics, Vot. 11, No. 10, 1971,
pp. 433--439.
[15] Mostovoy, S. and E. J. Ripling, "Factors Controlling the Strength of Composite
Structures," Final Report Contract No. N00019-70-C-0137, Naval Air Sys-
tems Command, 1971.
[16] McGarry, F. J. and Mandell, J. F., "Fracture Toughness of Fiberous Glass Re-
inforced Plastic Composites," Proceedings, 27th Annual Conference Reinforced
Plastics/Composites Institute, SPE, 1972, Section 9-A.
[17] Outwater, J. O., and Carnes, W. O., "Fracture Mechanics of Composite Mate-
rials," Proceedings, U.S. Army Solid Mechanics Symposium, Watervliet Ar-
senal, 1968.
[18] Murphy, M. C., Unpublished data, "On the Fracture Energy of Epoxy and
Polyester Resin Systems."
[19] Evans, A. G., Journal of Materials Science, Vol. 7, 1972, p. 1137.
[20] Williams, D. P. and Evans, A. G., Journal o] Testing and Evaluation, Aug. 1973.
[21] Evans, A. G., International Journal of Fracture, Vol. 9, 1973, p. 267.
[22] McKinney, K. R. and Smith, H. L., Journal of the American Ceramic Society,
Vol. 56, 1973, p. 230.
[23] Outwater, J. O. and Gerry, D. J., "On the Fracture Energy of Glass," U. S.
Naval Research Laboratory Report Contract NONR 3219(01)(1), 1966,
[24] Outwater, J. O. and Gerry, D. J., "On the Fracture Energy, Rehealing Velocity
and Refracture Energy of Cast Epoxy Resin," 22nd Society of Plastics Industry
Conference, Paper 13-D, 1967; also, Journal o] Adhesion, Vol. 1, 1969, pp. 290-
298.
[25] Outwater, J. O. and Austin, L. E. in Materials Per]ormance and the Deep Sea,
[26] Outwater, J. O. and Murphy, M. C., "On the Fatigue of Epoxy Resin," 26th So-
ciety of Plastics Industry Conference, Paper 10-A, 1971.
J. N . R o b i n s o n 1 a n d A . S. T e t e l m a n ~"
Measurement of on Small
Specimens Using Critical Crack
Tip Opening Displacement
REFERENCE: Robinson, J. N. and Tetelman, A. S., "Measurement of K~o

on Small Specimens Using Critical Crack Tip Opening Displacement," Frac-
ture Toughness and Slow-Stable Cracking, .4STM STP 559, American So-
ciety for Testing and Materials, 1974, pp. 139--158.
ABSTRACT: A method is described whereby on-loact values of crack tip
crack opening displacement (COD) can be measured at the midsection of
precracked three-point bend specimens by infiltration of silicone rubber. A
calibration curve relating midsection COD to clip gage displacement was
derived from measurements on the silicone rubber "castings." This calibra-
tion curve can be used to calculate midsection COD from on-load clip gage
displacement and specimen geometry only, without further infiltration mea-
surements. These values of COD have been shown to be simply related to
the stress intensity factor, independent of material, as theoretically predicted.
The central region of a Charpy specimen in three-point bend has been
found to remain in plane strain until well after general yield. Thus, plane
strain values of COD at fracture initiation, (COD)~, can be determined
from small specimens. Two initiation detection methods are described
whereby (COD)~ can be determined. These values of (COD)~ can be used
to accurately predict Kzo values which agree with data obtained on larger,
more expensive valid ASTM Kzc specimens.
KEY WORDS: fractures (materials), fracture strength, mechanical proper-
ties, cracking (fracturing), crack initiation, plastic analysis, strains, bend
tests, steels, alloy steels, pressure vessel steel, aluminum alloys, titanium
alloys
1Formerly, postgraduate research engineer, Materials Department, School of Engi-

neering and Applied Science, University of California, Los Angeles, Calif., 90024;
now, postdoctoral research assistant, Department of Metallurgy and Materials Sci-
ence, Imperial College, London, England.
-~Professor and chairman, Materials Department, School of Engineering and Ap-
plied Science, University of California, Los Angeles, Calif. 90024.
139
Copyright*1974 by
Downloaded/printed byASTM International www.astm.org
Linear elastic fracture mechanics provides an accurate one-parameter

fracture criterion. Unfortunately, the requirement to have contained
plasticity and plane-strain conditions over almost the whole specimen
results in large and expensive test specimens for all but low toughness,
high strength materials. There is therefore an urgent need for an alterna-
tive technique for measuring plane-strain fracture toughness using small,
inexpensive specimens. Such a technique must have several characteristics.
First, it should provide a single parameter which can be measured at the
specimen midsection, where, as shown later, plane-strain conditions exist
at fracture initiation even in small specimens of tough materials. Second,
it must be based on a model which allows large-scale yielding. In addi-
tion, a method for the determination of fracture initiation is required,
since the compliance method used for linear elastic fracture toughness
specimens cannot be used in the presence of large-scale yielding. Finally,
the critical value of the parameter thus obtained should be simply related
to a criterion of proven value, such as K~c.
Two parameters are available which are not limited to small scale
yielding, the J-integral[i] "~ and the crack tip crack opening displace-
ment[2, 3]. The only method yet available for measurement of the J-
integral[4] produces a thickness average value, rather than a plane strain,
and previous measurements of crack tip crack opening displacement on
sharply cracked specimens were made at the outside surface of the speci-
men. It would appear that a simple technique for measuring on-load crack
tip crack opening displacement (COD) at the specimen midsection would
satisfy all the requirements mentioned earlier. We have recently developed
such a method, which is described in the body of this paper.
Various techniques for the detection of fracture initiation have also been
investigated and those which have been found satisfactory for the deter-
mination of (COD)c, the critical plane strain COD at fracture initiation,
are discussed in detail later. The theoretical equation relating (COD)c to
Kic has been empirically verified and the justification for it is discussed.
This paper describes a technique whereby Kic values can be obtained
from (COD)c values on precracked Charpy specimens 10-ram thick.
These values are in agreement with valid ASTM Kic measurements made
on specimens 25 to 300-mm thick. The use of this technique results in
large savings in materials, machining time, and equipment capacity. Some
of the techniques reported herein have been outlined by Robinson and
Tetelman[5]. The present paper describes the work in greater detail, ex-
tends the technique to materials other than steels, contains further ex-
paper.
ROBINSON AND TETELMAN ON CRACK TIP DISPLACEMENT ld1
p e r i m e n t a l verification, and reports m o r e c o n v e n i e n t m e t h o d s of m e a s u r e -

m e n t of (COD)~..
Materials"
A wide r a n g e of metals were used for the e x p e r i m e n t s so as to d e t e r -
m i n e w h e t h e r the results were m a t e r i a l d e p e n d e n t . T h e metals used,
together with their h e a t t r e a t m e n t a n d stress-strain relations, are shown
in T a b l e 1. T h e A 5 3 3 B steel was p r o v i d e d by the H e a v y - S e c t i o n Steel
T e c h n o l o g y ( H S S T ) p r o g r a m f r o m Plates 01 a n d 02.
Specimen Design
T h e m a j o r i t y of the specimens used were of n o r m a l C h a r p y V d i m e n -
sions ( 1 0 b y 10 b y 55-ram long with a 2-ram deep m a c h i n e d n o t c h )
with either a fatigue p r e c r a c k o r a notch of 0 . 0 5 - m m r o o t radius. T h e
total c r a c k a n d notch length for p r e c r a c k e d specimens v a r i e d b e t w e e n
2.4 m m a n d 3.3 m m . S o m e half and q u a r t e r - t h i c k n e s s p r e c r a c k e d speci-
mens were also used. Kxo specimens were in a c c o r d a n c e with the A S T M
Test for P l a n e - S t r a i n F r a c t u r e Toughness of M e t a l l i c M a t e r i a l s ( E 3 9 9 -
7 2 ) , except that the m a c h i n e d notch was 3 . 2 - m m wide to facilitate inser-
TABLE 1--Test materials.
Relation between
Yield Effective True Stress
Stress, (MN/m 2) and Effective
Material Condition MN/m 2 True Strain
Steel
A533B 914 ~ C, 4 h, air cooled 470 ~=1260~~ < 0.08
871 ~ C, 4 h, water quenched = 1050~~ ~-> 0.08
663 ~ C, 4 h, furnace cooled
621 ~ C, 40 h, furnace cooled
4340 oil quenched, tempered 1 h,
tempering temperature 677 ~ C 720 ~ = 1290~~
616 ~ C 910 ~ = 1450~~176
565 ~ C 1100 ~=1560~ ~176
540 ~ C 1200
450 ~ C 1310 5:1710~ '~176
Aluminum
2024 O temper 90 ~-----380~~176
T4 temper 310 ~=905z ~176
Titanium
Ti-6As wrought 830 . . .
tion of the silicone rubber. All specimens were machined with axis
parallel to the rolling direction and crack propagation parallel to the width
direction (RW orientation). Knife edges to hold the clip gage were glued
on the top surface of the Charpy specimens, with the nearest attachment
point approximately 6 mm from the notch.
Transverse Strain Measurement

A set of lines parallel to the specimen axis was printed with Kodak
photoresist on the notch root of a Charpy V specimen. The line thickness
was 0.05 mm and the line spacing was 0.5 mm. The position of the
center of each line was measured using a travelling microscope with
graduations of 0.0025 ram. The specimen was then loaded in three-point
bending. At intervals in off-load angle of bend (0) of approximately 0.7
deg, the specimen was unloaded and the positions of the centers of the
lines again determined. The measured displacements were then converted
to strains.
Measurements of COD Using Silicone Rubber

A technique was developed whereby the crack tip COD was measured
by infiltrating the crack with a catalytically hardening silicone rubber,
Unitek Xantropren Blue dental impression material. Approximately one
quarter the amount of hardener recommended by the manufacturers was
used so as to obtain a lower viscosity mix with reasonable setting times.
The rubber was placed in the notch with the specimen unloaded. The
specimen was then loaded in three-point bend to a particular clip gage
displacement. The specimen was held at this clip gage displacement for
about 45 rain until the rubber had set. Bending was then continued until
the specimen was broken. The rubber "casting" normally remained on
one or other half of the specimen and could be examined in an optical
microscope to determine whether the rubber had penetrated to the tip of
the fatigue crack. "Castings" which had not penetrated to within 0.05 mm
of the tip of the fatigue crack were discarded. The rubber "casting" was
then removed, sectioned at the midsection, and examined in the scanning
electron microscope. The COD was measured at the point nearest the
crack tip where the crack faces were essentially undeformed (Fig. 1).
The microscope was calibrated using calibration grids of known dimen-
sions and the magnification factor could be determined to about •
percent. This technique allows on-load values of crack tip COD at the
specimen midsection to be measured directly. Values of the effective center
of rotation can be calculated from these COD measurements and the clip
gage displacement (CGD). This, then, allows the COD to be obtained
ROBINSON AND TETELMAN ON CRACK TIP DISPLACEMENT 143
FIG. 1--Rubber "casting" of crack tip in 4340 steel sectioned at the center of the
specimen.
directly from clip gage displacement without further use of the rubber
castings.
A "casting" was made of a scale of known length and measured to
determine whether any shrinkage or swelling occurs in the rubber upon
or after setting. The rubber was found to shrink very slightly with time,
0.5 percent after 250 h. Since all crack tip "castings" were measured
within 24 h, shrinkage is negligible.
Measurement of COD at Fracture Initiation

Double-Bend Technique The specimen was first bent to failure at the
required temperature and the fracture surface examined in the scanning
electron microscope to determine the mode of fracture initiation. For
those conditions where fracture initiation was by ductile rupture, speci-
mens were deformed in slow bending to a variety of clip gage displace-
ments at the required temperature and then fractured in impact at liquid
nitrogen temperature. Examination of the fracture surfaces in the scanning
electron microscope showed whether any ductile rupture was present,
which would indicate that fracture initiation, taken as corresponding to

0.05 mm crack growth (approximately 2 percent crack growth in the
precracked Charpy specimens used), had occurred in the first bending
operation. For those temperatures where fracture initiation occurred by
cleavage, specimens were deformed to a variety of clip gage displacements
and then ruptured at a temperature where fracture is entirely ductile. Any
cleavage present on the fracture surface then indicated that initiation had
occurred in the first bending operation. If the material did not undergo a
ductile-brittle transition then bending followed by further fatiguing was
used to detect fracture initiation. When examining the fracture surface,
care was taken to differentiate between the stretch zone and the region
of crack growth. Similar techniques have been described in the litera-
ture[6, 7]. The results from three or four specimens allowed a graph of
COD versus amount of crack growth to be plotted and this was then
used to determine the COD at fracture initiation.
Crack Profile Technique--Examination of silicone rubber castings from
specimens in which some crack propagation had occurred indicated that
the crack grew by the process shown in Fig. 2a. The total crack opening
displacement at the original fatigue crack tip (PQ in Fig. 2a) after crack
growth is made up of two parts, (1) the critical (COD), associated with
fracture initiation and (2) the opening of the propagated crack near the
original crack tip (RS in Fig. 2a). Thus, (COD)c=PQ-RS. The sole
difficulty with this method lies in assigning a value to the opening, RS,
of the propagated crack since this crack normally has fairly rough sides.
Satisfactory results were obtained by making the measurements on the
two sections at the nearest point to the original crack tip, where that
section of the. crack had essentially parallel sides. The indistinct region
between these points was ignored. This technique requires only one
specimen to obtain a value of (COD),, but may be expected to be less
accurate than the double bend technique since it involves an element of
judgment in measuring RS.
To obtain the best castings, the rubber should be placed in the notch
with the specimen unloaded and the specimen then bent up to, or slightly
before, maximum load. The specimen is held under load until the rubber
has set and then broken away from the casting. The casting is sectioned
at its midsection and examined in the scanning electron microscope, Fig.
2b. Castings taken after very extensive crack growth are difficult to
interpret.
Determination o] Stress-Strain Relationships

The 0.2 percent yield stress for 4340 steel was determined from tensile
tests. The tensile yield stress for A533B at the various temperatures was
o)4
(a)
(a) Schematic representation of crack growth from initial precrack. (COD)~-~PQ-RS.

(b) Rubber "casting" of crack in 4340 steel after some crack propagation has oc-
curred.
FIG. 2--Crack profile technique [or the determination o[ critical COD.
145
obtained from Witt[8]. In all other cases the 0.2 percent yield stress and
true stress-true strain curves were determined from uniaxial compression
of cylinders 13 mm in diameter and 13 mm in height using a fluorocarbon
lubricant.

Transverse Strain Measurement
Transverse strain was measured on a sharpy notched A533B Charpy V
specimen (10-ram thick, 0.05-mm root radius). The initial position of
the line center could be determined to an accuracy of approximately
_+0.005 mm, and, after considerable bending, to an accuracy of ap-
proximately • ram. This results in an accuracy for the displace-
ment between two lines of -+0.036 mm. Differences between initial and
final displacements of greater than 0.036 mm were taken as indicative
that some strain had occurred. Bending was continued up to fracture
initiation. The length of the central region of zero strain, as a function of
0, is shown in Fig. 3a. Since the smallest such region is about 3.3 mm
long, transverse strains less than about 1 percent are below the resolution
of the technique (for comparison, longitudinal strains at the specimen mid-
section reach values up to 33 percent[9]). Fig. 3a shows that plane strain
conditions are maintained in the central third of the specimen up to a
bend angle of at least 0 : 4 deg. All (COD)c values in the present work
were obtained from specimens with 0 ~< 4 deg and were always measured
at the midsection and thus well within the plane strain region.
Figure 3b shows grid lines on a standard Charpy specimen (0.254 mm
root radius) bent to an angle 0---20 deg. The lines on this specimen were
purposely made coarse so as to be clear on the photograph.
The Relationship Between COD and Stress Intensity Factor

The value of COD corresponding to a particular K value was determined
using standard ASTM fracture toughness specimens. The materials used
were 4340 steel, quenched and tempered to yield strengths of 1.20 and
1.31 G N / m 2, 6AE-4V titanium alloy, and 2 0 2 4 - T 4 aluminum. To ensure
that residual stresses from the fatiguing would be small, final fatigue
stress intensities for the steel and titanium specimens were approximately
one quarter of the stress intensity at which COD was to be measured.
This was not possible with the aluminum specimens since the fatigue
crack would not propagate at these low stress intensities. These latter
specimens were thus solution treated and re-aged after fatiguing to remove
residual stresses. The steel and titanium specimens were loaded up to K
values less than K~c and within the ASTM validity requirements, and the
(a) Results from an A533B steel Charpy specimen, 0.05-mm notch root radius.
(b) Coarse grid lines across notch of mild steel standard Charpy specimen, 0 : 2 0 ~
FIG. 3 - - T r a n s v e r s e strain m e a s u r e m e n t s .
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authori
C O D was measured by the rubber infiltration method. The aluminum

specimens were loaded to higher K values than permissible under the
ASTM standard since so much fatigue crack tunnelling occurred in
"valid" size specimens that the rubber failed to penetrate the majority of
the crack. The results of these studies are given in Fig. 4.
The titanium specimen was loaded up to two thirds of K~. The highest
points for the two heat treatments of 4340 steel are close to K~ and some
crack growth may be expected to have occurred. Two specimens of 4340
steel with 1.2 G N / m z yield stress were loaded up to K values correspond-
ing to the two lowest C O D values on Fig. 4 (for this steel) and the
amount of crack growth determined using the double bend technique.
It is difficult to detect very small amounts of crack growth, but in neither
case was there more than 0.015 mm. Such very small amounts of crack
growth would not be expected to significantly affect COD values[6].
From linear elastic fracture mechanics,
r-i 2024-T4 ALUMINUM, oy = 310MN/m2, W = 38 mm

0.075 O 4340STEEL, Oy= 1200MN/m2, W = 51 mm
4340 STEEL, oy = 1310MN/m2, W = 25 mm

n /
Q Ti-6A~-4V, WROUGHT, oy = 830MN/m2, W ~
E
E
0.05
o/o
g
0.025
I I
0.025 0.05
K2(I-~,2)
- - mm
EOy
FIG. 4--Correlation between K and COD.
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc
KS= - EG (1)
-- V2
From theoretical treatments [1, 2, 10-16]

G = X,~uCOD (2)
where ,ru is the yield stress and ;~ is a constant. From Eqs 1 and 2,
K s (1--v 2)
COD- (3)
,r~,,\E
The data of Fig. 4 confirms the relationship in Eq 3 between COD and
K2/o-u. A least squares fit to the data of Fig. 4 gives an experimental value
of ;t equal to 0.97. The measurement of COD values is accurate to within
approximately ___20 percent at COD=0.025 mm and + 1 0 percent at
C O D : 0 . 0 5 mm.
Space does not permit more than a very brief review of the various
theoretical analyses. Calculations by Burdekin and Stone[10] (if any
constraint factor on the yield stress is ignored), Bilby et ai[11], and the
deformation plasticity analysis of Rice and Rosengren[12], all yield results
identical with experiment, X= 1. Wells'[2] analysis, X : r / 4 , and the finite
element analysis of Sumpter et al[13] for single edge notch tension,
x=1.155, are also in close agreement with the experiment. Deformation
plasticity analysis by Rice[/] and Rice and Johnson[14] agree less, giving
;t values of 1.27 and 1.48. The theoretical treatments that might be
expected to be most closely applicable, finite element analyses by Hayes
and Turner[15] for three-point bend and by Levy et al[16] show least
agreement with experiment giving A values of 2 and 2.14, respectively.
Where necessary, a Poisson's ratio of 0.3 has been used to calculate the
prior A values.
Possible explanations for disagreements between theory and experiment
are:
1. differences in definition of COD,
2. possible nonidentity of G and the J-integral after large scale yielding
(this would affect both the deformation plasticity and the finite element
analyses),
3. difficulties in theoretically modelling the region near to the crack
tip. For example, no analysis includes representation of void formation
just ahead of the crack tip, and
4. the experiment compares a midsection value of COD with a through
the thickness average value of K. Comparison with a value of K at the
position of measurement of COD might be expected to increase ~ by a
factor of the order of 15 percent (see Schroedl and Smith[17]).
Rearranging Eq 3 and using X-- 1 for simplicity, we find
Kxc= V E o-~(COD)c
1 -v ~ (4)
Since Eqs 2 and 4 contain the uniaxial yield stress (~u), it might be
expected that the relation between K and COD would be affected by strain
hardening. That this is not so is evident from Fig. 4 since 4340 steel with
a 1.31 G N / m ~ yield strength has a very low strain hardening exponent,
n=0.033 in Eq 5, while 2024-T4 has a fairly high value of n=0.205.
~=C~" (5)
An explanation for this independence of strain hardening is provided

by considering the derivation of Eq 2. Since no analytical solution is
available for post-general yield conditions, we will consider only contained
yielding. Following Wells[2], and McClintock and Irwin[18], Eq 2 can be
obtained by considering COD to be the integration of the shear strain
around the elastic-plastic boundary.
f ~r}2
COD = Yu Rd~
d -rr/2
where 7u is the shear yield strain and R is the radius of the plastic zone
at angle 4~. The value of COD thus depends only on the shear yield strain
and the size and shape of the plastic zone. Since the shear strain at the
boundary is that for initial yield (Yu), strain hardening can only affect
Eq 2 through altering the dimension of the plastic zone. Rice and Rosen-
gren[12] have calculated, numerically, the relationship between strain
hardening exponent m, in Eq 6, plastic zone size, and COD using
~" ( Y y '
- - = -- (6)
~'u \ Yu /
as the stress-strain relationship inside the plastic zone, where r is the
shear flow stress at shear strain ~, and T~ and 7u are the values for initial
field. Most real materials have very small values of m, for example,
A533B, with an average value of n (Eq 5) to fracture initiation of about
0.2 has an average of m of about 0.08. This results in small changes in
plastic zone size with strain hardening exponent for real materials. Rice
and Rosengren's[12] numerically calculated values of COD are very
insensitive to hardening for m<0.2. The results shown in Fig. 4 thus
correspond to their calculations.
The Relationship Between COD and On-Load Clip Gage Displacement

Below room temperature the increase in viscosity of the silicone rubber
complicates the use of the infiltration technique for measurement of COD.
ROBINSON AND TETELMAN ON CRACK TiP DISPLACEMENT 1,51
Furthermore, it is inconvenient to measure COD in this way while using

the double bend or electrical potential fracture initiation detection tech-
niques. To circumvent these problems, measurements of the position of
the apparent center of rotation at particular values of COD were made.
The rotational constant, r, the depth below the crack tip at which the
apparent center of rotation is located expressed as a fraction of liga-
ment depth, can be calculated from the relationship[19].
r: aq-z (7)
(W--a)I~D 11
where W is the specimen width, a is the crack depth, and z is the height of
the knife edges above the specimen surface. The derivation of Eq 7 is shown
schematically in Fig. 5. The values of rotational constant refer only to the
knife edges on the top surface of the specimen and the crack tip at the
specimen midsection. Reference to other points in the specimen may be
expected to produce different apparent centers of rotation.
A curve of r versus COD is given in Fig. 6. As can be seen, r is a
single-valued function of COD independent of material (aluminum and
titanium alloys and steel) and of specimen geometry (within the limits
that this has been varied) for three-point bend specimens. This is im-
portant since it means that it is then possible to calculate midsection
COD from measurements of on-load CGD and specimen geometry only.
This curve can be used by all investigators, for specimens of similar
geometry, without having to repeat the infiltration measurements we have
made.
Z KNIFE EDGE
11.--,o.,c.--
rIW-a) FATIGUE CRACK
\ \ - A FARE. C E . E R
ATION
FIG. 5--Schematicshowingcalculationof rotationalconstant.
FIG. 6--Position of center of rotation versus COD.
The scatter band in r in Fig. 6 corresponds to the inaccuracy in

determination of COD shown in Table 2. Table 2 also shows the uncer-
tainty in measurement of K using Eq 3. Error in determining K from
COD values arises from: (1) uncertainty in COD values, (2) uncer-
tainty in the value of ~, and (3) uncertainty in measuring mechanical
properties, E, ~ and v. Error from (1) is shown as the third line of
Table 2. Uncertainty in ~ is more difficult to estimate. The generally close
agreement between the points on Fig. 4 and the line for Eq 4 indicates
that errors are not high. An estimate of ~ - - 1 . 0 + 5 percent for the defini-
tion of COD used herein is reasonable. A few percent inaccuracy in v
will have a negligible effect on K since vz ( ~ 0.1) is small and thus
( 1 - v z) is not heavily dependent on the value of v. A few percent in-
accuracy may be expected in E and ~y, resulting in an approximately
___5 percent error band in the product E~y. The product E~;~ is then
accurate to within ___10 percent. The resulting error band in K is shown
as the fourth line of Table 2. The fifth line shows the approximate total
error in K.
In the next section, the use of Eq 4 to determine K~e from (COD)e is
described. Comparison of calculated KI~ values with those determined on
A S T M standard specimens indicates that the size of the error bands in
Table 2 is overestimated.
TABLE 2--Error in determination of COD and in the corresponding

calculated values of K.
COD, mm 0.025 0.05 0.075 0.1 0.125 0.15 0.175 0.2 0.225
Total error band for 35 25 21 20 14 9 8 8 8
COD, %
Corresponding error ~9 • • • • • • • •
in stress intensity
factor K (from Eq
3),%
Error in K from • • • • • • • • •
other inaccuracies,
%
Approximate total • • • • • • ~7 ~7 •
error in K, %
The method is thus most suited for materials for which the critical C O D
is greater than 0.075 mm but can be used down to ( C O D ) c values of
about 0.03 mm. As an example, a steel with ( C O D ) c = O . 0 3 mm and
% = 500 M N / m 2, has, from Eq 4, a Kie of 59 M N / m ~/2. An aluminum alloy
with the same ( C O D ) c and % = 3 0 0 M N / m 2 has a K~e of 26 M N / m 3/2.
A least squares fit to the r / C O D curve is given by Eq 8 for C O D
between 0.025 mm and 0.25 mm, as shown in Fig. 6.
r = Ao + A 1 C O D + A 2COD2 + A 3 C O D "~ (8)
where C O D is in millimeters,
Ao= 0.03684,
AI= 3.899,
A2= - 16.34, and
A~ = 27.24.
Combining Eqs 7 and 8 gives
Bo + B I C O D + B z C O D 2+ B3COD ~+ B4COD 4 = 0 (9)
where C O D and C G D are in millimeters
Bo = A oCGD,
n~ = A ~ C G D -
(\W--~-a
a+z ] - Ao
B2 = A 2 C G D - A 1
B3 = A 3 C G D - A 2
B4= --A.~
Equation 9 can be solved using standard computer programs thereby
giving on-load, plane strain C O D values directly from measurements of
on-load clip gage displacement and specimen geometry only.
For small-scale yielding, the clip gage displacement is directly pro-

portional to the load, and thus to the stress intensity factor.
CGD=~K (10)
where ~ is a constant. Combining Eqs 4 and 10,
CGD 2( 1 - v~)
COD = (11)
~.2 E~ry
Combining Eqs 7 and 11,
~2 E,r~
COD=-(1 --v=) {1-+ r~)(a+z) -J\2 (12)
As COD .-+ 0, r must go to zero since all other parameters in Eq 12 are

fixed for a particular material and specimen geometry. This trend is evi-
dent in Fig. 6. At high values of COD, the value of r is close to the
value of about 0.5 predicted from slip line theory[20].
The Determination of KI~ from (COD)c
(COD)c was determined on precracked Charpy specimens of A533B

at - 5 5 , - 3 1 , 0, 24, and 60 ~ C using the double-bend technique to detect
the clip gage displacement and hence the COD at crack initiation. At
0 ~ C and above, initiation occurred by ductile rupture. At - 5 5 and
- 3 1 ~ C, initiation occurred by cleavage. In all cases initiation occurred
prior to maximum load, at room temperature considerably before, and
at - 5 5 ~ C very shortly before, maximum load. The double-bend tech-
nique has the disadvantage that three or four specimens are required to
determine (COD)c. The crack profile technique, which only requires one
specimen, was also used to measure (COD)c for A533B at room tem-
perature. Values of Kic for A533B at the various temperatures were cal-
culated from these (COD)~ values using Eq 4. The results are shown in
Fig. 7, superimposed on the scatter band for HSST "valid" Kie results[8]
from specimens up to 305-mm thick. The values can be seen to have
reached the upper shelf. It is noted that the agreement between K~e values
obtained from (COD)~ measurements and valid ASTM tests is good.
Precracked Charpy specimens of quenched and tempered 4340 steel
were also used to determined (COD)c and hence KI~ for a variety of
yield strengths using both the double bend and crack profile technique.
The results are shown in Fig. 8, together with measurements from valid
K~ tests. Once again, the agreement is good and both Figs. 7 and 8 indi-
INITIATION DETECTION TECHNIQUE

9 DOUBLE BEND
O CRACK PROFILE
200 r-
o9
E
loo
I I I I I I
-150 -100 -50 0 50 100
TEMPERATURE OEG C
FIG. 7--KIo calculated ]rom (COD)c ]or A533B superimposed on the scatter
band for HSST valid Kg, results [8] (shaded area).
cate that the two techniques for determining the critical value of COD
give very similar results.
It is of interest to note that, for a~<l.4 GN/m 2, ( C O D ) c increases
linearly with decreasing yield stress. The K~c, however, changes much
less with yield stress, particularly for ay<0.85 GN/m'-'. This results from
the increases in ( C O D ) c being balanced by decreases in ~, (see Eq 4).
At the other end of the scale, where (COD),. <0.03 mm and thus diffi-
cult to measure accurately, ~,> 1.28 GN/m'-' and K~e<94 MN/m 3/2 and
fracture occurs before general yield. The thickness of a Charpy specimen
(10 m m ) > l . 8 5 (K~J,~y) ~ We believe that this thickness is adequate to
ensure reasonably valid measurements of Kx~ in this regime. Thus, the
precracked Charpy specimen is able to provide an inexpensive measure-
ment of K~ over the entire range of toughness for 4340 steel.
The accuracv of the double-bend initiation technique may be expected to
I O (COD)9 AND KIC MEASURED USING DOUBLE BEND

INITIATION DETECTIONTECHNIQUE
"8" o ~ [] (COD)c AND Irlc MEASUREDUSINGCRACK PROFILE
TECHNIQUE
120
Z~ Kic DETERMINED BY ASTM STANDARD METHODS
o
z~ 9 KIC FROM THE LITERATURE'u,':~
o~:OD) c CALCULATED FROM VALID KIC DATA USING EQ (4)
80
=Z
o.o
40
I I I I 0
0.8 1.0 1.2 1.4 1.6 1.8
YIELD STRESS GN/m2
FIG. 8--Variation of (COD)c and K1~ with yieM stress for quenched and tem-
pered 4340 steel.
about _+ 15 percent for values of (COD),, down to about 0.03 mm. For
( C O D ) ~ > 0 . 0 7 5 mm, the scatter is similar to that found in standard Kic
tests (typically • 5 to 10 percent).
Summary
A method is described whereby on-load values of COD can be mea-
sured at the midsection of precracked three-point bend specimens by infil-
tration of silicone rubber. A calibration curve relating midsection COD to
clip gage displacement was derived from measurements on the silicone
rubber "castings." This calibration curve can be used to calculate mid-
section COD from on-load C G D and specimen geometry only, without
further infiltration measurements. These values of COD have been shown
to be simply related to the stress intensity factor, independent of material,
as theoretically predicted.
The central region of a Charpy specimen in three-point bend has been
found to remain in plane strain until well after general yield. Thus, plane-
strain values of ( C O D ) c can be determir~ed from small specimens. Two
initiation detection methods are described whereby ( C O D ) c can be deter-
mined. These values of (COD)~ can be used to accurately predict K~
University of Washington (University of Washington) pursuant to License Agreement. No further reprod
ROBINSON AND TETELMAN O N CRACK TIP DISPLACEMENT ]57
values u n d e r conditions w h e r e a valid A S T M K ~ s p e c i m e n w o u l d be m o r e

t h a n an o r d e r of m a g n i t u d e l a r g e r a n d one to two orders of m a g n i t u d e
m o r e expensive.
Acknowledgments
W e w o u l d like to t h a n k the U.S. A r m y R e s e a r c h O f f i c e - D u r h a m for
financial s u p p o r t u n d e r c o n t r a c t D A H C 0 4 - 6 9 - C - 0 0 0 8 . W e are also in-
d e b t e d to S.A. M o h a m e d , S. E n s h a , a n d I. R o m a n for their helpful advice
a n d to P.N. R a n d a l l a n d J. M e r k l e of the H S S T p r o g r a m for supplying
the A 5 3 3 B steel.
References
[1] Rice, J. R., Transactions of the American Society of Mechanical Engineers,
lournal oJ Applied Mechanics, Vol. 35, t968, pp. 379-386.
[2] WeUs, A. A., Proceedings of the Crack Propagation Symposium, The College of
Aeronautics, Cranfield, England, Vol. 1, 1961, pp. 210-230.
[3] Goodier, J. N. and Field, A. A., in Fracture o/Solids, D. C. Drucker and J. J.
Gilman, Eds., Interscience, New York, 1963, pp. 103-118.
[4] Begley J. A. and Landes, J. D. in Fracture Toughness, Proceedings of the 1971
National Symposium on Fracture Mechanics, Part H, ASTM STP 514, Ameri-
can Society for Testing and Materials, 1972, pp. 1-20.
[5] Robinson, J. N. and Tetelman, A. S., "The Determination of Kxo Values from
Measurements of the Critical Crack Tip Opening Displacement at Fracture Ini-
tiation," presented at the Third International Conference on Fracture, Munich,
Germany, 1973.
[6] Smith, R. F. and Knott, J. F., "Crack Opening Displacement and Fibrous Frac-
ture in Mild Steel," Con/erence on Practical Application of Fracture Mechanics
to Pressure-Vessel Technology, The Institution of Mechanical Engineers, Lon-
don, 1971, pp. 65-75.
[7] Steigerwald, E. A. and Hanna, G. L., Proceedings, American Society for Testing
and Materials, Vol. 62, 1962, pp. 885-913.
[8] Witt, F., "Heavy-Section Steel Technology Program Semiannual Progress Re-
port for Period Ending August 31, 1969, "ONRL Report 4512, Oak Ridge Na-
tional Laboratory, Oak Ridge, Tenn. 1970.
[9] Robinson, J. N. and Tetelman, A. S., "The Relationship between Crack Tip
Opening Displacement, Local Strain and Specimen Geometry," to be submitted
to International Journal of Fracture.
[10] Burdekin, F. M. and Stone, D. E. W., Journal o/Strain Analysis, Vol. 1, No. 2,
1966, pp. 145-153.
[11] Bilby, B. A., Cottrell, A. H., and Swinden, K. H., Proceedings o/ the Royal
Society (London), Series A, Vol. 272, 1963, pp. 304-314.
[12] Rice, J. R. and Rosengren, G. F., lournal of the Mechanics and Physics of Sol-
ids, Vol. 16, 1968, pp. 1-12.
[13] Sumpter, J. G., Hayes, D. J., Jones, G. T., Parsons, C. A., and Turner, C. E.,
"Post Yield Analysis and Fracture in Notch Tension Pieces," Paper 1-433, pre-
sented at the Third International Conference on Fracture, Munich, Germany,
April 1973.
[14] Rice, J. R. and Johnson, M. A. in Inelastic Behavior o/Solids, M. F. Kanninen,
Ed., McGraw-Hill, New York, 1970, pp. 641-672.
[15] Hayes, D. J. and Turner, C. E., "An Application of Finite Element Techniques
to Post-Yield Analysis of a Practical Fracture Test Specimen", presented at the
Institute of Physics Conference, Leicester, England, 1971 (to be published in

International Journal o[ Fracture).
[16] Levy, N., Marcal, P. V., Ostergren, W. J., and Rice, J. R., International Journal
o[ Fracture Mechanics, Vol. 7, No. 2, 1971, pp. 143-156.
[17] Schroedl, M. A. and Smith, C. W., "Influence of Three Dimensional Effects on
the Stress Intensity Factor for Compact Tension Specimens", presented at this
symposium.
[18] McClintock, F. A. and Irwin, G. R. in Fracture Toughness Testing and its Ap-
plications, A S T M STP 381, American Society for Testing and Materials and
National Aeronautics and Space Administration, 1965, pp. 84-113.
[19] Nichols, R. W., Burdekin, F. M., Cowan, A., Elliott, D., and Ingham, T., "The
Use of Critical Crack Opening Displacement Techniques for the Selection of
Fracture Resistant Materials," First Report of the Crack Opening Displacement
Application Panel of the Navy Department Advisory Committee on Structural
Steel, United Kingdom Atomic Energy Authority, Reactor Materials Laboratory,
Culcheth, England, 1968.
[201 Green, A. P. and Hundy, B. B., Journal of the Mechanics and Physics o[ Solids,
Vol. 4, 1956, pp. 128-144.
[211 Robinson, J. N. and Tuck, C. W., Engineering Fracture Mechanics, Vol. 4,
1972, pp. 377-392.
[22] Pellissier, G. E., Engineering Fracture Mechanics, Vol. 1, No. 1, 1968, pp. 55-75.
Saurindranath Majumdar 1 and JoDean Morrow 1
Correlation Between Fatigue Crack

Propagation and Low Cycle
Fatigue Properties
REFERENCE: Majumdar, Saurindranath and Morrow, JoDean, "Correla-

tion Between Fatigue Crack Propagation and Low Cycle Fatigue Properties,"
Fracture Toughness and Slow-Stable Cracking, A S T M STP 559, American
ABSTRACT: Fatigue crack propagation is treated as a succession of fatigue
crack initiation events. The cyclic stress and strain experienced by elements
ahead of the crack tip are computed by means of an elastic-plastic analysis
and the cumulative fatigue damage is evaluated from low cycle fatigue prop-
erties of the metal. If it is assumed that the appropriate element width is re-
lated to the microstructure size of the metal, a significant deviation from the
log-log linear relationship usually assumed to exist between the crack propa-
gation rate, da/dN, and the stress intensity range, AK, is predicted. The
analysis indicates that at small values of AK, da/dN rapidly decreases until
it approaches zero at a "threshold" value of AK.
Fatigue crack propagation data for eight steels from Barsom's work are
compared with predictions from the analysis using low cycle fatigue prop-
erties and values of the microstructure size based on Gurland's correlation
between yield strength and the dimensions of the microstructural units re-
sponsible for strengthening in steels. In all cases the estimate of da/dN for
a given AK was within a factor of two of the measured values.
KEY WORDS: fatigue (materials), crack propagation, stresses, strains,
damage, microstructure, mechanical properties.
Nomenclature
a Semi-crack length
b, c Fatigue strength and ductility exponents
cr Modified fatigue ductility exponent
n, n r Strain hardening exponent, cyclic value
1 Assistant professor and professor, respectively, Department of Theoretical and
Applied Mechanics, University of Illinois at Urbana-Champaign, Ill. 61801.
159
Copyright9 1974
Downloaded/printed byby ASTM lntcrnational Www.astIII.OI'g
P Fatigue crack propagation exponent

x, y Coordinates from the crack tip
A Fatigue crack propagation coefficient
COD Crack opening displacement at maximum load
E,G Young's modulus and shear modulus
Kc Mode I critical stress intensity factor
N, N I, Nt Cycles, cycles to failure, transition fatigue life, respectively
Rp Reversed plastic zone size
A Mean free path as defined by Gurland
p* Microstructure size related to X
Normal stress and normal strain
Crft , C7 Fatigue strength and ductility coefficients
~rff Monotonic (static) yield strength
~yP ~yt Cyclic uniaxial yield strength and yield strain
r, 7 Shear stress and shear strain
TOy ~10 Shear yield strength and yield strain
• Stress intensity factor range (defined as AS~/,m)
• Threshold stress intensity factor range
• Nominal stress range
Aep, Ace Plastic strain range and elastic strain range
A(r, Ac Normal stress range and normal strain range
AT, A 7 Shear stress range and shear strain range
The problem of Mode I fatigue crack propagation has been the sub-
ject of extensive study over the past decade. In 1961 Paris, Gomez, and
Anderson[/] 2 suggested a correlation between the fatigue crack propaga-
tion rate and the range of the applied stress intensity factor. Experimental
results are now generally presented in a form similar to Fig. 1 which is a
schematic log-log plot of the crack propagation rate, da/dN, as a function
of the stress intensity factor range, AK, (defined as AS~/,m). Most
metals appear to exhibit a "threshold" value of AK, below which fatigue
cracks do not measurably propagate. Above the threshold value there is
usually an intermediate range of AK over which da/dN is nearly linearly
related to AK on a log-log basis. At higher ranges, the crack propagation
rate increases rapidly as the maximum stress intensity approaches the
critical value of Kc. The crack propagation rate in the intermediate zone is
related to the range of stress intensity by the following equation
da = A (• (1)
dN
paper.
MAJUMDAR AND MORROW ON FATIGUE CRACK PROPAGATION 161
s
AS
tttttt
time
&K = & S ~
do_=~~ 14I
d'N- I
/
I
i
f
AKth Kc
IogAK
FIG. 1--Definition oJ terms and typical crack propagation data.
where A and p are experimentally determined constants.

A variety of tests[2-6] have shown that for most metals the exponent
p in Eq 1 usually varies between two and four, though values as large as
eight have occasionally been reported.
Using continuum mechanics variables in a dimensional analysis, Liu[7]
has shown that the exponent p in Eq 1 should be two. The experimentally
observed deviation of p from two can be accommodated within the frame-
work of continuum mechanics by permitting the coefficient, A , in Eq 1
to be a function of AK. This requires the introduction of a microstructural
"length parameter." In this paper a dimensionally compatible relation-
ship is proposed between /~K and d a / d N which has been derived by
analyzing a simple fatigue crack propagation model similar to that pro-
posed by Liu and Iino[8].
There are practical as well as theoretical reasons for attempting to
relate fatigue crack propagation to 'other material properties via the
analysis of a model such as used here. A significant problem facing
materials engineers is the selection of materials to resist fatigue crack
propagation at intermediate and low stresses. Few guidelines are currently
available for choosing the most crack resistant metal from a group of
candidates short of conducting fatigue crack propagation tests. From
University of Washington (University of Washington) pursuant to License Agreement. No further reprodu
this point of view, it would be desirable to relate A and p in Eq 1 to other

known mechanical properties of metals.
Review of Fatigue Crack Propagation Models
Several investigators have attempted to correlate fatigue crack propa-
gation with other mechanical properties. For example, Tomkins[9] has
analyzed fatigue crack propagation using the plastic cohesive stress
approach of Dugdale[10] and concluded that fatigue crack growth rate is
basically dependent upon the cyclic stress-strain characteristics of materials.
Hickerson and Hertzberg[ll] have considered the correlation of fatigue
crack growth resistance of materials with other mechanical properties and
found the cyclic strain hardening exponent, cyclic yield strength, and the
elastic modulus to be important parameters.
A number of continuum mechanics models of fatigue crack propagation
have been analyzed in the past. For example, McClintock[12] proposed a
fatigue crack propagation model on the basis that fracture at the crack
tip occurs whenever the accumulated average plastic strain reaches a
critical value in a small sector of material just ahead of the crack tip.
The radius of the sector which McClintock called the structure size, was
interpreted as the smallest region to which a macroscopic fracture cri-
terion can be applied (he used a value of 2• 10-4 in.). From his damage
analysis, McClintock concluded that the crack propagation rate is pro-
portional to the square of the plastic zone size.
Rice[13] has considered fatigue crack growth based on the plasticity
model of slip ahead of the crack. Tracing the deformation history of a
particular point after it is encompassed by the plastic zone until the crack
tip advances to the point, separation was assumed to occur when the total
absorbed hysteresis energy equalled a postulated critical value per unit
area. He obtained a relation similar to Eq 1 with p = 4 .
Using Neuber's rule[14], Weiss[15] derived a crack propagation law
under the assumption that the incremental growth per cycle is equal to
the distance over which the strain ahead of the crack tip exceeds some
critical value. The analysis of Weiss included a microstructure length
constant, 0", as introduced by Neuber to analyze the fatigue behavior of
sharp notches.
Liu and Iino[8] proposed a model in which it is assumed that the points
ahead of the crack tip constitute a set of uniaxial fatigue specimens. The
propagation of a fatigue crack is assumed to be due to the successive
fatigue failure of each of these imaginary specimens. By postulating equal
crack advance per cycle and experimentally measuring the plastic zone
size and the strain distribution ahead of the crack tip, they derived an
expression similar to Eq 1 with p = 2.
Fleck and Anderson[16] have analyzed a model similar to that used by

Liu and Iino[8]. They employed Rice's[13] discrete surface of slip or
tensile yielding analysis (for a perfectly plastic material) to estimate the
plastic zone size ahead of the crack tip. It was postulated that a fatigue
element fails when the crack has progressed halfway through the element.
By suitably choosing the uniaxial material properties and the element
size, Fleck and Anderson[17] obtained correlation between experimental
data and their analysis for various metals.
Scope
The present paper adopts Liu's model, but instead of experimentally
determining the plastic zone size and the strain distribution ahead of the
crack, the analytical results obtained by Rice[18] are employed. In its
simplest form, the analysis results in the exponent, p, of Eq 1 being
equal to two. The coefficient, A, can be calculated from the usual cyclic
stress-strain and fatigue properties of uniaxial specimens of a given metal.
Further, by introducing a microstructure size, 0", into the analysis it is
shown that a nonlinear relationship between log AK and log da/dN may
be obtained which quantitatively agrees with the trends in experimental
data. The nonlinearity is achieved by keeping the power, p, equal to two,
while the coefficient, A, becomes a function of AK. Thus, the dimension-
ality of Eq 1 is maintained.
Description of the Fatigue Crack Propagation Model

Figure 2 shows the model to be analyzed and a schematic representa-
tion of the stress and strain history experienced by an element ahead of a
propagating crack in a plate subjected to repeated tension so that AK is
constant. Figure 3a depicts an essentially undamaged element ab at the
reversed plastic zone boundary. Figure 3b shows the same element with
the crack at a distance x from it. The damage experienced by element ab
in one cycle at this location will be the same as experienced by element
cd in Fig. 3a (assuming AK is constant). Thus, the cumulative damage
required to cause fatigue failure of an element is equivalent to the sum
of the damage experienced during one cycle by each element lying within
the reversed plastice zone. Let N1(x ) be the fatigue life of a specimen of
the material subjected to constant cyclic strain equal to that which is
present at a distance x from the crack tip. Using linear fatigue damage
summation, failure of the element will occur when
1 :R~o-Nf(x ) (2)
University of Washington (University of Washington) pursuant to License Agreement. No further reprodu
"fi o f/NIl, AA
I VVVVV
laJ
~ Fati?ueElements
I.O
Distance of Element from Crack Tip
FIG. 2--Schematic of stress-strain history o[ a ]atigue element ahead oJ the crack

tip.
Assuming that AK is constant the crack advances in each cycle a

constant amount, AX= da/dN. Multiplying the left side of Eq 2 by da/dN
and the right side by • and approximating the sum by an integral as
illustrated in Fig. 3c, the following expression is obtained.
da _ ri n, 1
dN )o Nt(x----)dx (3)
Mechanics and Fatigue Analysis of the Crack Tip Region
Rice[t3, 18] considered a stationary crack loaded in antiplane shear

(Mode III) under small-scale yielding in a material with the following
shear stress-strain law
(a)
~ RP
II
Reversed Plastic Zone
(b)
~ Reversed Plastic Zone
Distance from Crock Tip

(c)
FIG. 3--Typical damage history of an element. (a) element at edge of reversed

plastic zone, (b) element at distance x from crack tip, and (c) damage history of an
element.
~=G7 for 7~<7o, r<To
(4)
'/ ~'~
T=ro -- fory>vo,~>To
\ 7o/
Assuming that the plastic strain components at each point within the
plastic zone remain proportional to each other, Rice derived the following
expressions for cyclic stress and strain ranges
~xr= 2to 4(l+n)rro"x

(5)
1
r AKIII 2 71+~
t,~ = 2,/0L 4 ( 1 ~ ro2XJ
A similar analysis for tensile loading (Mode I) is not available.
McClintock[12] has discussed the analogy between Mode III and
Mode I for the case where displacements parallel' to the crack are small
compared to those normal to the crack surface. In the present study this
analogy will be used to approximate the stress and strain ranges ahead
of a tensile crack in a material obeying the following cyclic stress-strain
law
cr= E~ for ~_<cu'
(6)
/ \
In terms of the product of stress and strain range, Eq 5 reduces to

AK z
• (7)
(l+n')~:Ex
The reversed plastic zone size, Rp, and the crack opening displacement
at maximum load (Km,x=,xK), COD, are given by the following
1 AK 2
Rp=4(l+n,) rr~u'2
(8)
C O D = 2c,' ~K ~ (9)
Even though the preceding equations are derived for a stationary crack,
they will be assumed to hold for a propagating crack. The use of the fully
stabilized cyclic stress-strain law may be justified, provided the crack
propagation rate is slow enough that each point within the plastic zone
spends sufficient number of cycles at each strain range.
In order to compute the fatigue life of an element ahead of the crack
tip subjected to a known stress-strain range, the following power func-
tions will be used
• = 2cr/(2NI) b (10a)
• = 2c/(2Nr)c (10b)
Note that in Eq (10b) the fatigue life has been assumed to be related
to the strain range by a power function. In most materials, however, the
MAJUMDAR AND MORROW ON FATIGUECRACKPROPAGATION ]67
total strain range is related to the fatigue life by the following equation
(Fig. 4)
2 '
Ar : Ae~q- Ace= 2~/(2N0 c + ~ (2Nf) b ( 1 la)
Equation l la, in general, cannot be solved for the fatigue life ex-
plicitly and consequently a numerical method has to be adopted. On the
other hand, a closed form solution to the present problem may be
obtained if the following approximations are made. The majority of the
damage experienced by a fatigue element ahead of the crack occurs well
inside the reversed plastic zone[17] where the plastic strain range is
much larger than the elastic strain range. Consequently, in the low cycle
(high strain) regime of fatigue, the fatigue life for a given strain range
will lie between the following two power functions with slightly different
exponents, c and c', as defined in Fig. 4
A, ~ • c (llb)
ar ~-- 2~/(2NI) c' (llc)
The exponent c is the usual fatigue ductility exponent or the slope of
the log A~p versus log 2N r plot. A straight line on the log-log plot con-
necting the intercept at one reversal (the fatigue ductility coefficient, ~/)
and the strain range at the transition fatigue life will have a slope of c'.
,1__
INt =
_ log 2
,;(ZN,)
A~
log T
,',-(ZN, f
AE , e E ~ b
i
~~T:'f(2Nf)+ "~f(2N,)
Ef
E
1
2N t
log (2Nf)
FIG. 4---Approximating total strain amplitude versus life plots by power functions.
Note that Eq 1 lb underestimates and Eq 1 lc overestimates the fatigue

life in the relevant range of the curve in Fig. 4. Thus, the crack propa-
gation analysis based on Eqs 1 lb and 1 lc should bracket the results that
would be obtained if the unabridged Eq 1 la were used to compute fatigue
lives.
Multiplying Eqs 10a and 10b and equating the result to Eq 7 gives the
damage sustained in one cycle by an element at a distance x from the
crack tip.
--1 1
1 =2[ AK2 tlb+c (x) b+c
(12)
NI 4(1 + n ' ) 7rE r
A disturbing aspect of Eq 12 is that the integral in Eq 3 is divergent for
usual values of b and c. This singularity can be avoided by introducing
crack tip blunting. At any instant the crack tip will have a finite radius
that is approximately equal to half of the crack opening displacement. To
treat this problem rigorously, a full nonlinear analysis is needed which
unfortunately is not available. In the absence of such an analysis Rice's
results are modified by replacing x in Eq 12 by x + C O D / 2 , Substituting
the modified form of Eq 12 into Eq 3 and using Eqs 8 and 9 gives
--1
da _ - 2 ( b + c ) (r(
dN b+c+ 1 4 ( 1 + ~ (r/q'
-b+c+13 dxK2 (13)

1 -- [4( 1 + n') ,v'] ~ 71" O'j/t2
Since all terms in Eq 13 except d a / d N and AK are constants for a given

metal, it may be written in a form similar to Eq 1 with p = 2 as follows
da = A (AK) 2 (13a)
dN
The coefficient, A , is a function of the cyclic stress-strain and fatigue prop-
erties of the metal. In the derivation of Eq 13, ~v', *v', and n' have been
used to characterize the cyclic deformation resistance and (r/, ~/, b, and c,
the fatigue resistance of the metal.
Influence of Material Properties on the Coefficient of Eq 13

Equation 13 indicates that the exponent of AK in Eq 1 is two and the
coefficient A is a function of the cyclic stress-strain and fatigue properties
of materials. For any fixed AK, the A's are proportional to the rate of
fatigue crack propagation for different metals. Thus the quantitative
influence of these properties on fatigue crack propagation resistance may

be evaluated by calculating their effect on the coefficient A.
For making such an evaluation, reference values for c/, ~/, ~/, n', b, c,
and E shown in Fig. 5 were chosen and each property in turn was varied
keeping the others fixed.
The effects of the various properties on the coefficient A are shown in
dimensionless form in Fig. 5. Though the variation in one material prop-
erty without accompanying variation in the others is highly unrealistic,
these plots still serve to indicate how sensitive the coefficient A is to small
changes in material properties. This helps separate those properties that
need to be known accurately from those that need not be.
It is seen in Fig. 5 that the coefficient A is more sensitive to relative
changes in fatigue strength coefficient (,r/), fatigue ductility coefficient
and exponent (c/ and c) and elastic modulus (E) than to changes in
other material properties. These four properties, therefore, have to be
known more accurately than the rest for proper evaluation of fatigue crack
resistance.
Note that according to Eq 13, the cyclic yield strength (~r/) and the
cyclic strain hardening exponent (n') are of relatively minor importance
in determining fatigue crack propagation resistance.
Effect of Microstructure Size

The preceding analysis assumes that the material is a homogeneous
continuum and that the mechanical resistance of the infinitesimal ele-
Reference V a l u e s o f the Variables

IO,C I I I I I I II I I I I J Tlll-
Q)
"1
,'f --I n' = 0.15
Cf & d ,
":f ~ b =--0.09
~
n'/bY ~ 1 5=0 , k si
g , o-y 6 0 ksi c =-0.60
E: 30,000 ksi
cf
Aref = 2x I0-e ksi-z
0.1 I , ,l~,,,J I ~ , ,,,,
0.1 I.O IO.O

'&'/Are f
FIG. 5--Effect o[ material properties on coefficient A of Eq 13.
ments may be represented by the resistance of smooth macroscopic labora-

tory specimens. In reality, there is a minimum size or width of element to
which macroscopic failure criteria may be applied. One might think of
this size as being related to the grain diameter in pure metal or to the
dimensions of the microstructural units responsible for strengthening in
more complex materials.
In an effort to account for the effects of microstructure of a metal,
assume that the fatigue elements ahead of the crack tip have a width of
20" which is determined by the microstructure size of the metal. Further,
the fatigue element next to the crack tip is assumed to be subjected to a
constant damage per cycle that is equal to the value at a distance of 2p*
from the crack tip as computed from the continuum mechanics solution.
For low values of • more than one cycle will be required for each
fatigue element to fail and for large values of AK several elements will
fail in one cycle. In either case, an average value of the crack propaga-
tion rate (not the actual cycle by cycle da/dN) is obtained by changing
the lower limit of the integration in Eq 3 from 0 to p*. Equation 3, with
the lower limit of integral changed will then be applicable provided the
reversed plastic zone size Rp is large compared to 2p*. Incorporating
these modifications into the analysis, Eq 13 is altered as follows
--1 I F
da_ - 2 ( b + c ) [ crv' 1~ 1+ 2 p * l bb++ cc+ l
dN -b~ i 4(1 +-Y-) / ,/ j -Ud J
b+~+_l 1
b+c ~ AKz
-- [4( 1 + n'),,,'] "'"r/-O'~t2 (14)
The condition Rp >>2p* implies
AK>>2(1 + n ' ) { crv'~/2 ~-p* (15)

Notice that for large enough load levels, the COD will be large compared
to p* and Eq 14 reduces to Eq 13. The usual upturn in the log AK versus
log da/dN plot that is observed experimentally at very large values of
AK is not explained by the present analysis.
For small loads where the COD is of the order of o*, Eq 14 will give
significantly smaller da/dN values compared to Eq 13. Thus, in general,
the log da/dN versus log AK plot will be nonlinear and will asymptotically
approach the linear relationship of Eq 13 for large values of AK. At very
small values of 2xK, where R~ and 20* are of the same order of magnitude,
replacing the sum in Eq 2 by the integral in Eq 3 will result in con-
siderable error. An estimate of the threshold value AKth may, however, be
obtained by equating Rp with 20*, that is, replacing the inequality with
equality in Eq 15.
Figure 6 contains plots of ~xK versus daMN for a typical steel (same
properties as used as reference in Fig. 5) and for various values of p*.
Note that the effect of p* is significant at low values of ~xK and tends to
be less significant at higher values of LxK. The fatigue crack growth
resistance of any metal may then be thought of as consisting of two parts.
Firstly, the fatigue crack growth resistance is controlled by the bulk low
cycle fatigue properties and is given by Eq 13. Secondly, an increase in
resistance beyond that predicted by Eq 13 is achieved by introducing the
microstructure size, o*. It is the latter effect that makes it necessary to
select an exponent p of Eq 1 that is greater than two in order to fit the
power function to experimental data over a limited range of AK.
The physical interpretation of the microstructure size, p*, is not clear.
10-3
Properties as in Fig. 5
10-4
.= IO-e
o iO.r
i
t I i llli,i , i t JllIL
IO-e I I0 IOO tOO0
Stress Intensity Factor Range, ksi i,/~"
F I G . 6--Effect of microstructure size on fatigue crack propagation rate.
Some possibilities would include the distance between slip bands, grain
size, carbide spacing, mean free ferrite path, and so on.
The effects of size or scale of the microstructure on the yield strength
of steels have been studied in some detail. Recently, Gurland[19] has
shown that in many steels of various microstructures, the yield strength
can be expressed as a function of a microstructural parameter, 2~, in a
form similar to the Hall-Petch equation, namely
~r~=cro + k,j A-~ (16)

where
,to = friction stress;
ky= constant; and
grain size in low carbon mild steels,
mean free path between cementite particles for spheroidic and
x = tempered steels, and
mean cell diameter in deformed pearlitic steels.
When /t is interpreted as shown, data for a variety of steels correlate

well with a single Hall-Petch equation. For the purposes of the present
paper p* for steels will be approximated by ~.
Comparison of Eq 14 with Barsom's Data on Steel

Published crack propagation data are seldom accompanied by cyclic or
low cycle fatigue properties of the metals studied. To compare crack
propagation data from the literature with Eq 14, therefore requires that
the appropriate mechanical properties first be estimated. Barsom's data[6]
on fatigue crack propagation in steels were chosen for comparison with
Eq 14. The fatigue strength and ductility coefficients were approximated
from the reported values of ultimate tensile strength and true fracture
ductility[20]. The fatigue ductility exponent, c, was approximated as
- 0 . 6 0 for the four soft steels and as - 0 . 6 5 for the harder steels. The
various properties assumed for computation are shown in Table 1.
As mentioned before, o* has been approximated by ~ as given in Eq
16. Gurland's[19] reported values for cro--10.5 ksi and k~=0.564 ksi Vin.
were used in Eq 16.
Figures 7 through 14 show the predicted values of crack propagation
rate according to Eq 14 together with Barsom's test data. The predicted
values have been based on two values of c in Eq 10b as discussed earlier.
For all practical purposes, Barsom's test data fall within the band of
predicted values in all cases. Notice that the value of p* for the high
yield steels are on the average a factor of ten smaller than those for the
MAJUMDAR AND MORROW ON FATIGUE CRACK PROPAGATION 1 73
TABLE 1--Reported and estimated mechanical properties oJ the steels.
Values Reported
by Barsom [6] Estimated Cyclic Properties a
0.2% %
Ultj- Offset Reduc-
mate Yield tion
Steel Strength, Strength, in ~t',
Designation ksi ksi Area n' ~t' c ksi
HY80 113 95 70 ' 0.12 1.25 --0.65 163 --0.080
HY 130 148 140 67 0.12 1.15 --0.65 198 --0.080
10Ni-Cr-Mo-Co 193 182 71 0.12 1.25 --0.65 243 --0.080
12Ni-5Cr-3Co 187 184 64 0.12 1.00 --0.65 237 --0.080
A36 75 36 68 0.14 1.15 --0.60 125 --0.085
ABS-C 63 39 66 0.14 1.10 --0.60 113 --0.085
A302-B 88 56 67 0.14 1.10 --0.60 138 --0.085
A537-A 83 59 73 0.14 1.30 --0.60 133 --0.085
NOTE: Modulus of Elasticity was taken as 30 000 ksi for all steels.
an' ~ 0.12 and c ~ --0.65 were used for the four harder steels while n' ~ 0.14 and
c ~ --0.60 were used for the softer steels; b was approximated as b=n'c; et' was ap-
proximated as the true fracture ductility which is defined as ct'~---In[100/(100--%
reduction of area)]; ,rt' was estimated by adding 50 ksi to the ultimate strength.
These techniques for estimating the cyclic properties have been employed for a num-
ber of years at Illinois and elsewhere when only the ordinary tensile properties of a
steel are known. See, for example, Ref. 20. Actual test data would, of course, be pre-
ferable to estimates. Potential errors in the estimated properties, however, will have
a relatively small effect on the calculated crack propagation resistance (see Fig. 5).
low yield steels. O v e r the range of • investigated b y B a r s o m f o r the

high yield strength steels, the effect of o* is small. O n the o t h e r h a n d , for
the low yield strength steels the i n t r o d u c t i o n of o* gives g o o d c o r r e l a t i o n
with b o t h the m a g n i t u d e of the c r a c k p r o p a g a t i o n rate a n d the slope of
the log AK-log d a / d N plots except at the highest A K levels w h e r e Km~x
a p p r o a c h e s Kc.
Conclusions
A m o d e l of fatigue c r a c k g r o w t h has b e e n a n a l y z e d using existing
mechanics solutions for the cyclic stress a n d strain n e a r the c r a c k tip and
b u l k low cycle fatigue p r o p e r t i e s as o b t a i n e d f r o m testing u n i a x i a l " u n -
n o t c h e d " l a b o r a t o r y specimens.
If the m i c r o s t r u c t u r e size of the m e t a l is ignored, this analysis leads to
E q 13 which p r o v i d e s an u p p e r b o u n d for the rate of fatigue c r a c k p r o -
p a g a t i o n for any given stress intensity factor range, a n d the slope of the
log zxK log d a / d N p l o t is two.
This u p p e r b o u n d in fatigue c r a c k p r o p a g a t i o n rate has b e e n r e l a t e d to
the b u l k m e c h a n i c a l p r o p e r t i e s of the metal. Of m o s t i m p o r t a n c e in de-
10-3
10-4
Based //~
on(:: ~
.E I0s
/E
I0"s
g ned
a_ e
(J iO-r
=I ICk Plote
+ 2in Thick Plote
~= 4.5xIO-=in
O-y= 95 ksi
i i = illll i = i llllJ
I0 i i i t llH
I0 I00 I000
Stress Intensity Factor Range, ksi ~V~-
FIG.7--Comparison of experimental data [6]with theoretical prediction for HY-

80 steel.
MAJUMDAR AND MORROW ON FATIGUE CRACK PROPAGATION 1 75
i0-~
10-4
.E I0~
IO-~
//
(3.
10-7
- I in Thick Plate
+ 2 in Thick Plate
/~= 2 . 0 x I0 -5 in
~/= 140 ksi
I0 I00 tO00
Stress Intensity Factor Ronge, ksi tv/~fi-
FIG. 8--Comparison of experimental data [6] with theoretical prediction for HY-
130 steel.
10-3
10-4
t~
/
.F. IO-5
t o "e
0..
iO-7
= I in T h i c k Plate
+ 2 in T h i c k P l a t e
/~= I.Ox I0 "s in
~= Is2 ~i
iE8 i , , , ..... I ........
I0 I00 I000
Stress Intensify Factor Range, ksiI~'T~-
F I G . 9---Comparison of experimental data [6] with theoretical prediction for 10

Ni-Cr-Mo-Co steel.
10-s
/
10-4
/
.E I0-s
g:
g
~ 10_s
13-
o 10_7
I in Thick Plate
* 2 in Thick Plate
/~= I.Ox 10"6in
~y= 184~
iO-e
I tO IO0 '1000
Stress Intensity Factor Range, ksi I#T~.
F I G . lO--Comparison of experimental data [6] with theoretical prediction for 12

Ni-5Cr-3Co steel.
Io-s
tO-4
i.
..o
.c IO-S
j
#_
g
io-6
o
~ IO-T
§ I~ in Thick Plate
/~ = 4.gxlO-4in
<ry= 361,~i
i I i i i iii I i i iiiitl I i i i iiii
Io's I I0 I00 I000
Stress Intensity Factor Range, ksiv~-
F I G . 1 1 - - - C o m p a r i s o n o[ experimental data [6] with theoretical prediction for A36

steel.
MAJUMDAR A N D MORROW ON FATIGUE CRACK PROPAGATION 179
10-3
//
J/
10-4.
e
..~
._= 10-5
s
o
n,'
~, 10-~
/,Y
o
Q.
o 10-7
f!
9x I in T h i c k Plate
/~= 3 . 9 x 10"4 in
Cry= 3 9 k s i
lO" , , , ,,,J,l , , = ,,,,i

I0 I00 )00
Stress Intensity Foctor Ronge, ksiv~fi-
F I G . 12---Comparison of experimental data [6] with theoretical prediction for

ABS-C steel.
10-3
,p
i.
10-4 J
.F. I0"~
@:
~, ~0"6
ID
10-7
+ l in Thick Plate
/~= 15xlO "~ in
~-- ,56 ksi
I0" , , ,,,,,, , , , , .... I , , ,,,,,,

I0 I00 I000
Stress Intensity Factor Range, ksi =v~-
F I G . 13--Comparison of experimental data [6] with theoretical prediction for

.4302-B steel.
I0 -s ,, ,,
10-4
.__. I0"5
o ICTS'
g
2
a.
iO-r
x § ~ in Thick Plate
~= 1.4 x I0"* in
~= 5 9 ~
io-e i i i ,,ill t i i J =ll,J t i i i Iiii
I0 I00 I000
Stress Intensity Factor Range, ksi,/T6-
FIG. 14--Comparison of experimental data [6] with theoretical prediction of

A537-A steel.
termining this kind of fatigue crack growth resistance are the fatigue
ductility coefficient and exponent, the fatigue strength coefficient and the
elastic modulus while the yield strength is theoretically of little significance.
By introducing a microstructure size into the analysis, the crack prop-
agation rate, for any given stress intensity factor range, is reduced from
the upper bound just discussed. The reduction is larger, the larger the
microstructure size. On the other hand, for a given value of 0", the crack
propagation rate is increasingly reduced as • decreases resulting in a
nonlinear log • log da/dN plot.
In steels, a Hall-Petch type equation may be used to compute the
relevant microstructure size from the yield strength. Since the micro-
structure size varies inversely as the yield strength, this kind of fatigue
crack growth resistance of steels is greater the smaller the yield strength.
Acknowledgment
S u p p o r t for this study was p r o v i d e d b y the A d v a n c e d R e s e a r c h Projects
A g e n c y of the D e p a r t m e n t of D e f e n s e u n d e r U. S. D e p a r t m e n t of the
A r m y No. D A H C - 1 5 - 7 2 - G - 1 0 , A R P A O r d e r No. 2169. H. T. Corten,
G. M. Sinclair, H. R. J h a n s a l e , R. H. Sailors, a n d J. P. G a l l a g h e r con-
tributed to this r e s e a r c h t h r o u g h helpful discussions a n d critical review
of the manuscript.
References
[1] Paris, P. C., Gomez, M. P., and Anderson, W. E., The Trend in Engineering,
University of Washington, Vol. 13, No. 1, 1961, p. 9.
[2] Paris, P. C. and Erdogan, F., Journal ot Basic Engineering, Vol. 85, 1963,
pp. 528-534.
[3] Swanson, S. R., Cicci, F., and HHoppe, W. in Fatigue Crack Propagation, ASTM
STP 415, American Society for Testing and Materials, 1967, p. 312.
[4] Wilhelm, D. P. in Fatigue Crack Propagation, ASTM STP 415, American So-
ciety for Testing and Materials, 1967, p. 363.
[5] Schijve, J. in Fatigue Crack Propagation, ASTM STP 415, American Society
for Testing and Materials, 1967, p. 415.
[6] Barsom, J. M., "Fatigue-Crack Propagation in Steels of Various Yield Strengths,"
presented at the First National Congress of Pressure Vessels and Piping, San
Francisco, 10-12 May 1971.
[7] Liu, H. W., Journal o] Basic Engineering, Vol. 85, No. 1, 1963, pp. 116-122.
[8] Liu, H. W. and Iino, N. in Proceedings, 2nd International Conference on Frac-
ture, Paper 71, Chapman and Hall, 1969, pp. 812-823.
[9] Tomkins, B., Philosophical Magazine, Vol. 18, 1968, pp. 1041-1066.
[10] Dugdale, D. S., Journal of the Mechanics and Physics oJ Solids, Vol. 8, 1960,
pp. 100-104.
[11] Hickerson, J. P., Jr. and Hertzberg, R. W., Metallurgical Transactions, Vol. 3,
Jan. 1972, pp. 179-189.
[12] McClintock, F. A. in Fracture of Solids, Drucker and Gilman, Eds., Wiley, New
York, 1963, pp. 65-102.
[13] Rice, J. R. in Fatigue Crack Propagation, ASTM STP 415, American Society
for Testing and Materials, 1967, p. 247.
[14] Neuber, H., Journal of Applied Mechanics, Vol. 28, 1961, pp. 544-550.
[15] Weiss, V. in Fatigue--An Interdisciplinary Approach, Burke, Reed, and Weiss,
Eds., Syracuse University Press, pp. 179-186.
[16] Fleck, W. G. and Anderson, R. B. in Proceedings, 2nd International Confer-
ence on Fracture, Paper 69, Chapman and Hall, 1969, pp. 790-802.
[17] Fleck, W. G. and Anderson, R. B. in Proceedings, Air Force Conference on
Fatigue and Fracture of Aircraft Structures and Materials, Miami Beach, 1969,
pp. 417-424.
[18] Rice, J. R., "Stresses Due to a Sharp Notch in a Work Hardening Elastic Plas-
tic Material Loaded by Longitudinal Shear," Brown University Technical Re-
port NSF GK-286/1, Dec. 1965.
[19] Gurland, J. in Stereology and Quantitative Metallography, ASTM STP 504,
[20] Landgraf, R. W. in Achievement of High Fatigue Resistance in Metals and Al-
loys, ASTM STP 467, American Society for Testing and Materials, 1970, pp. 3-
36.
J. M . B a r s o m 1 a n d R . C. M c N i c o l ~
Effect of Stress Concentration on

Fatigue-Crack Initiation in HY-130 Steel
R E F E R E N C E : Barsom, J. M. and McNicol, R. C., "Effect o f Stress Con-

centration o n F a t i g u e - C r a c k Initiation in HY-130 Steel," Fracture Toughness
and Slow-Stable Cracking, .4STM STP 559, American Society for Testing
ABSTRACT: The useful fatigue life of structural components is determined
by the sum of the elapsed load cycles required to initiate a fatigue crack
and to propagate the crack from subcritical size to critical dimensions.
Thus, to predict the service life of many steel structures and to establish safe
inspection intervals, an understanding of the fatigue-crack-initiation and
fatigue-crack-propagation behavior in steel is required. In the present study,
the fatigue-crack-initiation life in HY-130 steel was investigated by testing
specimens having widely varying notch acuities. The variation in notch
acuity covered the range from fatigue-cracked specimens to polished, un-
notched specimens, and fatigue-crack-initiation data were obtained in the
range 103 to 100 cycles. The data were analyzed by using linear-elastic
fracture-mechanics concepts and the theory of stress concentration in
notched specimens.
The results showed that the number of elapsed load cycles required to
initiate a fatigue crack in a notched specimen is related to the ratio of the
fluctuation of the stress-intensity facor, AK~, to the square root of the notch-
tip radius, p. Fatigue-crack-initiation life can also be expressed in terms of
the fluctuation of the maximum stress at the notch tip, Ae. . . . because
AKr/VT-Ocan be related to Ae. . . . Moreover, preliminary observations sug-
gested that the fatigue-crack-initiation threshold in martensitic steels of
various yield strengths tested under zero-to-tension load fluctuation can be
predicted by the equation
AK
_ =0.6
~rrsVO
where ~-s is the 0.2 percent offset yield strength.
By using a weakest-link statistical model, it was shown that surface ir-
regularities can appreciably decrease the fatigue-initiation life of structural
components.
K E Y W O R D S : crack propagation, fracture properties, steels, fatigue (mate-
rials), notch tests, fatigue tests
1 Associate research consultant and associate research engineer, respectively, United

States Steel Corporation, Research Laboratory, Monroeville, Pa. 15146.
183
Copyright* 1974 by
Downloaded/printed by ASTM International www.astm.org
The fatigue life of structural components is determined by the sum of

the elapsed cycles required to initiate a fatigue crack and to propagate the
crack from subcritical size to critical dimensions. Thus, the fatigue life of
structures in a benign environment may be considered to be composed of
three stages: (1) fatigue-crack initiation, (2) fatigue-crack propagation,
and (3) fracture. The useful life of cyclically loaded structural components
can be determined only when these three stages are evaluated individually
and the cyclic behavior in each stage is thoroughly understood.
Linear-elastic fracture-mechanics (LEFM) concepts have been used to
analyze the rate of growth of fatigue cracks and the subsequent fracture
of structural components. These concepts show that the rate of fatigue-
crack growth is dependent primarily on the fluctuation in the intensity of
the stress field in the vicinity of the crack tip[I,2], z Moreover, unstable
fracture is shown to occur when the stress intensity at the crack tip reaches
a critical value that depends on test temperature, loading rate, and stress
state[3, 4, 5].
The magnitude and distribution of the stress field in the vicinity of nar-
row elliptical and hyperbolic notches, and their relationship to the stress-
field characteristics in the vicinity of a fatigue crack, were determined by
Creager[6]. Thus, the applicability of LEFM concepts to analyze the
fatigne-crack-initiation behavior of notched specimens appears to be pos-
sible. The applicability of these concepts to sharp notches has been demon-
strated by the limited work of Forman[7], Constable et al[8], and Jack and
Price[9].
This report presents the results of a preliminary investigation of the
fatigue-crack-initiation behavior in steels. The test results were obtained by
testing HY-130 steel specimens having widely varying notch acuities. The
variation in notch acuity covers the range from fatigue-cracked specimens
to polished, unnotched specimens. The fatigue-crack-initiation data were
obtained in the range 103 to 108 cycles. The data were analyzed by using
LEFM concepts.
Materials and Experimental Work

A 1-in.-thick (25.4 mm) plate of HY-130 steel was used in the present
investigation. The chemical composition and the room temperature tensile
properties of the plate are given in Tables 1 and 2, respectively.
Specimen Preparation
Double-edge-notched specimens were machined from the midthickness
and quarter thickness of the 1-in.-thick HY-130 steel plate. The specimens
paper.
BARSOM AND McNICOL ON STRESS CONCENTRATION 185
TABLE 1---Chemical composition of the HY-130 steel investigated, percent.
C Mn P S Si Ni Cr Mo V A1" N O
0.11 0.87 0~002 0.005 0.30 4.85 0.58 0.48 0.070 0.018 0.010 0.0033
Acid soluble.
TABLE 2--Longitudinal mechanical properties o/ 1-in.-thick HY-130 steel plate

investigated.
Yield Strength Tensile Elongation Reduction

(0.2% Offset), Strength, in 1 In., of Area,
ksi ksi % %
146 153 19 68
were 20-in. long, 6-in. wide, and 0.125-in. thick (0.51 m by 0.15 m by
3.18 mm). The specimens were machined with their long axis parallel to
the final rolling direction of the plate. The geometry of the notches is shown
in Fig. 1 for notch-tip radii, p, ranging from 0.008 to 0.375 in. (0.20 to
9.53 mm). The polished, unnotched specimens that were tested had a 10-
in. (0.25-m) radius and are shown in Fig. 2.
T
6.00"
L ,~ 11.25"
19.25"
I~
j-
I It I I It I --f-
--i2,r- -
O. 900"
o.oo8"<p < 0.375"
I inch : 25.4 mm
ENLARGED NOTCH DETAIL
FIG. 1--Double-edge-notched specimens.
.00" RAOIUS
F0.125" ~-.0.001"
/ T
3.','50"
~o.oo3"
1
19.250" + O.ZSO"
I inch - 2 5 . 4 m m
FIG. 2--Geometry of unnotched specimens.
The notches and the tip radii in specimens where 0.008 < p < 0.375 in.
were made with an electric discharge machine ( E D M ) by using a graphite
electrode and by operating the machine under surface-finishing conditions
rather than metal-removal conditions. The surface of the notch and the
area surrounding the notch tip were then polished to a No. 8 finish (a
proftlometer indicated an arithmetic-average roughness of 8 ~in., or 203
mm). The specimens having a 10-in. radius were machined and then po-
lished to a No. 8 finish.
All specimens were tested at room temperature in a 50-kip (222 MN)
materials testing machine ( M T S ) . In general, the frequency of stress fluc-
tuation was 120 to 200 cpm for specimens with crack-initiation lives up to
about 5 x 104 cycles, and 400 to 600 cpm for specimens with lives greater
than 5 x 104 cycles. All the tests were conducted under tension-to-tension
loading at a stress ratio, R (the ratio of minimum applied stress to maxi-
mum applied stress), equal to § 0.1.
Fatigue-crack initiation was detected optically at x 10 to x 2 5 magni-
fication with a stereo-zoom microscope mounted on a traversing base.
Each test was terminated when the initiating crack extended 0.010 in.
(0.254 mm) from the notch root.
The specimens were tested over a range of fluctuating stresses, and the
number of cycles required to initiate an 0.010-in. crack was recorded.

Stress Analysis ol Notched Components
Stress Analysis of Sharp Cracks The fatigue behavior of structural
components that contain stress (strain) raisers is dependent on the magni-
tude and distribution of the stress (strain) in the vicinity of the stress
raiser. The magnitude and distribution of the stress field in the vicinity of
a fatigue-crack tip can be related to the nominal stress applied on the
structure, to the geometry and size of the defect, and to the material prop-
erties by using LEFM concepts.
The elastic-stress field in the vicinity of a sharp crack tip (that is, tip
radius, p, approaching zero) in a body subjected to tensile stresses normal
to the plane of the crack (Mode I deformation) is represented by the fol-
lowing equations[10].
o F 1 - s i n 20- sin
KI cos -~-
cry: (2,rrr)89
KI cos~[ l+sin -~-Osin - -~ ]

*u-- (27rr)89
K1 . 0 0 30
r~u= (2~rr)89 sm -~- cos-~- cos-~-
~=v(~+~), ~=~=0 (1)
where the stress components and the coordinates r and 0 are presented in
Fig. 3. These plane-strain equations show that the elastic-stress-field dis-
tribution in the vicinity of a sharp crack tip is invariant in all structural
components subjected to Mode I deformation, and that the magnitude of
the elastic-stress field can be described by a single-term parameter, Kx,
designated the stress-intensity factor. Consequently, the applied stress, ~,
the crack dimensions, and the structural configuration associated with
structural components subjected to Mode I deformation affect the value of
the stress-intensity factor but do not alter the stress-field distribution. Rela-
tionships between the stress-intensity factor and various body configura-
tions, crack sizes, and loading conditions have been presented by Paris and
Sih[ll].
The growth of fatigue cracks in structural components subjected to cyclic
stress fluctuations is a localized phenomenon that depends on the bound-
ary conditions at the crack tip. Consequently, it is logical to expect the
rate of growth of such fatigue cracks to be dependent on the stress-intensity
factor, K~, which provides a single-term parameter representation of the
stress conditions in the vicinity of the crack tip. Sufficient data are available
in support of this observation[I,2,12].
(7"
yAcoNx,x,\
ill X
Z Y
FIG. 3---Schematic illustration of the elastic-stress-field distribution near the tip
of a ]atigue crack (Mode 1 de[ormation).
Stress A n a l y s i s of B l u n t N o t c h e s - - - T h e useful cyclic life of many struc-

tural components is determined by the number of elapsed cycles required
to initiate a fatigue crack from a stress raiser having a finite radius of cur-
vature at the tip, plus the number of elapsed cycles required to propagate
the fatigue crack to critical dimensions. The latter portion of the useful
cyclic life is determined by the rate of fatigue-crack propagation in the
structural component, which has been shown to be, primarily, a function
of the stress-intensity-factor fluctuation, AKz[1,2,12]. Consequently, to
investigate the total useful life of cyclically loaded structural components
that contain notches of finite root radii, the magnitude and distribution of
the elastic-stress field in the vicinity of the notch tip must be established and
its relationship to the fatigue-crack-initiation behavior in blunt notches must
be determined.
The elastic-stress field in the vicinity of sharp elliptical or hyperbolic
notches in a body subjected to tensile stresses normal to the plane of the
notch is represented by the following equation[6].
K1 0[i_sin~_sin301 K1 ~ cos~
~- (2~rr)89c o s ~ - (2~rr)89 2r
01 0
cry- (2rr) 89 cos ~ - 1 + s i n ~ - sin -~ (2~rr)~ 2r
K~ . 0 0 30 Kz P sin 30 (2)
r~- (2rr) 89 s m - ~ - c o s ~ - c o s 2 (2~-r) 892r -2-
where the coordinates r, 0, and p are defined in Fig. 4. The first term in
Eq 2 defines the magnitude and distribution of the stress field in the vicinity
of a fatigue crack. The second term in these equations represents the influ-
ence of a blunt-tip radius on this stress field. Equation 2 also shows that on
the crack center plane the stress singularity for narrow elliptical and hyper-
bolic notches is centered on a line located at p/2 behind the crack front,
Fig. 4.
Notches in structural components cause stress intensification in the vi-
cinity of the notch tip. The material dement at the tip of a notch in a
cyclically loaded structural component is subjected to the maximum stress,
Crmax, and to the maximum stress fluctuations, Zx%~ax. Consequently, this
material element is most susceptible to fatigue damage and is, in general,
FIG. 4---Schematic illustration of the elastic-stress-field distribution near the tip

of an elliptical notch (Mode ! deformation).
the origin of fatigue-crack initiation. The maximum stress on this material

element is[13]
where
a = nominal stress, psi,
a = crack length, in., and
p = notch-tip radius, in.
The stress-intensity factor, K~, for the double-edge-notched specimens
investigated in this study, Fig. 1, is
K I = 1.1crVrra (4)
Substituting this expression for K1 in Eq 3, and for p<<a, we obtain
gm,~-- V~-p = 2 . 2 a
Va
This equation, which is applicable to sharp elliptical or hyperbolic notches,
(5)
can be obtained also from Eq 2 by substituting 0=zero and r~-p/2.

Because all the specimens tested in this investigation were loaded essentially
under zero-to-tension loading conditions, ~max= A~,x and K~ = AKI can be
substituted in Eqs 3 and 5.
Test Results
The fatigue-crack-initiation behavior of all the specimens tested in the
present investigation is presented in Fig. 5 in terms of the number of cycles
for fatigue-crack initiation, N~, versus the nominal-stress fluctuation, ~g.
A test was considered a "runout" when no indication of crack initiation
was observed at the notch tip after 10" cyclic-stress fluctuations. The data
show that the magnitude of the applied-stress fluctuation required to ini-
tiate a fatigue crack after a fixed number of elapsed cycles increased as the
root radius of the notch increased. Similarly, the number of elapsed cycles
required to initiate a fatigue crack in specimens tested under a fixed stress
fluctuation increased with increasing root radius. These general observa-
tions are expected, and further data analysis is necessary to establish the
relationships among the various parameters.
To minimize the data scatter encountered in this S-N type test, the data
presented in Fig. 5 were replotted in terms of the nominal-stress fluctuation,
A~, versus the notch-tip radius, p, Fig. 6. The data show that, within the
limits of this study, the relationship between ag and p at a fixed number of
elapsed cycles to initiate a fatigue crack, N~, is represented by the equation
FIG. 5--Dependence of fatigue-crack initiation of HY-130 steel on nominal-stress

fluctuations for various notch geometries.
FIG. 6--Relationships of nominal-stress fluctuation and notch-tip radius in HY-130

steel at various Jatigue-crack-initiation lives.
A ~ = A ( p ) 1/" (6)
where ,4 is constant for a given value of N~ and where the value of m in-
creases with decreasing value of cyclic-initiation life. The value of m
increases from 2 at N~-- 106 cycles to 3.2 at N ~ = 2 x 103 cycles. The appro-
priate values of ,4 and m were used in Eq 6 to re-present the dependence
of N~ on the magnitude of the fluctuation of the applied stress. This repre-
sentation is shown in Fig. 7 for 0 . 0 < p < oo. Data for o approaching a very
large number were obtained from tests conducted on unnotched specimens,
Fig. 2. The curve for p approaching zero, which represents a fatigue crack,
was calculated from the fatigue-crack-growth-rate data for HY-130 steel,
Fig. 8[14]. The procedure used to calculate the curve for p approaching
zero is presented in Appendix A.
The data presented in Fig. 7 for 0 < p < oo show that the initiation cyclic
life in the low-cycle region (defined by N~_<3 x 10' cycles) can be repre-
sented by the relationship
N~=t~(Acr) ~ (7)
where tt is a constant and n = 2 for a fatigue crack and increases to n = 8
for an unnotched specimen. The dependence of the value of the exponent,
n, on the notch-tip radius is shown in Fig. 9. Because of the monotonic
increase in the value of n and the change in the value of ~ as a function of
FIG. 7--Dependence of ]atigue-crack initiation by HY-130 steel on nominal-stress

fluctuations ]or various notch geometries.
FIG. 8--Fatigue-crack-propagation data Ior HY-130 steel.
increase in the value of p, the initiation cyclic life, N~, for different notch
acuities could not be related to the stress-intensity-factor fluctuation alone
but must also be related to the notch-tip radius. On the basis of the limited
work of Jack and Price[9], and because the stress-field equations in the
vicinity of a notch can be rendered nondimensional by using the nondimen-
sional stress-intensity parameter K / , r y ~ / p , N~ was plotted as a function
of AK~/~/p in Fig. 10. This figure shows that a fatigue-crack-initiation
FIG. 9--Dependence of the nominal-stress exponent on the notch-tip radius in the

region of finite fatigue-crack-initiation life of HY-130 steel.
FIG. lO--Correlation o[ ]atigue-crack-initiation life with the parameter Kr/V-pp

for HY-130 steel.
threshold for the various notched specimens having 0.008 <0<0.375 in.
(0.20 to 9.53 mm) occurs when
a K i _ 85 ksi (586 MN/m ~-) (8)

Vp
That is, the data show that fatigue cracks do not initiate in HY-130 steel
structural components when the body configuration, the notch geometry,
and the nominal-stress fluctuations are such that the value of the param-
eter, AK~/~/p, is less than 85 ksi. The data also show that at a constant
value of LxK~/~/p, the number of cycles required to initiate a fatigue crack
in the region of finite cyclic life (that is, when N~_<3 • 104 cycles) increases
as the notch-tip radius increases. A possible explanation of this behavior is
presented in a later section.
Fatigue-Crack-Initiation Threshold
The data presented in Fig. 10 show that the fatigue-crack-initiation
threshold at AK~/Vp= 85 ksi is reasonably applicable to notches having a
stress-concentration-factor value, kt, that ranged from 17.2 for the speci-
mens with 0--0.008 in. to 2.9 for the specimens with p=0.375 in. These
kt values were obtained by using a modified form of Eq 3 that includes a
finite-width correction factor[13]. Considering the wide range of kt values
BARSOM AND McNICOL ON STRESS CONCENTRATION ]95
for which this observation is applicable, one is tempted to apply Eq 8 to

polished, unnotched specimens (kt= 1.0) and to fatigue-cracked specimens
( k t ~ ~).
Equation 5 relates the maximum stress at the tip of a narrow notch to
the stress-intensity factor, K~, and to the notch-tip radius, p. This equation
was derived from Eq 2, which is applicable to elliptical and hyperbolic
narrow notches. Thus, the applicability of Eq 5 to the large notch-tip radii
tested in the present investigation may be questionable. However, the test
results presented in Fig. 10 suggest that fatigue-crack initiation from blunt
notches is governed, primarily, by the nondimensional stress-intensity
parameter, K/~rs~U-p(or K/Vo for a single steel, as in the present investi-
gation). The value of maximum stress, ~..... at runout of the notched
specimens and the variation in the value of K/Up at runout were within
_+5 percent.
Despite the fact that the term ~K~/\/o loses its significance for an un-
notched, polished specimen, let us assume that the fatigue-crack-initiation
limit in such specimens occurs when • 85 ksi. Substituting this in
Eq 5 shows that the fatigue-crack-initiation threshold of unnotched,
polished HY-130 steel specimens tested at a stress ratio, R, equal to 0.1
occurs when • is equal to 96 ksi (662 MPa). This value agrees quite
well with the experimental data obtained by testing unnotched, polished
specimens, Figs. 5 and 7. The agreement between the calculated value and
the experimental test results may be fortuitous, and further work is neces-
sary to establish its significance.
NonpropagatingFatigueCracks
Recent fatigue-crack-growth data suggest the existence of a fatigue-
threshold stress-intensity factor, Kth, below which fatigue cracks do not
propagate under cyclic-stress fluctuations[15-17]. A conservative esti-
mate of Kth for steels tested under tension-to-tension loads at a stress
ratio, R, equal to 0.1 is 5 ksiCin. (5.5 MN/m3/2).
The crack-opening displacement, 3, at the tip of a fatigue crack is given
by the equation
K2
3= (9)
where
K = stress-intensity factor, psi !/in.,

~rrs= yield strength, psi, and
E = Young's modulus, psi.
The value of ~ at maximum load in a fatigue-cracked HY-130 steel speci-

men tested under zero to 5 ksiVin, fluctuation in the stress-intensity
factor is equal to 0.6• 10 5 in. (152 nm). The value of the tip radius is
on the order of 8/2, or p = 0 . 3 x 10 ~ in. (76 nm). Substituting this value
of p in Eq 8 suggests that Kth ----0.15 ksi~/in. (0.16 MN/m3/:), which
appears to negate the existence of a fatigue threshold in fatigue-cracked
specimens of HY-130 steel.
Let us assume that a fatigue threshold in a fatigue-cracked HY-130
steel specimen does occur at Ktl~Z5 ksi~v/in. Then, Eq 9 suggests that
p=2.5 • 10 -3 in. (64~m) at the fatigue threshold. Judging by the value of
the crack-opening displacement and by extrapolating the data in Fig. 9 to
n = 2 , it appears that the value of p=2.5 • 10-~ in. is unrealistically high.
The previous observations appear to negate the existence of a fatigue
threshold in fatigue-cracked HY-130 steel specimens, and thus, do not
agree with available fatigue-threshold d a t a [ 1 5 - 1 7 ] . This apparent dis-
crepancy is based on the assumption that the fatigue-crack initiation be-
havior can be extrapolated to tip radii (that is, stress concentration factors)
well below the 0.008-in. radius tested. The discrepancy may be resolved if
the fatigue-crack-initiation behavior of specimens containing root radii
smaller than 0.008 in. is independent of the root radius (stress concentra-
tion factor). Consequently, the fatigue-crack-initiation behavior of speci-
mens containing notches of 0.001-in. tip radius and of the geometry shown
in Fig. 1 was investigated. The test results are represented in Fig. 11 super-
imposed on the data obtained by testing specimens containing notches of
0.008-in. tip radius. The data show, conclusively, that the fatigue-crack-
initiation behavior of the specimens tested in the present investigation is
independent of the notch-tip root radius, p, for p_<0.008 in. Equation 8
I i , I 'l" I i 9 i , i 9 1, I
200
~ ~ ,51o C O R R E S P O N D I N G TO ~ K t h
g 0 p = 0 0(~ inch
7-
Too 9 # 0~1 inch
4O 6
2t0 2l I J I i I J IJ I I , =,,,I I I , I ~ I I
4 6 8 10 2 2 4 6 8 ?0 3 2 4 G 8 104
FATIGUE-CRACK iNITIATrON CYCLES, N( x 10 3
FIG. 11--Fatigue-crack initiation o[ H Y - 1 3 0 steel ]or very small root radii
indicates that the value of the fatigue-threshold stress-intensity factor,

Kth, for p=0.008 in. is about 7.6 ksi ~/in. which is in good agreement
with the conservative estimate of Kth=5 ksi ~/in. for steels tested at a
stress ratio of 0.1, Fig. 11.
Fatigue-Threshold Parameter in Steels of Various Yield Strengths

The present data show that the fatigue-crack-initiation threshold of
HY-130 steel (yield strength, ~,.~, equal to 146 ksi, or 1007 M N / m z under
zero-to-tension loading is given by the equation
K
_ - - 8 5 ksi (586 M N / m 2) (8)
Vp
Normalizing this equation with respect to o,ys gives
K
_ =0.6 (10)
Ors ~r
Preliminary observations suggest that the fatigue-crack-initiation thresh-
old in steels of various yield strengths can be predicted by using Eq 10.
This equation may also prove to be applicable to metal alloys other than
steels. The use of this relationship to predict the fatigue-crack-initiation
threshold in various metal alloys may require normalizing the relationship
with respect to Young's modulus, E, or some power of E. Further work is
necessary to establish the general applicability of Eq 10 for predicting the
fatigue-crack-initiation threshold in steels and other metals and to estab-
lish the effect of the cyclic strain-hardening exponent on the fatigue-crack-
initiation behavior.
Finite-Initiation-Life Behavior
The fatigue-crack-initiation life of unnotched specimens is strongly
dependent on the surface conditions of the specimen. Surface damage and
surface irregularities can reduce the initiation life significantly because of
the stress concentration. On the other hand, the fatigue-crack-initiation
life of unnotched, polished specimens is caused by the plastic deformation
of the material. The plastic deformation causes the development of slip
steps that become the nucleus of the fatigue-crack-initiation site. These
observations are equally applicable to notched specimens. The fatigue-
crack-initiation life of a notched specimen can be reduced significantly by
surface irregularities in the vicinity of the notch tip.
The data presented in Fig. 10 show that, in the finite-initiation-life
region, the number of elapsed cycles required to initiate a fatigue crack
198 FRACTURE T O U G H N E S S AND SLOW-STABLE C R A C K I N G
at the tip of a notch, N~, is related to the parameter AK~/V-p (or Ki/av/p
for zero-to-tension loading) by the equation
Ni=fl( K1 ,V, (11)
\ x/o/
where/3 is a constant, and the value of the exponent, n, decreases as the
root radius decreases. More importantly, the data show that, at a con-
--
stant value of &K~/~/p, N~ is primarily a function of l / F ,e Fig. 12. This
functional relationship indicates that, at a constant value of ~XK~/1/p, the
' i ' I ' I'F'I ' I

08--
- - 2O
06
I DATA AT
04~ --i10
0
02
\
01
006
0O4
8
002 -- \. O6
O4
001
0OO8
\ O2
0.006
00O4
1 , n c h = 254 mm
0.002
0.001 , I , I ,/,LI , L
2 4 6 8 10 20
FATIGUE CRACK-INITIATION CYCLES, N x !03
FIG. 12--Dependence of fatigue-crack-initiation life on the notch-tip radius in the

region of finite-initiation life of HY-130 steel.
number of elapsed cycles required to initiate a fatigue crack at the tip of a

notch in a specimen of unit thickness is inversely related to the volume
of an element at the notch tip rather than to the surface of the notch
root. In other words, the fatigue-crack-initiation life in the specimens
tested in the present investigation decreased as the volume of the plastically
deformed material at the notch tip increased. Because of the high quality
of the surface of the notches that were tested, the surface did not affect
the number of elapsed cycles required to initiate a fatigue crack. The data
presented in Fig. 12 also suggest that the fatigue-crack-initiation life in
the finite-life region is independent of notch-tip radius when o is large
(that is, when the volume of the plastically deformed metal is large) and
when p is small. This behavior occurred in HY-130 steel when o was
greater than or equal to 0.25 in. (6.4 mm) and when p was less than or
equal to 0.008 in. (0.20 m m ) .
The fatigue-crack-initiation life of structural components is influenced
by surface irregularities, by internal inclusions and other volumetric stress
(strain) raisers, and by the volume of the plastically deformed metal. A
mathematical model that depicts the dependence of fatigue-crack-initiation
life on surface and volumetric characteristics is presented in Appendix B.
The model is based on a weakest-link theory and Weibull's distribution
function[18,19,20]. It shows that fatigue-crack initiation in structural
components can be represented by an areal-risk parameter, a volumetric-
risk parameter, a zero-probability stress range, a flaw-density exponent,
and a scale parameter.
Summary
The results of a preliminary investigation on the fatigue-crack-initiation
behavior of notched HY-130 steel specimens under zero-to-tension loading
may be summarized as follows:
1. The number of elapsed load cycles required to initiate a fatigue
crack in a notched specimen is related to the ratio of the stress-intensity-
factor fluctuation to the square root of the notch-tip radius, AKI/X/-p.
2. The number of elapsed load cycles required to initiate a fatigue crack
in notched specimens can be related to the fluctuation in the maximum
stress at the notch tip, a~ ....... because a~ .... is related t o AKi/k/p.
3. A fatigue-crack-initiation threshold in notched HY-130 steel speci-
mens occurred at AKI/!/p = 85 ksi (586 MN/m~-).
4. The fatigue threshold in polished, unnotched HY-130 steel speci-
mens was predicted by substituting AK~/1/pz85 ksi in the equation
A~max=2( AKI) /~/~p.
5. The fatigue-crack-initiation threshold under zero-to-tension loads in
notched specimens of martensitic steels may be approximated by using the

relationship
K1 __=0.6
~ys ~/p
where ~ys= yield strength.
6. The fatigue-crack-initiation behavior appears to be independent of
root radius (that is, stress concentration factor, kt) for root radii smaller
than a given value and larger than a given value. In the present investiga-
tion, these values were p_<0.008 in. and p~_0.25 in. which correspond to
kt > 17.2 and kt ~ 3.4, respectively.
7. Second-order dependence of the number of cycles to initiate a
fatigue crack in notched specimens on the magnitude of the tip radius
appears to be related to the size of the plastically deformed material in
the vicinity of the notch tip.
8. A weakest-link statistical model showed that surface irrr
can appreciably decrease the fatigue-initiation life of structural com-
ponents.
APPENDIX A
Derivation of Equation for Expressing the Number of Cycles for Crack Ini-
tiation (N~) in Terms of the Applied-Stress Fluctuation (Act) in the Case of a
Preexisting Fatigue Crack
The fatigue-crack-growth-rate data presented, in Fig. 8 can be approxi-

mated by the equation
da
= l A x 10-8 ( A K I ) 2 (12)
dN
where
da
=fatigue-crack-growth rate per cycle, inch per cycle, and
zXK~= stress-intensity-factor fluctuation, ksil/in.
The stress-intensity factor for the double-edge notched specimens tested
in the present investigation is given by Eq 4, and therefore, the stress-
intensity-factor fluctuation is
~XK~--1.1 (• ~/,ra (13)
Because Aa was held constant for each test, substituting this expression of
~KI in Eq 12 and separating variables gives
d N = 1 9 x 10 6 d a (14)
(4r 2 a
The number of elapsed cycles for crack initiation, N~ [where fatigue-crack
initiation was defined in this study as an 0.01-in. (0.25 mm) crack
extension from the notch tip], is obtained by integrating Eq 14:
19• 6 fa~
f N dN
N
da
(15)
N , = 19• 106 In ao+0.01 (16)

(Act) 2 ao
where ao was equal to 0.90 in. (23 mm) in the present investigation.
Equation 16 relates N~ for a fatigue crack to the applied-stress fluctuation,
• and was used to plot the curve for p ~ 0 in Fig. 7.
APPENDIX B
Areal and Volumetric Effects on the Probability of Fatigue-Crack Initiation
Statistical theories have been applied to explain the dependence of

strength of specimens on their shape, length, volume, and surface area.
The basic concepts underlying these theories can also be used to analyze
the effects of the physical parameters of specimens on the probability of
fatigue-crack initiation.
The statistical models used to determine the probability of failure of a
material are based on the assumption that the variation in the strength of
specimens is caused by flaws of various severity that are distributed within
the test section. Moreover, the distribution of flaws within an element of
area or of volume is assumed to be random with a certain density. The
strength of a specimen subjected to uniform stress is a function of the
severest flaw that exists in the specimen. Because the strength of the
whole is only as great as its weakest link, these approaches have been
designated the "weakest-link" theories of fracture. To allow for analytical
formulation of these theories, a specimen is visualized as being composed
of individual infinitesimal links (areas or volumes), each having a certain
probability of failure that corresponds to the stress level applied to the
link.
The most widely accepted formulation of the weakest-link theory of

fracture was proposed by W e i b u l l [ 1 8 , 1 9 ] , who suggested a special dis-
tribution function of wide applicability[20], t h e weakest-link theory of
fracture that is based on Weibull's distribution function postulates that
failure in an isotropic, homogeneous material is fully described by the
zero-probability strength, a flaw-density exponent, and a scale parameter.
This theory is applicable to various stress states and to materials the
fracture strength of which depends on surface area or volume, or on both.
The probability of failure of a homogeneous, isotropic material sub-
jected to uniaxial stress, ~, was expressed by Weibull as follows.
P(x) = I - eIx p - - f
d
(--~m.dx~
x k
(7-- O-ox
O'bx / _J
= 1 - e Br for cr~ r (17)

where
B= i ('-'o0
X \
- -
O'bx /
mxdx was termed by Weibull as the "risk of rup-
ture,"
x= surface area, A , or volume, V, of the tested material,
~o~= zero-probability strength,
m, = flaw-density exponent, and
~b~ = scale parameter.
The probability of failure of a material governed by both a surface-flaw
distribution and a volume-flaw distribution, P ( A , V ) , is given by the
relationship
P(A,V) =P(A ) +P(V) -P(A)P(V) (18)
Thus, by using Eq 17,
P(A,V) = 1 -- e - ' ( A , v ) = 1 -- e -t'(a~+'(v)~ (19)
where
if__ r ma ( T - - ff o v my
for ~o,<~>~o~ (20)

The prior statistical analysis of fracture strength can be adapted to
analyze the fatigue-crack-initiation behavior of unnotched and notched
specimens. The fatigue-crack-initiation life in a specimen subjected to a
stress fluctuation, ~, is assumed to be a function of the severest flaw con-
tained in the material. Thus, the weakest-link fatigue-crack-initiation
theory would postulate that fatigue-crack initiation in an isotropic, homo-
geneous material is fully described by the stress fluctuation corresponding

to zero probability of fatigue-crack initiation, by the flaw-density exponent,
and by a scale parameter. The area risk of fatigue-crack initiation, B(A),
depends on the severest strain raiser contained in the surface area under
consideration. The volume risk, B(V), depends on the severest strain
raiser contained within the material and the volume of the plastically
deformed metal. The dependence of B (V) on the volume of the plastically
deformed metal is important in inclusion-free, polished specimens. The
combined effects of the distribution of inclusions within the material and
of the volume of the plastically deformed metal are accounted for by the
zero-probability fatigue-crack-initiation stress range, • by the effective
flaw-density (inclusion-density) exponent, m~, and by the scale parameter,
• Numerical values of these three parameters can be obtained graphi-
cally from experimental data in a manner similar to that proposed by
Ansevin[21].
The surface of the HY-130 steel specimens tested in the present investi-
gation and the surface of the notches in these specimens were polished
surfaces. Consequently, the effect of the area risk, B(A ), on the probability
of fatigue-crack initiation was minimal. Moreover, because the HY-130
steel tested was free of inclusions that cause internal fatigue-crack initia-
tion, the dependence of P(A,V) on the volumetric distribution of internal
flaws (inclusions) was also negligible. Consequently, the probability of
fatigue-crack initiation of the HY-130 steel tested must be dependent,
primarily, on the volume of the plastically deformed metal. This functional
dependence was verified by the data presented in Fig. 11.
The fatigue-crack-initiation life of notched specimens can be reduced
significantly by surface irregularities in the vicinity of the notch tip. Thus,
the probability of fatigue-crack initiation in such specimens could be gov-
erned, primarily, by the areal-risk parameter, B(A), rather than by the
volumetric-risk parameter, B (V).
References
[ll Paris, P. C., "The Fracture Mechanics Approach to Fatigue," Proceedings,
Tenth Sagamore Army Materials Research Conference, Syracuse University
Press, 1964.
12] Barsom, J. M., "Fatigue-Crack Propagation in Steels of Various Yield
Strengths," ASME Paper 71-PVP-12, ASME Transactions, Journal oj Engineer-
ing/or Industry, Nov. 1971, pp. 1190-1196.
[3] Barsom, J. M. and Rolfe, S. T., "K~e Transition-Temperature Behavior of
A517-F ,Steel," Engineering Fracture Mechanics, Vol. 2, No. 4, June 1971.
[4] Shoemaker, A. K. and Rolfe, S. T., "The Static and Dynamic Low-Temperature
Crack-Toughness Performance of Seven Structural Steels," Engineering Fracture
Mechanics, Vol. 2, No. 4, June I971.
[5] Greenberg, H. D., Wessel, E. T., and Pryle, W. H., "Fracture Toughness of
Turbine-Generator Rotor Forgings," Engineering Fracture Mechanics, Vol. 1,

No. 4, 1970.
[6] Creager, M., "The Elastic Stress-Field Near the Tip of a Blunt Crack," Master
of Science thesis, Lehigh University, Bethlehem, Pa., 1966.
[7] Forman, R. G., "Study of Fatigue-Crack Initiation From Flaws Using Fracture
Mechanics Theory," Air Force Flight Dynamics Laboratory Technical Report
AFFDL-TR-68-100, Sept. 1968.
[8] Constable, I., Culver, L. E., and Williams, J. G., "Notch-Root-Radii Effects in
the Fatigue of Polymers," International Journal of Fracture Mechanics, Vol. 6,
No. 3, Sept. 1970.
[9] Jack, A. R. and Price, A. T., "The Initiation of Fatigue Cracks From Notches in
Mild-Steel Plates," International Journal of Fracture Mechanics, Vol. 6, No. 4,
Dec. 1970.
[10] Irwin, G. R., "Analysis of Stresses and Strains Near the End of a Crack Travers-
ing a Plate," Journal of Applied Mechanics, Vol. 24, 1957.
[11] Paris, P. C. and Sih, G. C., "Stress Analysis of Cracks," Fracture Toughness
Testing and Its Applications, ASTM STP 381, American Society for Testing
and Materials, 1965.
[12] Clark, W. G., Jr., "Fatigue-Crack-Growth Characteristics of Heavy-Section
ASTM A533 Grade B, Class 1 Steel Weldments," ASME Paper 70-PVP-24,
American Society of Mechanical Engineers, presented at the Petroleum Mechani-
cal Engineering and Pressure Vessels and Piping Conference, Denver, Colo.,
13-17 Sept. 1970.
[13] Baratta, F. I. and Neal, D. M., "Stress Concentration Factors in U-Shaped and
Semi-Elliptical Edge Notches," Journal of Strain Analysis, Vol. 5, No. 2, 1970.
[14] Barsom, J. M., Imhof, E. J., Jr., and Rolfe, S. T., "Fatigue-Crack Propagation
in High-Yield-Strength Steels," Engineering Fracture Mechanics, Vol. 2, No. 4,
June 1971.
[15] Schmidt, R. A., "A Threshold in Metal Fatigue," Master of Science thesis,
Lehigh University, Bethlehem, Pa., 1970.
[16] Paris, P. C., "Testing for Very Slow Growth of Fatigue Cracks," MTS Closed
Loop Magazine, Vol. 2, No. 5, 1970.
[17] Bucci, R. J., Clark, W. G., Jr., and Paris, P. C., "Fatigue-Crack-Propagation
Growth Rates Under a Wide Variation of K for an ASTM A517 Grade F (T-I)
Steel," presented at the Fifth National Symposium on Fracture Mechanics, Uni-
versity of Illinois, Urbana-Champaign, Ill., 31 August to 2 September 1971.
[18] Weibull, W., "A Statistical Theory of the Strength of Materials," Ingeniors
Vetenskaps Akademien, Handlingar, Vol. 151, 1939.
[19] Weibull, W., "The Phenomenon of Rupture in Solids," Ingeniors Vetenskaps
Akademien, Handlingar, Vol. 153, 1939.
[20] Weibull, W., "A Statistical Distribution Function of Wide Applicability," Journal
of Applied Mechanics, Vol. 18, 1951.
[21] Ansevin, R. W., "Analysis of Fracture Probabilities in Stressed Glass," Master
of Science thesis, University of Pittsburgh, 1965.
W . G . C l a r k , Jr. 1
Evaluation of the Fatigue Crack Initiation

Properties of Type 403 Stainless Steel
in Air and Steam Environments
REFERENCE: Clark, W. G., Jr., "Evaluation of the Fatigue Crack Initia-

tion Properties of Type 403 Stainless Steel in Air and Steam Environments,"
Fracture Toughness and Slow-Stable Cracking, ASTM STP 559, American
ABSTRACT: Fatigue crack initiation data were generated for Type 403
martensitic stainless steel in laboratory air and oxygenated steam environ-
ments. Tests were conducted with 1-in.-thick blunt notched compact tension
specimens and the results analyzed and expressed in terms of both fracture
mechanics and maximum stress parameters.
KEY WORDS: fatigue (materials), crack propagation, steels, environ-
mental tests, air, steam, corrosive environments, stainless steels, stresses
Existing linear-elastic fracture mechanics technology provides a unique

quantitative approach to the evaluation of the fatigue crack growth rate
properties of structural materials. Specifically, it has been shown that the
rate at which a pre-existing crack grows as the result of cyclic loading is
dependent upon the cyclic range of the applied stress intensity factor,
• 2. From knowledge of the fatigue crack growth rate properties ex-
pressed in terms of fracture mechanics parameters and knowledge of the
critical failure conditions for a given structure, it is possible to accurately
determine the number of loading cycles required for a pre-existing crack
to grow to the critical size for failure[2]. However, many structures are
required to withstand an almost infinite number of loading cycles and the
t Fellow engineer, Mechanics Department, Westinghouse Research Labs., Pitts-

burgh, Pa. 15235.
paper.
205
Copyright9 1974 by
Downloaded/printed by ASTMlntcrnational www.astm.org
useful life of such a structure is dependent primarily on the fatigue crack

initiation properties of the material rather than the crack growth rate
behavior. For such high cycle applications, once a fatigue crack has
initiated, the time required for the crack to grow to failure is generally
relatively short and crack initiation behavior becomes a primary design
consideration.
Over the years, many attempts have been made to characterize the
fatigue crack initiation properties of structural alloys[3]. Although sev-
eral techniques have been developed which have proven invaluable in
the evaluation and prediction of fatigue crack initiation behavior, none of
these techniques can be considered truly quantitative in nature. Many
limitations associated with these techniques are the direct result of inade-
quate stress analysis procedures and the failure to adequately distinguish
between crack initiation and growth behavior.
Recent developments and advances in the areas of fracture mechanics
and stress analysis now provide techniques whereby it is possible to more
accurately analyze the stress conditions associated with crack initiation
behavior. Consequently, it appears feasible to develop a more quantitative
approach to this phenomenon. Recent studies have produced substantial
evidence which indicates that the stress intensity concept of fracture
mechanics technology may apply to the evaluation of fatigue crack initia-
tion behavior in the presence of notches[4,5,6]. Since the stress intensity
factor is essentially a measure of the magnitude of stresses in the vicinity
of a stress concentration and since crack initiation obviously depends upon
these stresses, it is reasonable to assume that a relationship may exist
between the applied stress intensity factor and crack initiation behavior.
Data exist which show that such a correlation does in fact exist[4,5,6].
In this study, an experimental program was undertaken to evaluate the
applicability of a linear elastic fracture mechanics approach to the evalua-
tion of the fatigue crack initiation behavior of Type 403 martensitic stain-
less steel. High cycle fatigue crack initiation data were generated for Type
403 stainless steel in both room temperature air and 212~ oxygenated
steam environments. All tests were conducted with blunt-notched, 1-in.
thick compact tension specimens and the result expressed in terms of both
maximum stress and fracture mechanics parameters.
Material
The material involved in this investigation was Type 403 martensitic
stainless steel supplied as 1.5 in. by 3.25 in. by 5 ft long forged bar stock.
The bar stock was austenitized at 1760~ for 21/~ h, oil quenched,
tempered at 1225 ~ for 4 h, and air cooled. The chemical composition and
CLARK ON AIR AND STEAM ENVIRONMENTS 207
room temperature mechanical properties of this material are summarized

in Tables 1 and 2.
TABLE l--Chemical composition, weight percent.
C Mn P S Si Cr Ni Mo Cu Fe
0.10 0.44 0.016 0.028 0.26 12.49 0.25 0.45 0.08 balance
TABLE 2--Mechanical properties at 70~
0.2 Yield Ultimate Elongation Reduction Charpy

Strength Strength in 2 in. in Area Impact Energy
93.5 ksi 110.5 ksi 22% 68.4% 102 ft-lb
Air Environment Tests
All fatigue crack initiation testing involved in this investigation was con-
ducted with 1-in.-thick WOL (wedge-opening-loading) type compact
tension specimens modified for crack initiation studies. Two specific test
geometries were involved. One geometry included a single machined notch
with a radius of either 0.010 in., 0.040 in., or 0.188 in. A sketch of
this test specimen showing the various notch geometries is presented in
Fig. 1. Note that the total notch depth as measured from the centerline of
loading for each notch geometry is 1.000 in. Figure 2 shows a photo-
graph of one set of the single notch specimens. The single notch specimens
were used to develop both fatigue crack initiation and growth data. Once
a crack had developed at the notch radius, the test specimen was then
used to generate fatigue crack growth rate data.
The other test specimen used in this study is shown in Figs. 3 and 4.
This geometry includes three different notch radii (0.010 in., 0.050 in.,
and 0.100 in.) in the same specimen. Once crack initiation is complete at
the first radius, the crack is grown through to the first drilled hole and
another initiation test conducted. This procedure is repeated for the next
hole. Preliminary tests with this multiple radius specimen indicate that for
the applied loads used in this study, the presence of the different holes
does not have a significant effect on crack initiation behavior. In tests with
this specimen, the load was adjusted such that the maximum stress or
stress intensity level associated with each separate initiation test remained
the same or increased progressively from the first radius to the third
FIG. I - - W O L type compact tension crack initiation test specimen.
FIG. 2 - - W O L type compact tension crack initiation test specimens (r=O.OlO,

0.040, and 0.188 in.).
3.200
2.550
F 0.500 Dia.
)
--- 1.000 .450
T
40~ inc
~1 ~ - -
II
- ~j ~ . 0 5 0 r ~k-
0100r
II
II
II
II
~--0, 625-18 Thd-3

O.625 Deep
FIG. 3--Multiple radius crack initiation test specimen.
radius. In no case was the maximum stress on the preceeding radius

higher than that developed on the next radius. A detailed photoelastic
study of this multiple radius specimen is currently underway to establish
the limits of applicability[7]. Obviously, one way to eliminate any problem
that may arise due to the multiple holes is to machine one hole, generate
the initiation data and then drill the next hole. However, it is usually
less expensive as well as less time-consuming to have all holes drilled
initially.
All testing involved in this study was conducted with specimens
machined from the "as received" bar such that the cracks grew in the
3.25-in. direction (transverse direction). The notch radii or drilled holes
were all honed (to a finish of at least 16 ~in.) prior to testing. In all
cases the notch radii were examined carefully prior to testing and any
specimens which revealed tool marks were returned for further honing.
All crack initiation testing was conducted on constant-load hydraulic
fatigue machines under sinusoidal tension-tension loading conditions at a
maximum stress ratio, R=crmin/cr.... of 0.1 or less and a test frequency
of 30 Hz.
FIG. 4----Multiple radius crack initiation test specimen (r~_O.010, 0.050, and
0.100 in.).
Detection of Crack Initiation

In most cases, the number of loading cycles required to initiate a fatigue
crack was determined with three different monitoring techniques. One
procedure involved visual examination of the notch radius with a • 30
binocular microscope positioned to look directly down the slot. Another
procedure involved visual examination of the notch radius on both sides
of the specimen with • 45 telescopes. In both visual techniques, a strobe
light was used to aid in the detection of cracking. The third monitoring
technique involved the use of an ultrasonic procedure similar to that used
to monitor fatigue crack growth behavior[8]. In this case, a 10-MHz,
1A-in.-diameter ultrasonic transducer is positioned on the top of the test
specimen such that the sound beam enters the specimen perpendicular to
the plane of the notch. The ultrasonic instrument (Sperry Type UM Re-
flectoscope) is adjusted to obtain a 2-in. sweep to peak second back
reflection signal through the solid portion of the specimen. At this tuning
level, the transducer is positioned just ahead of the radius to be examined
such that a 0.2-in. sweep to peak signal is obtained from the uncracked
radius. (The exact position of the transducer with respect to the radius
being studied varies slightly with the different radii.) The crack initiation
test is begun with the transducer positioned as just described. When
cracking takes place, the initial ultrasonic signal from the radius begins to
increase in amplitude. The number of cycles required for the initial 0.2-in.
ultrasonic signal to grow to 0.3 in. was used to define Ni, the number of
cycles required to initiate a crack. Although this point was arbitrarily
selected to describe N~, the crack initiation behavior was monitored con-
tinuously until the ultrasonic signal grew to 2.0-in. sweep to peak. Ex-
amination of broken specimens showed that an increase in the ultrasonic
signal from 0.2 to 0.3-in. sweep to peak corresponds to the develop-
ment of a crack less than 0.030-in. long and less than 0.005-in. deep
on the notch radius surface. In most cases, the ultrasonic procedure
indicated the development of a crack just before any cracking or damage
could be observed with the binocular microscope positioned to scan the
surface of the notch radius. An increase in the ultrasonic signal from
0.2-in. sweep to peak to 2.0 in. was found to correspond to a crack
entirely across the specimen thickness (1.0 in.) and about 0.150-in. deep.
A comparison and further discussion and analysis of the various tech-
niques used to detect crack initiation in this study is presented in the
results section.
Steam Tests
The oxygenated steam environment tests included in this study were
conducted with a technique similar to that used for the air tests. However,
the specimens were tested in an enclosed environment chamber and the
detection of crack initiation was limited to the ultrasonic procedure just
described. Figure 5 shows the test chamber as well as the ultrasonic
crack detection instrumentation. All steam tests were conducted with
saturated steam at 212~ and 1 atm pressure. The oxygen content of the
steam was 40 ppm. The high oxygen steam was prepared by boiling
water from a make-up tank which contained distilled water with an oxygen
gas over pressure of 2 psig.
Experimental Results
Evaluation of Crack Detection Techniques
Table 3 presents a summary of crack initiation data generated with
the three crack detection procedures involved in this study. Note that the
ultrasonic procedure is the first technique to indicate the development of
cracking. Visual examination of the notch radius surface yields results
bo
I,o
T A B L E 3--Comparison of techniques used to determine fatigue crack initiation. r
o- r
Cycles to Crack Initiation Z
Visual Examination
z
Specimen Notch ap, AKI, Ultrasonics Notch
Identification Radius, r lb ksix/in. Technique Surface Sides o~
8988-B55 0.010 2 800 14 76 100 121 000 237 800

5
8988-B56 0.010 2 800 14 92 100 120 000 199 200
8988-B57 0.010 2 800 14 108 500 113 100 234 900
8988-B58 0.010 3 800 19 24 300 25 600 36 300
8988-B59 0.010 4 800 24 6 600 6 000 12 000
8988-B61 0.038 4 750 23.8 133 800 133 800 174 600
8988-B62 0.038 3 800 19 5 400 000 5 530 000 5 800 000
8988-B63 0.038 6 800 34 50 700 60 000 70 000
8988-B67 0.189 10 800 54 45 000 54 000 60 000
8988-]368 0.189 10 800 54 45 000 50 000 53 000
8988-B69 0.189 9 800 49 120 000 121 200 122 000
FIG. 5--Environment chamber used lor fatigue crack growth rate testing.
similar to that obtained with the ultrasonic procedure. However, results

obtained by visual examination of the notch radius on the sides of the
specimen are substantially different in many cases than the data generated
with the other two techniques. This difference in results is due to the fact
that the fatigue cracks always grew deeper in the center of the specimen
than at the sides. Figure 6 shows a typical fracture surface of a crack
initiation specimen. Note the bowed shape of the initial crack. After ap-
proximately 0.100 in. of growth the crack fronts began to straighten out.
Further examination of the data in Table 3 shows that the difference in
the namber of cycles between the first development of cracking and the
FIG. 6--Typical fracture appearance of multiple radius crack initiation test speci-
men (Type 403 stainless steel).
first observation of cracking on the sides of the specimen depends on the

notch radius as well as the applied load. This behavior is not unexpected
since for the small radii and low loads, the time required for the first
crack to grow to the surface of the specimen is relatively long because of
the low • level involved. However, for the large radii and high load
tests, the number of cycles required for the first crack to grow to the sur-
face is relatively small because of the associated high • level. Because
of the dependence of side measurements on the notch radius and applied
load, this technique of detecting crack initiation was not considered ade-
quate for this study. For specimens substantially less than 1-in. thick, the
correlation between the first development of cracking and visual measure-
ments on the sides of the specimen may be adequate.
In view of this analysis of crack detection techniques, it was concluded
that either visual measurements of cracking on the notch radius surface or
measurements made with the ultrasonic technique were completely ade-

quate. However, since visual examination of the notch radius surface
could not be adapted to enclosed environment testing, the ultrasonic pro-
cedure was selected for all subsequent crack initiation testing. Except for
Table 3, all crack initiation data reported here refer to data generated
with the ultrasonic procedure described previously.
Crack Initiation Data

Table 4 presents a complete summary of all the pertinent data and
results associated with the crack initiation portion of this investigation.
The nominal stress range values, A~N, reported for each test were com-
puted from the following nominal stress expression for compact tension
specimens[9].
a~ aP
(W-a) I 1+3 - W+a-1
-~._] (1)
where AP is the applied load range, B is the specimen thickness, a is the
notch depth measured from the centerline of loading, and W is the speci-
men width also measured from the centerline of loading. The maximum
applied stress range, A~. . . . was determined from the generalized stress
intensity expression which shows the effect of blunt notches on K~;
where[10]
,- /
AKI= lira -YhY-~/7~m.x (2)

r---->0 z
and
Ao'max~-- 2__ 1 AKI (3)

V~ Vr
The elastic stress concentration values, Kt, reported for each notch con-
figuration were determined from A~maxdivided by Aex.
Although Eqs 2 and 3 are considered exact only when r approaches
zero and thus, cannot be arbitrarily applied to large radii, Wilson and
Gabrielse have shown as the result of a detailed finite element analysis of
blunt notches in compact tension specimens that this expression is accurate
to within 10 percent for notch radii up to 0.180 in.[//]. (The accuracy
for notches smaller than 0.040 in. approaches 4 percent.)
The stress intensity range parameters, AKx, reported in Table 4 and
also those used in Eqs 2 and 3 were determined from the crack-notch
stress intensity expression for the compact tension specimens where[12]
-- - PmaxV'a Pn,inVa
zkKI~-KImax-l~'min= r B-W Y BW (4)
f'o
O,
ci
8t -
O
"r
Z
TABLE 4 - - S u m m a r y o/ /atigue crack initiation data generated ]or Type 403 stainless steel.
Maximum
Alternating Alternating
Specimen Notch Notch Nominal Applied __ Cycles to
Identifi- Radius, Depth, Applied Stress, Stress, ~Kx, ~/r, aKi, Initiation,
cation r, in. a, in. Kt Load, lb Attar, k s i Atrmax, k s i ksiqin__ in. qr N~
Single Notch Specimens

8988-B55 0.010 1.00 11 200to 3000 14.2 156 14.0 0.1 140 76100
8988-B56 0.010 1.00 11 200 to 3 000 14.2 156 14.0 0.1 140 92 100
8988-B57 0.010 1.00 11 200 to 3 000 14.2 156 14.0 0.1 140 108 500 t~
8988-B58 0.010 1.00 11 200to 4000 19.4 214 19.0 0,1 190 24300
8988-B59 0.010 1.00 11 200to 5000 24.5 270 24.0 0.1 240 6600 t')
8988-B60 0.038 1.00 5.6 200 to 3 000 14.2 80 14.0 0.195 72 16 190 600 -~ Z
8988-B61 0.038 1.00 5.6 250to 5000 24.2 135 23.8 0.195 122 133800 0
8988-]362 0,038 1.00 5.6 200 to 4 000 19,4 108 19.0 0,195 97.5 5 400 000
8988-B63 0.038 1.00 5.6 200 to 7 000 34.6 194 34.0 0.195 174 50 700
8988-B64 0.038 1.00 5.6 200 to 4 500 22.0 123 21.5 0.195 110 180 000
8988-B65 0.189 1.00 2.6 200 to 5 000 24.5 68 24.0 0.434 55.4 26 300 000 ~
8988-B66 0,189 1.00 2.6 200 to 8 500 42.3 110 41.5 0,434 96.0 4 800 000 ~
8988-B67 0,189 1.00 2.6 200 to 11 000 55.0 142 54.0 0,434 124 54 000
8988-B68 0,189 1.00 2.6 200 to 11 000 55.0 142 54.0 0.434 124 45 000
8988-B69 0.189 1,00 2.6 200 to 10 000 50.0 130 49.0 0,434 113 120 000
Multiple NotchSpecimens
8988-B85 0.009 1.00 11 200 to 2 400 11.2 123 11.0 0.095 113 234 000
0.050 1.50 4 300 to 3 300 33.0 132 25.7 0.225 114 113 000
0.100 2.00 1.7 200 to 2 I00 86.0 146 41.0 0.316 130 167 000
8988-B86 0.009 1.00 11 200to 2400 11.2 123 11.0 0.095 113 257000
0.050 1.50 4 300to 3300 33.0 132 25.7 0.225 114 180000
0.100 2.00 1.7 200 to 2 250 92~0 156 43,0 0.316 136 75 000
8988-B87 0.009 1.00 11 200 to 2 000 9.2 101 9.0 0.095 95. 600 000
0.050 1.50 4 200 t o 2 600 26.5 106 20.5 0.225 91 24 000 000 -~
4 200 to 2 800 27.6 110 22.4 0.225 99 10 000 000 -'
4 200to 2900 29.8 119 23.0 0.225 102 205000
8988-B93 0.038 1.00 5.6 200 to 8 000 39.8 222 39.0 0.195 200 11 160
0.050 1.50 4 200 to 5 600 60.0 240 46.5 0.225 206 14 400
0.I00 2.00 1.7 200 to 3 100 130.0 220 62.0 0.316 196 8 640
8988-B90 0.009 1.00 11 200 to 6 000 29.5 325 29.0 0.095 300 4 800
0.050 1.50 4 2 0 0 to 8 000 86.0 345 67.0 0.225 300 4 800
0.100 2.00 1.7 200 to 4 600 200,0 340 95.0 0.316 300 Specimen
failure, 1 cycle
Oxygenated Steam Tests, 212~
8988-B88 0.009 1.00 11 200 to 2 400 11.2 123 11.0 0.095 113 630 000
0.050 1.50 4 200 to 3 000 31.0 124 24.0 0.225 105 795 600 N
0.100 2.00 1.7 200 to 2 100 86.0 146 42.0 0.316 134 21 600 7.
8988-B94 0.038 1.00 5.6 200 to 4 200 20.5 114 20.0 0.195 102 200 000
0.050 1.50 4 200 to 3 700 38.5 154 30.0 0.225 134 108 000 O
0.100 2.00 1.7 200 to 2 300 95.0 162 45.0 0.316 145 27 000 Z
8988-B90 0.009 1.00 11 200 to 2 000 9.2 101 9.0 0.095 95 6 840 000 _>
0.050 1.50 4 200 to 2 600 26.5 106 20.5 0.225 90 12 000 000 -~
0.100 2.00 1.7 200 to 1 700 67.5 114 32.0 0.316 100 2 000 000 ),
Z
~7
T,
<
O
X
-,4
in this case, a is the notch depth measured from the centerline of loading,
P ..... and Pmin are the maximum and minimum loads per cycle, B is the
specimen thickness, W is the specimen width, and Y is a compliance
constant which depends upon the notch depth.
Figure 7 presents the number of cycles required to initiate a fatigue
crack versus the parameter AK~/Vr. Figure 8 presents a similar curve
expressed in terms of maximum applied stress range, A~r..... Note that
relatively little data scatter was encountered for the various radii tested.
Note also that the steam environment does not have a significant effect
on the number of cycles required to produce a crack. From the data
shown in Fig. 7, it is apparent that for the number of cycles involved in
this study, no crack initiation occurred when AKI/Vr was below 95 ksi.
This value of 6K~/Vr corresponds to a maximum stress range of approxi-
mately 100 ksi (Fig. 8). Thus, it appears that for the Type 403 stainless
steel involved in this study, no fatigue cracks develop at maximum applied
stress levels below about 100 ksi. At maximum stress levels about 100 ksi,
the number of cycles required to initiate a fatigue crack can be estimated
from Fig. 8.
In addition to generating fatigue crack initiation data in this study, an
attempt was also made to evaluate the effect of the notch radius on
fatigue crack growth rate behavior immediately ahead of the notch radius.
However, because cracking initiated as a point defect and eventually
grew to a small semielliptical surface defect on the notch radius surface,
it was not possible to accurately measure the flaw depth. In addition,
for the case of the larger radii, several crack initiation sights were
developed before a single crack began to grow to a significant size. Even
if the size and shape of the first cracking could be established, it is un-
likely that existing stress intensity expressions would be adequate to
analyze such small surface defects. Thus, it was not possible to develop
meaningful data regarding the effect of existing stress concentrations on
early crack growth. However, it was noted that when the initial crack had
grown across the thickness of the specimen (corresponding crack depth
approximately 0.150 in.) subsequent fatigue crack growth rate behavior
was identical to that generated with the more conventional fatigue pre-
cracked specimens.
Discussion
The experimental results generated in this study clearly show that a
correlation exists between either AK~/Vr or A~r.... and the number of
loading cycles required to develop a fatigue crack in a notched body.
However, many questions exist with regard to the quantitative nature and
1000 n I I I I I I I I I I I I L I I I [ I L I I l I I [ I I I" I I 1 I
Type403 Stainless Steel

oys = 93,500 psi
IT WOL Specimens
R=O.I
Test Frequency30 cps
r = Notch Radius
Room Temp., Air Environment
0 & 9 0 / AKI\
L~ ,00--
~oo ~|
l'O
Notch Radius, r, in.
9 0.010
o - 0.038
", - 0.050 Cycles to Initiation Determined
9 - 0. I00 with Ultrasonic Instrumentation
a - 0.289
Steam Environment (212~ 1 Atm.)
I I I I L I II I I I I h I L L I I I I I I I I I I I I L I L I J I I t i L l
I~ 104 105 106 10? 108
Ni, Cycles to Crack Initiation
F I G . 7--Cycles to initiation as a function oJ A K z / x / r Jor Type 403 stainless steel in air and steam environments.
1000 I L L l l I ~ I 1 I I I I I 1 1 I I I I I I I
Type 403 Stainless Steel

oys = 93,500 psi
IT WOL Specimens
R=0.1
x" Test Frequency 30 c ~
r = Notch Radius
o 9
o
F~oom Temp., Air Environment
no I9 l( ) lOO i
%0"~ |174 ax th
I00--
Notch Radius, r, in. t~
9 - 0.010
E
== o - 0.038
x
- 0.0f,0
Cycles to Initiation Determined
9 - 0.100
with Ultrasonic Instrumen~tion
- 0.189
| Steam Environment (212~ 1 Atm.)
lO03~ i ~ i i k i la104 i i i i t I L I105 i x i i h i ~s106 i ~ J L i , L 107 t ~ I I L J ~al

08
Ni, Cycles to Crack Initiation
FIG. 8--Cycles to initiation as a/unction o /ma xi m u m alternating stress/or Type 403 stainless steel in air and steam environments.
applicability of this empirical relationship. It is encouraging as well as

somewhat surprising to find that the stress intensity concept which is based
on cracked-body analyses can also be used to estimate the stresses asso-
ciated with rather large radii.
Wilson and Gabrielse have noted that their finite element analysis of
compact tension specimens with blunt notches was limited t o two-dimen-
sional plane strain conditions and they suggest a detailed three-dimen-
sional elastic-plastic analysis to further evaluate fatigue crack initiation
behavior in notched bodies[//]. They also point out that further studies
regarding the effect of the degree of triaxiality and stress gradient on
fatigue crack initiation behavior are required to permit the development
of a quantitative approach to the evaluation of this behavior. Obviously, an
extensive amount of additional work is required in these areas of tech-
nology. The effect of material, loading spectrum, geometry, section thick-
ness, and surface finish on fatigue crack initiation behavior are some of
the areas that must be investigated further.
The results of this study are presented in terms of both a fracture
mechanics parameter, aK~/V~ and a more conventional maximum al-
ternating stress parameter, Aa. . . . The most convenient form of presenting
such data will obviously be the users choice. However, it must be noted
that many of the older stress analysis techniques available for computing
nominal stress values and for selecting Kt parameters for a given geometry
are not as accurate as the more recent stress analysis techniques used to
develop stress intensity parameters. Thus, it is suggested that when
possible, the A~m~xvalues used to interpret fatigue crack initiation behavior
be computed from the appropriate stress intensity expressions (if one is
available) such was the case in this analysis.
It is interesting to note from the results of this study, that no cracks
developed at applied maximum stress levels below about 100 ksi. Upon
first examination, it may appear that such results contradict the more
conventional endurance limit approach to predicting fatigue life (where
the 108 cycle endurance limit is approximately one half the ultimate
tensile strength, 55 ksi for the Type 403 stainless steel studied here).
However, it must be kept in mind that conventional endurance limit data
are usually generated with R. R. Moore fatigue specimens which are
subjected to complete reversed loading. Therefore, an endurance limit of
55 ksi reported as the result of R. R. Moore fatigue testing actually
represents a maximum applied stress range of about 110 ksi, similar to
the A~maxthreshold level reported here (Fig. 7).
Although the endurance data generated here appear consistent with
more conventional fatigue test results, it is very unlikely that data gen-
erated above the endurance limit will be the same because of the large
differences in the degree of triaxiality and stress gradient between the
small R. R. Moore specimen (approximately 0.25-in. diameter) and
the relatively large 1-in.-thick compact tension specimen.
Although the results of this study show that a steam environment does
not have a significant effect on the fatigue crack initiation behavior of
Type 403 stainless steel, this observation must be qualified with regard to
the duration of the test. The longest run steam tests involved in this study
took approximately two weeks to complete. Thus, the maximum ex-
posure time in the steam environment was no more than two weeks. It is
possible, however, that if the exposure time were much longer, general
corrosion may take place at the notch radius surface, thus changing the
surface finish of the radius and resulting in the development of a fatigue
crack in less cycles than that predicted here. Thus, it is obvious that total
exposure time in the environment must be considered in the development
of meaningful environmental fatigue crack initiation tests.
Summary
An evaluation of the fatigue crack initiation properties of Type 403
stainless steel has shown that the number of loading cycles required to
initiate a fatigue crack in a notched body can be described in terms of
existing linear elastic fracture mechanics technology. Specifically, it has
been shown that the localized maximum applied alternating stress as
computed from fracture mechanics considerations is the primary factor
controlling the number of cycles required to initiate a fatigue crack in the
presence of a blunt notch. In view of these observations, it appears that a
fracture mechanics approach to the evaluation of fatigue crack initiation
behavior may ultimately lead to a unique quantitative design procedure
for predicting the number of loading cycles required to develop a fatigue
crack in a structural component. Such a design procedure combined with
existing techniques available for predicting the growth of fatigue cracks
will permit the accurate analysis of the total cyclic life of structural
components. However, a substantial amount of additional testing is re-
quired to establish the applicability of a fracture mechanics approach to
fatigue crack initiation. Several areas of additional research are suggested
in this paper.
The data generated in this study show that an oxygenated steam en-
vironment (212~ 1 atm) does not have a significant effect on the
fatigue crack initiation behavior observed for Type 403 stainless steel.
However, it has been noted that the total exposure time in the test environ-
ment may ultimately affect such test results if sufficient general corrosion
occurs to alter the surface finish of the crack starter notch.
This paper also describes a unique approach to the detection of

fatigue crack initiation behavior in blunt notched compact tension speci-
mens. The technique involves the use of commercially available ultrasonic
nondestructive testing instrumentation. This ultrasonic technique is ex-
tremely accurate and is applicable to crack initiation testing in enclosed
environments.
Conclusions
1. Existing linear elastic fracture mechanics technology can be used to
describe the fatigue crack initiation properties of Type 403 stainless steel.
2. The number of loading cycles required to initiate a fatigue crack is
directly related to the fracture mechanics parameter AKI/Vr which in
turn, is related to the maximum alternating stress in the vicinity of the
crack starter notch.
3. Fatigue cracks did not initiate in Type 403 stainless steel when the
AK1/Vr level was less than 95 ksi or when the maximum applied al-
ternating stress was less than 100 ksi.
4. An oxygenated steam environment at 212~ and 1 atm pressure
did not influence the fatigue crack initiation properties of Type 403
stainless steel.
5. A multiple radius compact tension crack initiation test specimen has
been developed which yields three separate crack initiation tests per
specimen.
6. A unique ultrasonic procedure for monitoring fatigue crack initia-
tion behavior in compact tension specimens has been developed. This
technique can be used for enclosed environment testing.
Acknowledgments
The use of blunt notched compact tension specimens to study fatigue
crack initiation behavior was first suggested to the author by B. F. Langer
of Westinghouse Nuclear Energy Systems[13].
The author gratefully acknowledges the contributions of R. R. Hovan
of the Westinghouse Research Mechanics Department, and J. R. Malley
and D. Peters of the Materials Testing and Evaluation Laboratory to the
various phases of the experimental program.
The continued interest and support provided by the Westinghouse
Marine Division made this work possible.
References
[1] Johnson, H. H. and Paris, P. C., "Subcritical Flaw-Growth," Engineering Frac-
ture Mechanics, Vol. 1, No. 3, 1968.
[2] Clark, W. G., Jr., "Fracture Mechanics in Fatigue," Experimental Mechanics,
Sept. 1971.
[3] Wundt, B. M. in Effects of Notches on Low-Cycle Fatigue, ASTM STP 490,

American Society for Testing and Materials, May 1972.
[4] Jack, A. R. and Price, A. T., "The Initiation of Fatigue Cracks from Notches
in Mild Steel Plates," International Journal of Fracture Mechanics, Vol. 6,
No. 4, Dec. 1970.
[5] Forman, R. G., "Study of Fatigue Crack Initiation from Flaws Using Fracture
Mechanics Theory," 4th National Symposium of Fracture Mechanics, Aug. 1970.
[6] Toolin, P. R., unpublished Westinghouse Research Laboratories data.
[7] Johnson, R., unpublished Westinghouse Research Laboratories data.
[8] Clark, W. G., Jr. and Ceschini, L. L, "An Ultrasonic Crack Growth Monitor,"
Materials Evaluation, Aug. 1969.
[9] Wessel, E. T., Clark, W. G., Jr., and Wilson, W. K., "Engineering Methods for
the Design and Selection of Materials Against Fracture," DDC Report AD
801001, 1966.
[I0] Paris, P. C. and Shi, G. C., "Stress Analysis of Cracks," Symposium on Tough-
ness Testing and Its Applications, ASTM STP 381, American Society for Testing
and Materials, 1965.
[11] Wilson, W. K. and Gabrielse, S. E., "Elasticity Analysis of Blunt Notched Com-
pact Tension Specimens," unpublished Westinghouse Research Laboratories data.
[12] Wilson, W. K., "Stress Intensity Factors for Compact Tension Specimens," West-
inghouse Research Memo 67-1D6-BTLFR-M1, June 1967.
[13] Langer, B. F., "Proposed Research on Low Cycle Fatigue of Notched Members,"
Westinghouse Electric Corp., Research Memo 67-IDO-CODES--M1, 31 Oct.
1967.
F. H. Gardner ~ and R . 1. S t e p h e n s 2
Subcritical Crack Growth Under Single

and Multiple Periodic Overloads in
Cold-Rolled Steel
REFERENCE: Gardner, F. H. and Stephens, R. I., "Subcrifical Crack

Growth Under Single and Multiple Periodic Overloads in Cold-Rolled
Steel," Fracture Toughness and Slow-Stable Cracking, ASTM STP 559,
ABSTRACT: Fatigue crack growth retardation, delayed retardation, initial
acceleration, and optimum fatigue crack growth life under single, single
periodic, and multiple periodic tensile overloads were investigated in cyclic
strain softening AISI 1020 cold-rolled steel using compact type specimens.
Macroscopic delayed crack growth retardation was observed following both
single and multiple periodic overloads and was best observed at high over-
load ratios and larger crack lengths. Initial crack growth acceleration at the
beginning of an overload interval was not quantitatively evident, however
initial acceleration could have occurred before the crack extended to the
least measurable scale division. Under constant load range testing, multiple
periodic overloads of about 10 to 30 gave optimum or maximum fatigue
life in most cases, however values of 1 gave maximum life in some cases.
The largest increase in fatigue life was 453 percent. Maximum or optimum
fatigue life obtained from multiple periodic overloads was always greater
than that obtained from a single overload. In general, the results suggested
maximum fatigue life can be obtained by applying a few multiple tensile
overloads at high overload ratios only when accelerated crack growth begins
following low load crack retardation.
KEY WORDS: fractures (materials), mechanical properties, fatigue (mate-
rials), stresses, crack propagation, plastic zone, delayed retardation, residual
stress, cold-rolled steel, striations
The influence of periodic single and multiple peak tensile overloads on

total fatigue life and fatigue crack growth behavior is not well estab-
1Project engineer, Mid Equipment Co., Inc., Grundy Center, Iowa 50638.
Professor, Mechanical Engineering Department, The University of Iowa, Iowa
City, Iowa 52242.
225
Copyright*1974by
lished. Research in the 1960's and early 1970's determined that ap-
preciable crack growth retardation can occur following tensile overload-
ing[1-25]. 3 Crack growth retardation as shown in Fig. la has occurred
following overloading[i,3,7,8,14,20,22,25] and delayed retardation as
shown in Fig. lb has also occurred following overloading[I,13,18,19,22].
The presence of delayed retardation appears to be dependent on the
overload plastic zone size, the ratio of the overload stress intensity to
the lower load stress intensity and the least reading of the crack moni-
toring system. The first two imply the importance of an effective stress
intensity obtained from crack closure concepts[10,13,16,19]. Delayed
retardation has been found in both cyclic strain hardening and cyclic
strain softening materials[18,22]. During multiple overloading, higher
paper.
iI .
.-I "1II
"~
(..)L~RETARDATION--~
CYCLES
(a ) 1 i Z I l l --
i,.,.j~I[~RETARDATION-~
, //, J
o CYCLES
(b) . _ i ~ _
T,ME
(c) - ~ i
FIG. 1 - - ( a ) Crack retardation, (b) delayed retardation, and (c) loading spectrum.
GARDNER AND STEPHENS ON PERIODIC OVERLOADS 227
crack growth rates have been observed in the early cycles relative to
that of later cycles[15,19]. This higher crack growth rate has been called
initial acceleration.
Several models have been used to explain observable crack growth
behavior during and following single or multiple overloading, namely:
crack tip blunting[2], residual stresses ahead of the crack[3], and crack
closure[10]. Raju et all20] have shown that crack growth retardation fol-
lowing overloading could be eliminated by annealing at elevated tem-
peratures. Since crack tip blunting would not be eliminated by temperature
effects, it was shown that blunting is not the significant factor in crack
growth retardation. The main effects that can be removed by annealing are
residual stresses. This affects not only residual stresses ahead of the
crack, but also crack closure. Several papers have recently qualitatively
explained delayed retardation following overloads and initial acceleration
during multiple overloading using the crack closure model and an effective
stress intensity[10,16,19]. It is the objective of this paper to further
investigate and clarify crack growth retardation, delayed retardation,
initial acceleration, and optimum fatigue life under single, single periodic,
and multiple periodic tensile overloads.
Material and Experimental Procedures

The material used for this investigation was cyclic strain softening
cold-rolled AISI 1020 steel from the same stock as used by Vargas and
Stephens[22]. The monotonic tensile ultimate strength, 0.2 percent yield
strength, and reduction in area obtained in the rolled direction was 97.5 ksi,
91 ksi, and 62 percent, respectively. The ASTM grain size number was
approximately 8 and 7 in the transverse and longitudinal directions,
respectively.
All overload crack growth tests were conducted at room temperature
using a prefatigue cracked compact type specimen shown in Fig. 2.
Specimens were machined from 1,4-in. plate in the cold-rolled condition
and loaded in the direction of rolling. A chevron starter notch was used
in all specimens to ensure an initial uniform crackfront. Side grooves
0.0125-in. (5 percent of specimen thickness) deep with a 88 radius
were milled on both sides of the specimen to control the plane of crack
extension. The side grooves were polished to facilitate macroscopic crack
growth observations. Crack growth was monitored with a traveling, opti-
cal microscope at x 36 magnification using stroboscopic illumination. A
scale with a least division of 0.01 in. was taped to the specimen and
positioned parallel to the side grooves such that the crack tip and scale
could be observed simultaneously. Tests were performed on a 20 000-1b
228 FRACTURE
TOUGHNESSANDSLOW-STABLE
CRACKING
_i
~:" i i I_~~
4.0 ~'1
g
g
rd
0.78D I A . ~
,.2
0.78 DIA.
L
FIG. 2--Compact specimen dimensions, (in.).
capacity, closed-loop electrohydraulic test system using load control and

positive haversine waveform.
The fatigue loading pattern is shown schematically in Fig. lc. The
steady-state load interval, designated with a subscript ~, always refers to
the lower load level, while the overload interval, designated with a sub-
script 0, always refers to the high load level. Constant steady-state load
ranges, Ap~, were applied at 15 Hz while constant overload ranges, •
were applied at 1.5 Hz. The minimum load, Pmi,, for both steady-state
loading and overloading was always 200 lb. Constant load ranges rather
than constant stress intensity ranges were specifically chosen to closer
simulate actual situations.
The test program consisted of steady-state tensile loading •
3.0, or 3.5 kips, with superimposed tensile overloading aP0=5.0 or 6.0
kips. The load ratio R = m i n i m u m load/maximum load varied from
approximately 0.03 to 0.07 for the complete spectrum of overloads and
steady-state loads. For a given specimen, the ratio of the overload range
to the steady-state load range was always constant. This overload ratio
varied from 1.43 to 2.40 within the test program. The number of steady-
state cycles, ANs, applied between overloads was 5000, 10 000, or 20 000
cycles for a given specimen while the number of overload cycles applied
per overload interval, • was 1, 10, 30, 50, 100, or 200 cycles for a
given specimen. AN0 = 100 or 200 cycles was used only in conjunction with
ANs-----10000 cycles. Single overload and • results reported pre-
viously[22] are included for completeness. A total of 84 specimens sub-

jected to various overloads are thus included in this paper. Test parameter
combinations along with fatigue crack growth life are given in Table 1.
All specimens were fatigue precracked at the steady-state load, •
to an initial crack length, a, of i.00 in. The fatigue life then was taken as
the number of applied cycles to grow the crack from 1.00 in. to fracture.
The first overload interval was always applied at a crack length of 1.05 in.
The initial steady-state stress intensities at this crack length were 37, 44.5,
and 52 ksi~/in., for ~xP,=2.5, 3.0, and 3.5 kips, respectively. The over-
load stress intensity at the start of the first overload interval with a = 1.05
in. was 74 and 89 ksiCin, for • and 6.0 kips. The fracture tough-
ness, Ke, for this material and thickness was 95 ksi~/in, for a 1.00 in.
crack length. Thus, initial overload stress intensities were 78 and 94 per-
cent of K~ while initial steady-state stress intensities were 39, 47, and
55 percent of K~. Stress intensities were calculated from
K-B~-WI 29.6 (-~-~)v~ - 185 (-~-~) ~2
+655"7 (~-~)5/2--1017 ( ~ - ) 7/2+638"9 ( - ~ - ) ~ 1
where P is the applied load, B is the net thickness, W is the specimen

width, and a is the crack length (ASTM Test for Plane-Strain Fracture
Toughness of Metallic Materials (E 399-72) ).
Test Results
Table 1 contains the test results for no overload, single overload, single
periodic overload, and multiple overload sequences. Results are tabulated
with respect to AP,, LxP0, AN~, and ANo with fatigue life indicated. The
number of overload intervals is shown in parentheses, and a failure during
the overload interval is shown with an asterisk. Approximately 70 percent
of the specimens subjected to periodic overloads failed during the over-
load interval. Final crack length at fracture ranged from about 1.4 in. to
about 2.1 in. About 40 percent of the specimens with AN0=I, that is,
single periodic overloads, tended to have shorter crack lengths at failure
and thus lower applied stress intensities at failure. This may have been
due to normal scatter or to smaller plastic zone sizes and less crack
closure following the single overloads. In all cases, however, applied
stress intensities at failure were greater than the monotonic fracture tough-
ness Ke which can be attributed primarily to residual compressive stresses
in front of the crack and crack closure. In all specimens, except one
(~P~ = 2.5 kips, • = 6.0 kips, • = 10 000 cycles, and • = 200 cycles),
the overloading sequence increased life compared to no overloading. The
1,o
Co
0
e-
~m
TABLE 1 - - E x p e r i m e n t a l [atigue crack g r o w t h li]e.
c
O
ANs .-----5000 cycles AN, = 10 000 cycles AN~ ~ 20 000 cycles -r
Z
No Single ,ANo z~No ANo

APs, APo, Over- Over-
lb lb load load 1 10 30 50 1 10 30 50 100 200 1 10 30 50
~7
6000 48 070 105 560 (27) ~ (23)* (18) * (14)* (10) (17) * (14)* (11)* (6) * (4) * (7) (12) (12) * (9)*
2500 [89] 140430 120803 97460 76625 155800 176010 145700 116981 65950 46340 186080 266045 246830 186660
]37] a 5000 48 070 67 760 (25) (25) (23)* (19)* (10) (15)* (15)* (13)* (11)* (8)* (5) (7)* (9)* (8)* 0
[74] 125 110 127 820 113 305 101 770 96 300 157 090 157 390 136 902 115 922 87 480 86 050 145 500 186 760 165 690 d~
6000 29270 44190 (10)* (12)* (12)* (9)* (9)* (8)* (8)* (7)* (5)* (3)* (4) (5)* (4)* (5)* m~
3000 [89] 63 500 63 941 64 200 48 870 82 710 84 170 84 170 74 052 53 900 34 260 64 690 103 760 83 643 103 730
[44.5] 5000 29270 31680 (10)* (8)* (11)* (10)* (4) (6) (6) (6) (6)* (4)* (2) (2) (2)* (2)* m
[74] 53 800 43 990 59 570 53 710 42 690 63 670 64 382 63 300 63 970 44 760 37 580 40 610 42 620 44 230
6000 18610 23410 (7) (7)* (7)* (5)* (3) (4) (4)* (3) (4)* (2)* (1)
3500 [89] 36 930 37 550 38 491 27 840 31 310 39 320 42 261 32 860 43 520 22 680 23 410 . . . . . . . . . :~
[52] 5000 18610 19910 (5) (5) (5) (6) (2) (2) (2) (2)* (2)* (2)* (1) "~
[74] 24 810 27 340 27 400 31 300 20 018 21 020 25 960 23 010 23 400 23 130 19 900 . . . . . . . . . O
a [ ] Initial steady state or initial overload stress intensity, ksi 3/in.

( ) N u m b e r of overload intervals.
* Failed during overload.
maximum increase in life was 453 percent while the decreased life was 4
percent.
Typical crack length versus applied cycles for various loading sequences
are shown in Figs. 3 to 6. The open circles represent the crack length and
applied cycles at the beginning of an overload interval. The number of
overloads, ANo, within an interval has been labeled on each curve. Dotted
regions without a solid line represent crack growth during an overload
interval, and solid lines represent crack growth during the steady state or
lower load interval. Curves labeled N.O. refer to no overload specimens
and S.O. refers to single overloads with the prescribed values of • and
• An arrow terminating a curve indicates failure during the steady-state
loading and an X indicates failure during the overload interval.
In Fig. 3 it is evident that maximum life for • kips, AP0=6.0
kips, and • cycles occurred with • 1. As • increased,
the fatigue life decreased due to crack growth during the overload inter-
vals. The single overload was not as beneficial as multiple periodic over-
loads with • 1 and 10, but was more beneficial than • or 50.
In Fig. 4 the effect of AN0 for • kips, • kips, and •
5000 cycles is just the opposite to that of Fig. 3. That is, as ANo increased
from 1 to 50 the fatigue life increased, however, ANo=10 and 30 gave
2.0
~---A N o = 5 0 l/--S.O. ~
1.8 / \ r.o=,O
r-N.O.* L q'o"X'\ X "x,~,/~/ / Jm
1.6'
/ > //yj
z
W 1.4
-J
1.2
0
1.0 ~ N.O.-NO OVERLOAD

~* S.O.-SINGLE OVERLOAD
0 2'o 4 '0 60' 8'0 I00
' 120
' 140
'
APPLIED CYCLES(x I0 -a)

FIG. 3--Crack length versus applied cycles for AP.:2.5 kips, APo:6.0 kips, AN~
=5000 cycles.
2.0 ~--ANo=l
=o ~o\ ~.-~,o.,O
~
1.8
N.O.= . // / ~ A N
, . ,.
Iv
O =50
, 1.6-
"I-
I--
(..9
Z
ta
_1
1.4' Y J
v
n.- 1.2.
1.0-
I I I I I i
0 5 10 15 20 25 30
APPLIED CYCLES ( x 10 -3)
FIG. 4-iCrack length versus applied cycles [or AP==3.5 kips, APo=5.0 kips,
AN==5000 cycles.
7 - A No= 50
2.0
i/-so /-,,No=l (
/-N.o. / P( "" l/ ]~/ANo=IO
1.8
1.6
1"' U ]::7oo
_1
1.4 i
o 1.2
1.0
I I I I i I
0 10 20 30 40 50 60
APPLIED CYCLES ( x l O "3)

FIG. 5---Crack length versus applied cycles/or AP==3.0 kips, APo=5.0 kips, AN=
=10 000 cycles.
,.oi .
"I"
!
,--i/ / /-,,,o.1OO
I'-
(.9
Z
I.iJ N O= 10
--I
v
0
r 1.2_ ~ ~No= 1
1.0-
0 20 40 60 80 100 120 140 160 180
APPLIED CYCLES (x 10 "~)

FIG. 6--Crack length versus applied cycles for AP.,=2.5 kips, APo=6.0 kips,
ANs ~ 10 000 cycles.
essentially the same life. This increase in fatigue life with increased ANo is
due to the lower crack growth rates during the steady-state loading. The
single overload gave only slightly better fatigue life than no overload
and gave less life than all periodic overload sequences.
In Fig. 5 with • kips, • kips, and • 000 cycles,
it is seen that for aNo= 10, 30, 50, or 100 the fatigue life is essentially
independent of ANo. This is attributed to essentially similar crack growth
rates during loading. However, with ANo= 1 or 200, appreciably less
fatigue life occurred than with other periodic loads. Figure 6 is somewhat
similar to Fig. 3 except that due to the short crack length at fracture for
• the maximum fatigue life occurred with • 10. In all cases
crack growth rates tended to increase at the larger crack lengths during
both overloading and steady-state loading due to the increased applied
stress intensities. Figures 5 and 6 show that a larger percentage of crack
growth occurred during overloading when AN0 was large. For t, N o = 100
or 200, more than half the crack growth occurred during the overload
intervals, while at the smaller overload intervals, ANo=10 or 30, most
crack growth occurred during the steady-state load interval. The maximum
number of overload cycles for a given specimen was 1710 cycles.
In Figs. 3 to 6 it appears that delayed retardation has occurred following
some overload intervals, while in other cases delayed retardation is not

evident. The delayed retardation is more evident at the higher crack
lengths where applied overload stress intensities are quite high. Figure 7a
is an expanded scale of the last three steady-state and overload intervals
of one specimen from the fifth overload interval to fracture. The no over-
load growth behavior, N.O., is also shown for more complete comparison.
Figure 7b shows the macroscopic steady-state crack growth rates which
were calculated by the tangent method from the best-fit curve through the
data points at 0.01 in. intervals. The fifth steady-state interval shows a
/ aP~
I/
A Ps = 5
A N s : 10,OOO
1.70 A N o = 50
/'NO
F Ps = 3
1 . 5 0 I)
o 1o,;oo zo,ooo 3o,boo
APPLIED CYCLES
(a)
w'~ 30
\
I c 20.
x
o ~ 10,
[ . . . . . I
I
I I
' I I
O,~ , i | I I i i
1.~0 1.40 1.45 1.50 1.55 1.60 1.65 1.70
CRACK L E N G T H - in.
(b)
FIG. 7----Expanded scale of crack growth during steady-state loading.
constant growth rate, the sixth interval shows a small amount of delayed
retardation, while the seventh interval shows appreciable delayed retarda-
tion. This expanded scale clearly indicates the variety of crack growth
behavior which occurred following multiple overloading. An inflection
point, and thus an increase in da/dN, was also noted for some specimens
during the steady-state load interval at large crack lengths and large AN=.
Delayed retardation was previously reported[22] for single overloads and
single periodic overloads (AN,=1) and thus, the number of applied
periodic overloads is not a limiting factor for delayed retardation.
Crack growth behavior during overloading is not clear in Figs. 3 to 6
because of the small scale. Figure 8 shows an expanded scale of nine
representative overload intervals for different • AP0, AN~, AN0, and
crack lengths. Since the least scale reading for measuring crack length
was 0.01 in., the curves are shown dashed near the extrapolated origin.
These representative curves do not show the initial acceleration at the
beginning of the overload cycles as found by others[15,19]. In a few
cases, however, extrapolation indicates some initial acceleration may have
occurred. It is possible that a smaller least reading would show initial
acceleration. However, the extent of the initial acceleration would be less
than 0.01 in. During crack growth monitoring, it appeared that some
initial acceleration occurred before the 0.01 in. Matthews et al[15]
.125-
I
"I-
I.- .I00-
0
n,,
(.9
.075-
fo
oc
~o
,O50-
O
._1
rr
to
> .025-
0
0
0 5'0 I00 150 260
APPLIED OVERLOAD CYCLES
FIG. 8--Expanded scale of crack growth during overloading.
showed initial acceleration occurred in 4340 steel for greater than 0.050-in.
crack extension under constant stress intensity overload testing. Loads were
reduced, in their work, at 0.025-in. crack increments which indicates that
initial acceleration is not just a loading procedure phenomenon.
A scanning electron microscope (SEM) was used in an attempt to
correlate the observed macroscopic crack growth rate with microscopic
striation spacing. The large number of periodic overloads, however,
made it very difficult to find fatigue striations as observed by Vargas
and Stephens[22]. Less than about 0.01 percent of the surfaces contained
fatigue striations. Representative fatigue striations during an overloading
are shown in Fig. 9a and 9b during the following steady-state region. Al-
though complete microscopic crack growth studies were not possible due
to the difficulty in locating striations, enough were found to indicate that
steady-state crack growth behavior observed macroscopically was correct.
Figures 10 to 12 summarize the effect of overload interval, AN0, on
total fatigue life. Single overload and no overload life has also been shown
for added comparison. Approximately three-fourths of the curves appear
to be somewhat parabolic which indicates an optimum or maximum
fatigue life can often be obtained for a given Ap~, Ap0, and ANs. For
most loading conditions, the optimum number of multiple periodic over-
loads, AN0, was between 10 and 30 cycles. In a few cases, however,
ANo = 1 gave maximum fatigue life. Optimum aNo greater than 10 occurred
primarily for low overload ratios. In all cases, however, single overloads
and no overloads resulted in less fatigue life than optimum periodic over-
load fatigue life. In Fig. 10, with AP.~=3.5 kips, only small relative
changes in fatigue life occurred as aNo varied from 1 to 50 or 200 cycles
for a given AP0. This is due to the relatively low overload ratios of 1.4 and
1.7. In Fig. l l with AP~=3.0 kips, greater relative changes in fatigue life
occurred for 6P0=6.0 kips due to the larger overload ratio of 2.0. In
Fig. 12, with AP,=2.5 kips, appreciable changes in relative fatigue life
occurred due to the high overload ratios of 2.0 and 2.4.
The effect of the steady-state interval, ANs, on total fatigue life for
ANo= 1 and 50 cycles is shown in Fig. 13. Similar behavior occurred for
/xN0= 10 and 30. The more complete data for a N 0 = l was obtained by
Vargas and Stephens[22]. It is evident that for a given Ap~, aP0, aN0 an
optimum value of AN~ occurred which gave maximum fatigue life. This is
particularly evident for AN0= 1 where additional data with AN~=2000
cycles was obtained. For larger overload ratios greater than 2.0 the optimum
value of AN~ for maximum fatigue life was equal to or greater than the
largest applied interval of 20 000 cycles.
Macroscopic fracture surfaces are shown in Fig. 14 for AP,=3.0 kips
FIG. 9--Fatigue striations during overloading and steady-state loading.
237
CYCLES
o ANs= 1 0 , 0 0 0 9 NO OVERLOAD LIFE
a ANs= 5,000 CYCLES 9 ONE OVERLOAD AT APo = 5
9 ONE OVERLOAD AT APo = 6
60-
(/)
LU
_1
0>-
L.) 40-
!
0
20
x
LU
h
0
0 lo 3o s'o ,oo 2Go
ANo- CYCLES
FIG. lO--The effect of overload interval ANo for AP,:3.5 kips.
and a P 0 = 6 . 0 kips. The effect of the overload interval AN0 is shown in

Fig. 14a while the effect of the steady-state interval is shown in Fig. 14b.
All specimens show the first overload at a crack length of 1.05 in. Areas
that appear flat and smooth are steady-state crack growth, while areas
that appear rough are the overload regions. The crack front and over-
load regions show curved regions which are concave up in the direc-
tion of crack growth and are attributed to the side grooves. It is quite
evident in Fig. 14a that as AN0 increases from 10 to 200 (left to right)
the percentage of total crack growth during overload increases. Like-
wise, Fig. 14b shows the percentage of total crack growth during steady-
state loading increases as ANs increases (left to right).
t, ANs=20,O00 CYCLES 1NO OVERLOAD LIFE

o ANs= 10,000 CYCLES 9 ONE OVERLOAD AT APo=6
[] ANs= 5,000 CYCLES 9 ONE OVERLOAD AT APo=5
120- f APo=6 7APo= 6

(/)
bJ
..J
0
>-
80.
I
i
0
40
x
bJ ~"~APo= 5 '--APo = 5 "-AP o = 6
I,
._I
1.0 3'0 16o 2bo

A No- CYCLES
FIG. l l - - T h e effect of overload interval ANo for APs:3.0 kips.
GARDNERAND STEPHENSON PERIODICOVERLOADS 239
9 NO OVERLOAD LIFE
: POo:
r 24C
/ "~. ,', AN s = 2 0 , 0 0 0 CYCLES
L~ . ~. o A N s = I O , O00 CYCLES
._l ~ J ~ [] AN s = 5,000 CYCLES
O
O
I
x
I..U
120!80
! ~ 8P~
h
d
/APo= 6 - 0" ~ A PO 6
40 I
0 1 10 30 50 100 200
&N0-CYCLES
FIG. 12--The effect of overload interval &No for A P , : 2 . 5 kips.
Discussion of Results
Figures 10 to 12 indicate that at higher overload ratios (about 1.7 or
greater) an optimum or maximum fatigue crack growth life using multi-
ple periodic overloads can be obtained with proper values of • Maxi-
mum life in most cases was obtained with • 10 or 30 cycles while in
some cases ANo----1 cycle gave maximum fatigue life. The actual optimum
• values were between 1 and 30 since only ANo increments of 1, 10,
~~~
ANo=1 ~ A P s = ?..5 hNo= 50
200- AP~= 2 5
~/.,,,// APo=6 APo=6
03 / r P, =2.5 APs=2.5
UJ
"J 150-
_/. / &Po;5 APo=5
o
>-. /7 / /-&Ps=~
o
t //&Po;6
b lOO- APs=3
APo=6
x
LLI - //-~ ps= APs=3
u_ 50- APo=5
._1
f ;-""~, ~,'%APs=3.s ~APs = 3.5 ~APs=3. 5
i i l
APo=5 APo=5 APo=6
2 5 10 20 1'o 2'0
ANs x IO-3-CYCLES ANs x ]O-3-CYCLES
FIG. 13--The effect of steady-state load interval ANs.
O
r.
t-
O
.I-
Z
t~
t'l
7~
-7
0
FIG. 14--Macroscopic fracture surfaces for AP~=3.0 kips, APo=6.0 kips.
30, 50, 100, and 200 were used. An optimum AN0 greater than 1 implies
that a larger degree of crack closure or intensification of compressive
residual stresses ahead of the crack or both occurs from a few periodic
overloads as compared to just single periodic overloads. Thus for high
overload ratios only, the effective stress intensity at the termination of a
short multiple overload period tends to be less than that of a single over-
load for this cyclic strain softening steel. As AN0 becomes larger, crack
growth during the overload period becomes predominant due to the con-
stant overload amplitude AP0 and fatigue life decreases. However, under
constant stress intensity range 6K, using load shedding techniques, total
fatigue life increased with increasing number of overloads • in 2024-T3
aluminum alloy[/9].
In Fig. 13 the maximum fatigue life for a given ap~, /xp0, and ANo
occurred over a wide range of AN.~. At high overload ratios (greater than
2.0) optimum values of AN, were larger than 20 000 cycles while at lower
overload ratios (less than 2.0) optimum fatigue life occurred within the
test range. Similar behavior was found for AN0= 10 and 30. Vargas and
Stephens[22] noted from Fig. 13a, that optimum intervals of AN, for
single periodic overloads compared very favorably with intervals of crack
growth retardation obtained from just single overload tests. This retarda-
tion interval was defined in Figs. la and lb. Thus, they implied that to
achieve maximum fatigue life, overloads should be applied at the end of
the constant growth rate region of Fig. la or lb, that is, when the crack
growth rate begins to increase again. Examination of crack growth versus
applied cycles data, as shown in Figs. 3 to 7, representatively, shows the
retardation period was not a constant during a multiple or single periodic
overload test. As the crack grew, the applied stress intensity for both
overload and steady-state intervals increased which changed the retarda-
tion interval. Tests performed in this research with constant values of
APo, Ap~, aN0, and AN~ would not then give true optimum or maximum
fatigue life. Only "relative" maximums were achieved. To obtain a true
optimum or maximum fatigue life, overloads should be applied after the
retardation interval which is a function of the effective stress intensity
which, in turn, is a function of aP0, ap~, aN0, AN~, and crack length.
However, for constant values of load range and cycle intervals, this
research indicates that appreciable increases in fatigue crack growth life
can occur by applying a high overload ratio (2.0 or greater), multiple
overload intervals of about 10 cycles, and steady-state cycles similar in
magnitude to retardation intervals found from single overload tests.
Crack length versus applied cycles were not converted to da/dN versus
applied AK because of the many interacting second order affects caused by
multiple periodic overloading. Even the simplest single overloading could

cause difficulty in interpreting an equation such as
d a / d N = A AK"
since n can take on negative, zero, and positive values due to retardation
and delayed retardation. Twenty to fifty percent of the single overload
life of a given specimen in this investigation had negative or zero values
for n. Up to 80 percent of the total life for single overloaded specimens in
austenitic manganese steel had negative or zero values[18]. The most logi-
cal approach then is to work toward the use of piecewise continuous effec-
tive stress intensity range as suggested from crack closure concepts which
incorporate history effects. This approach has recently received appre-
ciable interest[ l O,12,13,19,23,24,25].
Conclusions
1. Macroscopic delayed crack growth retardation was observed for
both single and multiple overload intervals, and hence, its occurrence is
not necessarily dependent upon the number of applied overloads within
an overload interval. Delayed retardation was best observed at high over-
load ratios and larger crack lengths.
2. Initial crack growth acceleration at the beginning of an overload
interval was not quantitatively evident. It appeared, however, that initial
acceleration could have occurred before the crack extended to the least
measurable scale division of 0.01 in.
3. All multiple periodic overloading tests performed increased the
fatigue crack growth life except for one specimen. The largest increase
was 453 percent.
4. For a given LxP~, • and AN~, with higher overload ratios (about
1.7 or greater) an optimum AN0 existed which gave maximum fatigue
crack growth life. In most cases this value was 10, but in a few cases
AN0=1 or greater than 10 existed. At low overload ratios for a given
APs, APo, and AN8 fatigue crack growth life was rather insensitive to
change in aNo.
5. For a given Ap, and APo, the optimum or maximum fatigue crack
growth life obtained from multiple periodic overload testing was always
greater than life obtained from a single overload.
6. For a given ap.~, AP0, and aNo, optimum or maximum fatigue crack
growth life occurred when AN, was similar in magnitude to the retardation
life obtained from a single overload test.
7. In general, this research has indicated that maximum fatigue crack
growth life can be achieved by:
(a) using a high o v e r l o a d ratio APo/APs,

(b) using a multiple o v e r l o a d interval, AN0, of a p p r o x i m a t e l y 10 cycles,
and
(c) a p p l y i n g the o v e r l o a d intervals only w h e n a c c e l e r a t e d c r a c k growth
begins following low l o a d c r a c k r e t a r d a t i o n .
References
[1] Schijve, J. in Advances in Aeronautical Sciences, Pergamon Press, 1962, pp.
387--408.
[2] Christensen, R. H. in Proceedings, Crack Propagation Symposium, Cranfield,
England, Sept. 1961, Vol. 2, pp. 326-374.
[3] Hudson, C. M. and Hardrath, H. F., "Effects of Changing Stress Amplitude on
the Rate of Fatigue-Crack Propagation in Two Aluminum Alloys," NASA,
TN D-960, Sept. 1961.
[4] Jacoby, G., Experimental Mechanics, Vol. 5, No. 1, March 1965, pp. 65-82.
[5] McMillan, J. C. and Pelloux, R. M. N. in Fatigue Crack Propagation, ASTM
STP 415, American Society for Testing and Materials, 1967, pp. 505-532.
[6] Hertzberg, R. W. in Fatigue Crack Propagation, A S T M STP 415, American
[7] McMillan, J. C. and Hertzberg, R. W. in Electron Fractography, ASTM STP
436, American Society for Testing and Materials, 1968, pp. 89-123.
[8] Hudson, C. M. and Raju, K. N., "Investigaton of Fatigue-Crack Growth Under
Simple Variable-Amplitude Loading," NASA, TN D-5702, March 1970.
[9] McMillan, J. C. and Pelloux, R. M. N., Engineering Fracture Mechanics, Vol. 2,
July 1970, pp. 81-84.
[10] Elber, W. in Damage Tolerance in Aircraft Structures, A S T M STP 486, Ameri-
[11] Wheeler, O. E., "Spectrum Loading and Crack Growth," American Society of
Mechanical Engineers, Paper No. 71-Met-X, 1971.
[12] Willenborg, J., Engle, R. M., and Wood, H. A., "A Crack Growth Retardation
Model Using an Effective Stress Concept," Air Force Flight Dynamics Labora-
tory, Technical Memorandum 71-1-FBR, Jan. 1971.
[13] vonEuw, E. F. J., Hertzberg, R. W., and Roberts, R. in Stress Analysis and
Growth of Cracks, ASTM STP 513, American Society for Testing and Materials,
1972, pp. 230-259.
[14] Jofias, D. and Wei, R. F., International Journal o/ Fracture Mechanics, Vol. 7,
No. 1, 1971, pp. 116-118.
[15] Matthews, W. T., Barratta, F. I., and Driscoll, G. W., International Journal o/
Fracture Mechanics, Vol. 7, No. 2, 1971, pp. 224-228.
[16] Adams, N. J. I., International Journal o/ Fracture Mechanics, Vol. 8, No. 1,
1972, pp. 105-106.
[17] Dotsenko, A. M., "Development of Fatigue Cracks Under Infrequent Compres-
sive Stress," Translated from Zavodskaya Laboratoriya, Vol. 37, No. 3, March
1972.
[18] Rice, R. C. and Stephens, R. I. in Progress in Flaw Growth and Fracture Tough-
ness Testing, A S T M STP 536, American Society for Testing and Materials, 1973,
pp. 95-114.
[19] Trebules, V. W., Jr., Roberts, R., and Hertzberg, R. W. in Progress in Flaw
Growth and Fracture Toughness Testing, ASTM STP 536, American Society for
[20] Raju, K. N., Ningiah, V., and Rao, B. V. S., International Journal ol Fracture
Mechanics, Vol. 8, No. 1, 1972, pp. 99-102.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authori
[21] Pitoniak, F J., "Experimental Study of Fatigue Crack Propagation and Retarda-
tion Using Polymethylmethacrylate," Technical Report AFML-TR-72-235,
Nov. 1972.
[22] Vargas, L. G. and Stephens, R. I. in Proceedings, Third International Con[erence
on Fracture, Munich, Germany, Vol. VI, April 1973, p. V-325.
[23] Probst, E. P. and Hillberry, B. M., "Fatigue Crack Delay and Arrest due to
Single Peak Tensile Overloads," AIAA Paper No. 73-325, 1973.
[24] Corbly, D. M. and Packman, P. F., Engineering Fracture Mechanics, Vol. 5,
1973, pp. 479--497.
[25] Jones, R. E., Engineering Fracture Mechanics, Vol. 5, 1973, pp. 585-604.
H. P. C h u 1
Effect of Mean Stress Intensity on

Fatigue Crack Growth in a 5456-Hl17
Aluminum Alloy*
REFERENCE: Chu, H. P., "Effect of Mean Stress Intensity on Fatigue

Crack Growth in a 5 4 5 6 . H l 1 7 A l u m i n u m Alloy," Fracture Toughness and
Slow-Stable Cracking, ASTM STP 559, American Society for Testing and
Materials, 1974, pp. 245-263.
ABSTRACT: Fatigue crack growth data on a 5456-Hl17 aluminum alloy
have been obtained by testing compact tension and contoured double canti-
lever beam specimens. The effect of tensile m e a n stress on crack growth
rate is found to reach a m a x i m u m when stress ratio R >_ 0.25. The crack
propagation laws are discussed and two new equations are developed. It is
shown that the new equations can represent the entire range of fatigue crack
growth rate versus applied stress intensity for different R values from fa-
tigue threshold to unstable crack propagation.
K E Y WORDS: fatigue (materials), crack propagation, cracking (fractur-
ing), fractures (materials), fatigue tests, stress ratio, aluminum alloys
Aluminum alloys have been increasingly used in the construction of

advanced engineering structures. To safely sustain repeated loads during
service, the alloys must have adequate fatigue properties. The purpose of
this study is to provide experimental data of fatigue crack growth in a
5456-Hl17 aluminum alloy. The specimens have been tested under dif-
ferent tensile mean stresses, and the data are analyzed in terms of fracture
mechanics. The current fatigue crack growth laws are discussed and shown
* The opinions or assertions made in this paper are those of the author and are not
to be construed as official or reflecting the views of the Department of the Navy or
the naval services at large.
1 Metallurgist, Naval Ship Research and Development Center, Annapolis, Md.
21402.
245
Copyright9 1974byby ASTM lntcrnational
Downloaded/printed Www.astIII.0F~
to be inadequate to correlate the data. Two new equations are proposed

which not only account for the stress ratio effect but completely describe
the crack-growth behavior from fatigue threshold to unstable crack
propagation.
Material
The material studied was a standard aluminum alloy, 5 4 5 6 - H l 1 7 .
This is a new temper which is included in Interim Federal Specification
QQ-A-250/20. The alloy was obtained commercially in the form of a
;/~6-in.-thick rolled plate, and tested in the as-received condition.
Specimens
Both compact tension ( C T ) and contoured double cantilever beam
(DCB) specimens were tested in this study. The former had the overall
dimensions according to ASTM Test for Plane-Strain Fracture Toughness
of Metallic Materials (E 399-72) ; the latter had a tapered section similar
to that used by Mostovoy et al[1] 2 for testing aluminum alloys (Fig. 1).
All specimens were cut from the plate with the orientation such that fatigue
crack would propagate in the rolling direction.
Tests
The specimens were tested in a Sonntag fatigue machine, which provided
sinusoidal loading by the rotation of an unbalanced mass at 1800 cpm.
Crack length was measured by means of a traveling microscope of • 10
magnification.
When testing the DCB specimens, loads were increased periodically,
and crack growth data were recorded for 0.1 to 0.2 in. under each load
setting. On the other hand, the CT specimens were tested without any
change in loading. The increment in crack measurements was about 0.03
to 0.05 in. All tests were conducted in ambient air environment.

Use oJ Data
Typical crack growth curves are shown in Fig. 2. Such curves for the
DCB specimens were always linear in consequence of the constant stress
intensity in the contoured section. Thus, the fatigue crack growth rate,
-"The italic numbers in brackets refer to the list of references appended to this
paper.
CHU ON MEAN STRESSINTENSITYIN ALUMINUM 247
NOTCH WING--,, NOTCH WING
-, !Ii- I }=__~->=~_~_~
0.002 R ~
/ A-A
t Llm r" 5.75
1.25
9 5.5 ~
CONTOURED D O U B L E CANTILEVER BEAM SPECIMEN
4 1.83
_•0"25R2 R
~-1.30-"~
9
'
4.16
3.33
d
COMPACT TENSION SPECIMEN
FIG. 1--Fatigue specimens (all dimensions in inches).
da/dN, at a load level could be determined simply by the slope of the

curve. However, in the CT specimens, the stress intensity increased and the
crack propagated ever faster during testing. In this case, the fatigue rates
were calculated by graphical differentiation.
The load parameter is expressed as stress intensity factor range,
~xK=maximum K - m i n i m u m K. The formulas for stress intensity cal-
culations for the CT and DCB specimens may be found in ASTM E
399-72 and Ref 1, respectively. The data thus obtained are listed in
Table 1.
0.3 I I I I I I
DCB SPECIMEN
MAX LOAD = 900 LB _ o "~
0.2- 050 -
0.1
j o/~
I I I I I I
:::- o 10 20 30 40 50 60
NUMBER OF CYCLES, x 1000
Z
I i i I I
1.6 8
CT SPECIMEN o
MAX LOAD : 1600 LB O
o
R = 0.50 o
t.9 1.2
/
/
o
0.~
O OO
0.4 O OO
oOO OO
o
0 o
0 I | I I I I
200 300 400 500 600 700 800
NUMBER OF CYCLES, x 1000
FIG. 2--Typical Jatigue crack growth curves.
Comparison ol Data
Because of the new temper, data on the 5 4 5 6 - H l 1 7 alloy are not
found in the literature. Bates and Clark[2] have conducted fatigue tests
on an aluminum alloy, 5456-H321. Their crack growth data for R = 0
are intermixed with those on 5 4 5 6 - H 1 1 7 (Fig. 3). Apparently, the differ-
ence in these two tempers does not have any significant effect on the
fatigue crack growth properties of this alloy.
The data on 5456-H321 show that there is an abrupt increase in crack
growth rates at A K - - 2 8 ksi~/in., which reflects a transition from slow to
fast crack growth. The present experiment was terminated before this
phenomenon was observed. Based on the closeness of the two sets of
data, it is reasonable to assume that the 5456 alloy in H l 1 7 temper has
the same transition as in H321 temper (see Fig. 3).
Conversely, there is a sudden drop in crack growth rates when
UJ
_J
u> - 100
u
a 10
IdJ
I---
t~
'~ "1-
I,,,,-
0e v
0
kJ
iii
I-- 0.1
u
6 810 20 40 6 810 20 4 6 810 20 4 6 810 20 4 6810
STRESS INTENSITY RANGE (AK) KSIv/TN.
FIG. 3--Fatigue crack growth rate versus stress intensity range of five R values.
T A B L E 1---Summary of fatigue crack growth data on 5456-11117 aluminum alloy.
R:0" R:0.125 R:0.25 R:0.50 R:0.75
da/ da/
Specimens AK b dN C AK da/dN AK da/dN &If. da/dN AK dN
Contoured 4.87 0 '~ 6.58 1.90 4.06 0.22 4.06 0.37
Double 5.20 0a 7.31 3.77 4.87 0.48 4.87 1.06
Cantilever 5.52 0~ 7.92 5.80 5.68 1.32 5.28 1.58
Beam 5.85 0a 8.76 7.31 6.50 2.48 5.68 2.11
6.17 0.12 9.75 9.75 7.31 3.70 6.09 2.70
6.49 0.27 12.20 15.89 8.12 4.57 6.49 4.29
6.49 0.87 14.61 26.00 8.93 6.51
7.47 0.42 17.07 36.00 9.74 9.44
8.12 0.73 11.37 13.89
8.12 3.67 12.99 23.64
9.09 4.82 14.61 40.33
10.39 6.67
12.34 10.06
14.94 12.86
16.24 9.57
19.48 31.20
22.73 30.00
25.98 39.33
27.93 52.50
Compact 7.46 3.63 6.16 2.37 5.97 2.66
Tension 7.61 3.75 6.22 2.58 6.07 2.83
7.76 3.86 6.29 2.80 6.19 3.00
7.97 3.98 6.38 2.94 6.31 3.25
8.03 4.15 6.49 3.24 6.45 3.45
8.28 4.32 6.61 3.53 6.58 3.70
8.40 4.64 6.76 3.74 6.79 3.97
8.53 4.80 6.92 4.12 7.01 4.35
8.68 4.94 7.10 4.56 7.24 4.75
8.84 5.18 7.22 4.74 7.38 5.22
9.02 5.38 7.35 4.92 7.71 5.75
9.19 5.68 7.60 5.14 7.89 5.95
9.79 7.04 7.73 5.40 8.09 6.15
9.91 7.36 7.90 5.86 8.28 6.80
10.21 7.72 8.08 6.26 8.54 7.23
10.50 8.00 8.28 6.72 8.84 8.10
10.85 8.44 8.51 7.14 9.19 9.00
11.25 9.20 9.00 8.00 9.50 9.58
11.69 9.72 9.20 8.30 9.97 10.50
12.22 10.92 9.33 8.40 10.52 12.10
9.51 8.60 11.20 14.60
9.73 8.90 12.24 17.70
10.00 9.70 12.75 21.00
10.21 10.10 13.45 23.30
10.50 11.40 14.27 29.00
10.80 12.20
11.11 13.40
11.35 14.20
11.67 14.80
12.20 17.10
12.89 18.70
13.46 20.05
14.13 23.80
15.02 28.20
16.13 33.20
16.85 40.10
250
CHU ON MEAN STRESS INTENSITY IN ALUMINUM 251
A K < 8 ksiVin. It indicates that there exists a stress intensity level below
which fatigue crack will not propagate. This particular • value is termed
as threshold stress intensity and designated as • Gross has discussed
the minimum limit of ~K indicative of development of nonpropagating
cracks[3]. The engineering significance of s should be equivalent to
that of the endurance limit obtained by conventional fatigue tests.
Harrison[4] has analyzed the fatigue results of 14 materials, including
an A1-Cu alloy, and concluded that "for all materials with the one excep-
tion of pure aluminum, crack will not propagate if AK/E<IO-~Vin. ''
The Young's modulus, E, is expressed in pounds per square inch. Since
aluminum alloys have a typical value of E-----10 • 106 psi, their threshold
stress intensity would be AK ~ 1 ksiVin. This is a rather low value.
As shown in Table 1, a DCB specimen was subjected to four different
loads of a K < 5 . 8 5 ksiVin. Under each of these loads, no crack growth
was observed after one million cycles. It should be pointed out that this
specimen was precracked at AK=6,49 ksiVin., and then loaded succes-
sively at AK----4.87, 5.20, 5.52, and 5.85 ksiVin. At the next load of
AK=6.17 ksiVin., the crack was found to grow at a rate of 0.12 ein./cycle.
By these results, the threshold stress intensity for R = 0 is estimated as
AKth= 6 ksiVin. A similar AKth value could be estimated from Pearson's
data on five aluminum alloys for R=0.0515]. Ideally, • should be
determined by long-term tests without any overloading which might retard
subsequent crack growth at lower loads. The duration of one million cycles
is an arbitrary choice, and the slight overloading in precracking the speci-
men is considered tolerable for practical purposes.
Comparison of Specimens
The obvious reason for using two types of specimens was to find out if
the fatigue data could be influenced by specimen design. Furthermore,
the results of DCB specimens were obtained under constant stress intensity
whereas those of CT specimens were under constant load. Testing of the
two types of specimens could also give a comparison between loading
methods.
Figure 3 illustrates that the data of CT and DCB specimens are essen-
tially the same for both R = 0 . 2 5 and R = 0 . 5 0 . This can be considered as a
simple but convincing evidence that (1) the stress intensity factor is
indeed the rate-controlling parameter in fatigue crack propagation, and
R = minimum-to-maximum stress ratio.

b AK=stress intensity factor range, ksi Vin.
da/dN~-fatigue crack growth rate, ~ in./cycle.
No crack growth after 106 cycles.
(2) the data analyzed in terms of fracture mechanics are independent of

types of specimen and loading. In this connection, it is interesting to note
that Bates and Clark[2] tested WOL specimens and their data compare
closely with those of the present work (Fig. 3). It also indicates that
good correlation of interlaboratory fatigue data can be achieved by using
fracture mechanics test methods.
Effect of Mean Stress

Traditionally, the mean stress in fatigue studies is expressed as the
ratio of minimum-to-maximum stress, R, and its influence is often termed
as stress ratio effect. The R value does not change whether the load is
reduced to stress or stress intensity factor. All of the present data on
5 4 5 6 - H 1 1 7 alloy for five R values are plotted in Fig. 4 for a direct com-
parison. Three features are readily seen, namely:
l. When mean stress was relatively low ( R = 0 . 1 2 5 ) , the fatigue rates
were slightly increased to the upper limit of the data for R = 0.
2. The fatigue rates were further increased by moderate or high mean
stress. But the data for R--0.25, 0.50, and 0.75 united and formed a
narrow band.
3. The threshold stress intensity for R > 0 was lower than that for
R---0.
When the mean stress is in tension, it usually promotes fatigue crack
growth in aluminum alloys. For example, the fatigue-rate data on 7 0 7 5 - T 6
alloy were separated into discrete bands with higher rates (relative to AK)
for higher R values[6,7]. Data on other aluminum alloys[5] also showed
an increase in fatigue rate corresponding to an increase in mean stress.
But Fig. 4 shows that under three different mean stresses (R_>0.25) the
crack growth rates for same AK was not varied. Thus, the alloy could
become insensitive to increases i n m e a n stress after it was raised to a
certain level. This peculiar behavior was probably associated with the
plastic deformation at the crack tip during testing. The implication is
that the stress ratio effect on the tested alloy could reach a maximum
for all values of R > 0.25.
Another consequence of mean stress was the reduction in threshold
stress intensity. Judging by Fig. 4, a single value of AKth might be assumed
for all the data for R > 0 . This was consistent with the results of other
aluminum alloys[5,6], as the da/dN versus AK curves for R > 0 converged
when fatigue rates were reduced to the order of one t~in./cycle. In the
case of a 7 0 7 5 - T 6 alloy[6], AKth could be taken as 2.5 ksi~/in, for five
R values from 0.2 to 0.8. For the 5 4 5 6 - H l 1 7 alloy, the threshold stress
intensity for R > 0 is estimated to be 3.6 ksi~/in, with good correlation of
data as discussed later.
100
z
:L
10
z
-r-
l--
LD
t_J
0.1
1 10 100
STRESS INTENSITY RANGE (AK}, KSI
FIG. 4--Effect of R on [atigue crack growth rate.
Fatigue Crack Growth Laws

Paris[8] has established that fatigue crack growth rate, d a / d N , can be
related to stress intensity range, AK, by
da
aN =c (• n (1)
Where c and n are constants. According to this question, a plot of d a / d N

versus AK should be linear in log-log coordinates. Figure 3 shows that
the relation holds only for a relatively short range of data. For example,
part of the data for R = 0 can be represented by
da = 0.04 (AK) 5.1~, tzin"/cycle (2)
dN
Where • is expressed in ksiVin. The equation is valid only for AK
values between 8 and 28 ksi~/in.
Forman et al[9] have discussed the stress ratio effect and pointed out
that a correct crack-growth law should satisfy the physical reality that
instability of crack propagation will occur when maximum stress intensity
approaches the critical value, Kc, for fracture. Based on these considera-
tions, they modified the Paris function as
da c(•
dN = ( 1 - R ) K ~ - • (3)
This equation is often used; it has been shown to have good representation
of fatigue data on aluminum alloys[6,7].
Theoretical curves by Eq 3 are compared with the present data. The
following constants are used to fit the data
c = 1.9,
n = 2, and
Kc= 48 ksiVin.
The correlation is fairly accurate for AK>AKth (Fig. 5 ) ) The Kr value
reported by Bates and Clark[2] is used in the calculations.
Donahue et al[10] have reviewed fatigue crack growth data for R = 0
on about 30 materials, nine of which are aluminum alloys. By considering
crack opening displacement under cyclic loading, they have derived an
equation which includes a term of AKth
da 4A [AK2 AKth2] (4)

dN ~rSyE
where
A = a material constant,
E = Young's modulus, and
S~ = yield strength.
This equation has a special feature in that it predicts zero crack growth
rate at threshold stress intensity. In order to use Eq 4, the quantity, A,
must be determined first. This step can be bypassed, however, by taking
8In Figs. 5 through 8, correlation of data for R=0.125 is graphed separately on

the right-hand side for ciarity.
100 i i
_ R= 0.125
Z
:=L
.--" 10
Z
-o
Z
-/
-I-
<IS
0.1 i t
2 4 6 8 10 20 40 6 8 10 20
STRESS INTENSITY RANGE ( A K ) , KSI
F I G . 5---Correlation oT equation by Forman et al [9].
the whole term (4A/rSuE) as a proportionality constant. Thus, the data

for R = 0 are correlated by
da
dN = 0.07 (AK 2-- AKth 2 ) , / ~ i n . / c y c l e (5)
where Agth = 6 ksi~/in.

Equation 4 does not include the factor of stress ratio, and the propor-
tionality constant must be determined individually for a particular mean
stress. The equation becomes
for R = 0 . 1 2 5
dd~= 0.08 (• 2 - AKth 2) tAm/cycle (6)
for R = 0.25
da
= 0.12 (AK 2- • 2) ~in./cycle (7)
dN
where • 3.6 ksiVin.
As pointed out before, threshold stress intensity is influenced by the mean
stress. The previous value of AKth seems to be a good estimation for it
gives an adequate correlation of data (Fig. 6).
In Figs. 5 and 6, a theoretical curve is extended by a dashed portion.
The purpose is to demonstrate that although the equation by Forman et al
predicts fast crack propagation when stress intensity approaches Kc, it fails
to describe the threshold behavior. In contrast, the equation by Donahue
et al dictates zero crack growth at AKth, but it does not portend the fatigue-
rate transition to final rupture. Apparently, there is a need of a fatigue-
crack growth law that will both express the stress ratio effect and predict
the fatigue rates for the entire range of loading from AKth to ge.
In view of this, a correct and complete crack-growth equation should be
developed which must satisfy the following criteria
lim da ----00
maxK~ K od N
and
lim da = 0
AK --> AKth dN
According to Forman et al[9], since
AK
max K = - -
1--R
then
lim da ~o9
AK~ (l--R) Kcd N
Thus, the equation should have a singularity at [ ( 1 - R ) K ~ - A K ] . But in
general, the singularity still exists when this quantity is raised to any power
[ ( 1 - - R ) K~-zxK] m. To satisfy the second criterion, the exponential form
of the equation may be simply modified as either (AK--AKth)" or
( a K " - a K t h - ) . Therefore, a generalized equation can be chosen as
da Cl (AK-- AKth) ~1 (8)
dN = [(l-R) K~-•
or
da cz ( A K " ~ - AKth%)
d N ----[ ( l - - R ) K~-AK]~2 (9)
where cl, c2, ml, m2, nl, and n2 are constants.
@
I
9 /
100
- R : 0.125
z
-{2
10 /
r
o
o
0.1 I I
2 4 6 8 10 20 40 6 8 10 20
STRESS INTENSITY RANGE ( A K ) , KSI
F I G . 6--Correlation o[ equation by Donahue et al [10].
These two equations, which are formulated by modifications of previous

work[5-10], are correlated with the data on alloy 5456-Hl17 in Figs. 7
and 8, respectively. The empirical constants, computed by trial-and-error,
are as follows:
For Eq 8
c~= 16,
n~ = 1.5, and
ml-~ 1.
_ R :0.125
U
Z
::1.
/
Z
"o
<
-'t-
<
t9
t--
..<
O. I I
2 4 6 8 10 20 40 6 8 10 20
STRESS INTENSITY RANGE ( & K ) , KSI ~ .
FIG. 7---Correlation of Eq 8.
For Eq 9
c~= 6.72,
n2 = 2, and
m s = 1.27.
As discussed before, the theoretical curves are calculated by using AKth= 6

ksiVin, for R = 0 and AKth=3.6 ksiVi~, for R > 0 . It is considered that
either of the two equations can be used to represent the data with adequate
accuracy from fatigue threshold to unstable crack propagation. The equa-
I00 i i
- R- ' 0 . 1 2 5
I I
z
:::L
.---10
z
/
!
"r
_)
o
o
L9
0.1 I I
2 4 6 8 10 20 40 6 8 ]0 20
STRESS INTENSITY RANGE (~K), KSi
FIG. 8ICorrelation of Eq 9.
tions also predict the stress ratio effect correctly. Since they are equally
simple in form and easy to use, both equations should be retained for
future applications. A choice may be made, of course, by seeing which
one can better fit the particular data than the other.
Conclusions
1. Test data for R = 0 indicate that the 5456 alloy in H l 1 7 temper
appears to have the same fatigue crack growth behavior as in H321
temper.
260 FRACTURETOUGHNESS AND SLOW.STABLE CRACKING
2. The fatigue crack growth results obtained by testing compact tension

and contoured double cantilever beam specimens are essentially the same.
3. The general effect of mean stress is to increase fatigue rates and
to lower the threshold stress intensity. The effect reaches a maximum on
the 5456-H117 alloy when R>_0.25.
4. Two crack growth equations are developed; they have been shown
to be capable of describing the stress ratio effect and the fatigue crack
growth rates over the entire range of loading from AKth to Kc. The
equations are
da c~ ( A K - - AKth)"x
dN [(l-R) K~-AK]%
and
da co- (AK"~-- AKth%)
2"-
dN [(l-R) Kc-AK]'~
where cl, co, ml, m2, nl, and n2 are constants.
Acknowledgments
The support of this work by the Naval Ship Systems Command is
gratefully acknowledged.
References
[1] Mostovoy, S. et al, Journal of Materials, Sept. 1967, p. 661.
[2] Bates, R. C. and Clark, W. G., Jr., Transactions, American Society for Metals,
June 1969, p. 380.
[3] Gross, M. R., Naval Engineers Journal, Feb. 1970, p. 44.
[4] Harrison, J. D., Metal Construction and British Welding Journal, March 1970,
p. 93.
[5] Pearson, S., Engineering Fracture Mechanics, March 1972, p. 9.
[6] Hudson, C. M. and Seardina, J. T., Engineering Fracture Mechanics, April 1969,
p. 429.
[7] Hartman, A. and Schijve, J., Engineering Fracture Mechanics, April 1970, p. 615.
[8] Paris, P. C. in Proceedings, 10th Sagamore Army Materials Research Confer-
ence, Syracuse University Press, 1964, p. 107.
[9] Forman, R. G. et al, Journal of Basic Engineering, Sept. 1967, p. 459.
[10] Donahue, R. J. et al, International Journal oJ Fracture Mechanics, June 1972,
p. 209.
DISCUSSION ON MEAN STRESS INTENSITY IN ALUMINUM 261
DISCUSSION
R. J. Bucci t (written discussion)--I should like to caution the reader

that care should be exerted when interpreting the sigmoidal shape of the
low end of the da/ dN versus AK growth rate curve in terms of a fatigue
crack growth threshold, AKt~. Upon examining Dr. Chu's data, I am
concerned that the author may not have carried his experiments to low
enough growth rates nor did he obtain sufficient number of growth rate
increments at low AK to accurately assess a "true" constant amplitude
fatigue crack growth threshold. That is, based on this reader's experience,
the Agth values cited within appear large (that is, nonconservative) and
inconsistent with experimental observations reported in the literature for
aluminum alloys where considerable growth rate data have been obtained
at rates below 10 -7 in./cycle.
From this point on, my comments must remain as conjecture for I know
of no other published data on this same alloy system and temper. However,
my reasons for this dispute are as follows.
1. Past experience with AKth determination on aluminum alloys (where
significant data have been obtained at growth rates below 10-r in./cycle)
has indicated to this reviewer that AKth for aluminum alloys is of the
order of 2 ksi V'~. and lower 5, 3, ~, 0, as the author so states in reference
to Harrison's comments (author's Ref 5).
2. The following observation on a "Threshold Knee" is relevant.
Upon examination of low growth rate data (below 10 -7 in./cycle) in the
literature (for example, refer to data of Linder, Paris, Bucci, Schmidt, et
al, A S T M STP's 513 and 536) one may approximate the shape of the
AK versus da/dN data by drawing two straight lines. The first of these
lines would be parallel to the da/dN axis and pass through AKth. The
second line would represent a best straight line fit to the data above the
threshold (usually that data in range of 10-6 and above that is often
1Senior Engineer, Engineering Design Division, Alcoa Laboratories, Alcoa Center,

Pa. 15069.
Linder, Johnson, and Paris, Journal of Engineering and Fracture Mechanics, Vol.
1, No. 1, p. 27.
Schmidt, R. A., "Extremely Slow Fatigue Crack Propagation," Ph.D. thesis, Le-
high University, 1972.
Alcoa Research Laboratories, unpublished data on 5083 alloy.
Del Research Corp., unpublished data.
represented by a power law expression of form d a / d N = C ( a K ) U ) . The

intersection of these lines would define a "knee" point in the da/dN curve
which contains • For a wide variety of alloys (high and medium
strength steels, stainless steels, titanium alloys, aluminum alloys, copper
alloys, etc. 6) where sufficient low growth rate data had been obtained no
"knees" were found to occur above 10-7 in./cycle. The author's data
appear in contradiction to this, in that "apparent knees" appear in the
10-6 in./cycle growth rate regime.
3. Several reasons may be cited as causes for elevating the apparent
&Kth threshold of the material.
(a) Faulty precrack procedures can lead to erroneous conclusions for
short cracks.
(b) Overload retardation. At low AK experience has indicated that a
very small percentage overload (order of 10 percent) can cause crack
growth delays lasting many millions of cycles.
(c) Inappropriate K calibrations--see, for example, work of Newman
(this same symposium) on effect of pin holes on compact specimen K
calibration for short cracks. (Most threshold data are obtained on short
cracks.)
(d) Environmental effects can give a sigmoidal shape to the da/dN
versus AK behavior pattern.
(e) Crack front shape changes can lead to false indications of threshold.
(f) Residual stresses or deformations or both left during precrack or
specimen fabrication or inherent in the material can inhibit crack growth.
In summary, the experience of this reviewer in examining low crack
growth data in metal leads him to regard the author's threshold determina-
tion with some suspect. Moreover, the doubt expressed in the determination
of AKth raises questions regarding the applicability of "universal" fatigue
data fitting expressions which contain an uncertain experimentally deter-
mined parameter in the fit expression.
Perhaps a cure for many of these raised uncertainties may be provided
by a better understanding of fatigue crack growth rate testing at low AK.
Hopefully, ASTM Subcommittee E24.04 can provide the leadership and
guidance necessary in the direction.
H. P. Chu (author's closure)--It is common knowledge that a precise

determination of 6Kth requires some special effort, which is not included
in this investigation. However, it is clearly indicated in the paper that the
AKth values are estimates and that "ideally, AKth should be determined by
long-term tests without any o v e r l o a d i n g . . . " This is exactly the discusser's
Del Research Corp., published and unpublished works.
DISCUSSION OF MEAN STRESS INTENSITY IN ALUMINUM 263
opinion. The two crack growth equations can represent the present data
with accuracy. They should be applicable to other fatigue crack growth
data as well. This discussion is devoted entirely to AKth determinations.
The discusser has stated that his comments must remain as conjectures.
Suffice it to say that the main topic of this paper is not concerned with
AKth determinations anyway.
M . K a t c h e r 1 a n d M . K a p l a n "~
Effects of R-Factor and Crack Closure

on Fatigue Crack Growth for Aluminum
and Titanium Alloys
R E F E R E N C E : Katcher, M. and Kaplan, M., "Effects of R-Factor and Crack

Closure on Fatigue Crack Growth for Aluminum and Titanium Alloys,"
ABSTRACT: Following the crack closure studies of Elber, experiments were
performed which appear to corroborate the link between the load required
to open a crack at its tip and the shift in crack growth rate curves with stress
ratio. Measurements of fatigue crack growth rates and closure loads were
made on compact tension specimens for positive stress ratios of R = 0 . 0 8 to
R = 0 . 8 and at rates between 10-~ and I0 -~ in./cycle. An Reut exists for 2219-
T851 and recrystallization annealed Ti-6A1-4V wherein da/dN versus AK
data does not shift to faster rates with increasing stress ratios above R = 0 . 3 2
and R = 0 . 3 5 , respectively. The physical basis for the Rout can be adequately
explained in terms of crack tip closure. Fatigue S-N data also exhibited an
Reut when replotted as AS-N.
K E Y WORDS: stress ratio, fatigue (materials), crack propagation, fracture
properties, titanium alloys, aluminum alloys, cyclic loads
Nomenclature
B Thickness, in.
C Forman constant
Kt Stress concentration factor
K1U~tX Maximum stress intensity, ksiVin.
K~ Failure stress intensity, ksiVin.
Plane strain, failure stress intensity, ksi~/in.
Closure stress intensity, ksiVin.
1 Lead engineer, Materials and Processes, Rockwell International, Los Angeles,

Calif. 90009.
Aerospace engineer, Aeronautical Systems Division, Wright-Patterson AFB, Day-
ton, Ohio. 45426.
264
Copyright9 1974 by
Downloaded/printed by ASTM lntcrnational www.astm.org
KATCHER AND KAPLAN ON ALUMINUM AND TITANIUM ALLOYS 265
&K Stress intensity range, ksi~/in.

AKeff Elber effective stress intensity range, ksi~/in.
N Cycles
P Load, lb
Ap Load range, lb
R Stress ratio, minimum stress/maximum stress
Reut Stress ratio cutoff
Sttlax Maximum stress, psi
aS Stress range, psi
U &Keff/ AK
W Width, in.
Y f (a/W)
a Crack length, in.
rn Walker exponent
n Forman exponent
Much of the accuracy in predictive analysis depends on how one
accounts for the shift in constant amplitude crack growth rates with
increasing stress ratio R. One attempt to quantitatively account for this
shift was tried by Forman, Kearney, and Engle for 2024-T3 aluminum
alloy[/].~
da C • '~
dN (l-R) Kc-zxK
It represented a modification of the Paris equation by the addition of the
denominator ( l - R ) K e - A K [ 2 ] . This addition accounted for the shift
of CGR with stress ratio as well as the asymptotic nature of the log-log plot
of d a / d N versus aK. While accurate for 2024-T3, the Forman equation
was unsuccessful in predicting the shift of C G R data with increasing
positive stress ratios for 2219-T851 aluminum and recrystallization
annealed ( R A ) T i - 6 A 1 - 4 V titanium alloys[3].
Another attempt to account for the R-factor shift was introduced by
Walker[4]. His process was a curve fitting technique. He noted that, for
certain alloy systems, the stress intensity range, • was more important as
an influence on crack growth rate than the maximum stress intensity factor,
K . . . . In other alloy systems, the reverse was true or they were of equal
importance. His representation of an effective stress or stress intensity
worked equally well for fatigue crack growth and notched fatigue coupons.
Still another attempt to account for the R-factor shift is due to Elber
[5,6]. He concluded that fatigue cracks are closed for a significant portion
paper.
of the tensile load cycle and that an effective stress intensity range, AKe~f,
exists which is less than or equal to the applied stress intensity range, •
Elber suggested that the cause of the R-factor shift is due to the fact that
the closure load must be exceeded before crack growth can occur.
It was the purpose of this program to develop an accurate equation for
fatigue crack growth in 2219-T851 aluminum and RA Ti-6A1-4V.
Elber's crack tip closure concepts were extended for use with these alloys.
Crack tip closure was also found to strongly influence notched fatigue
behavior. In addition, a clearer picture of the influence of Kmax and AK
on fatigue crack growth rates was obtained.
Material
The alloys chosen for this study were 2219-T851 aluminum and RA
Ti-6A1--4V. The chemical contents of these alloys are shown in Table 1.
The heat treatments along with their mechanical and fracture toughness
properties are shown in Table 2. Fracture toughness testing was performed
in accordance with ASTM Test for Plane-Strain Fracture Toughness of
Metallic Materials (E 399-72) using compact tension specimens. If the
fracture toughness test did not meet all ASTM criteria, the toughness value
is recorded as KQ and noted by placing it in parenthesis. All tested speci-
mens were oriented in the LT direction.
Crack Growth Rate Tests

Dimensions for the two types of compact tension specimens used to
obtain fatigue CGR data are shown in Fig. 1. The 2219-T851 specimens
were machined from Material 7 and the Ti-6A1-4V specimens were
machined from Material 67. Initially, the specimen recommended by
ASTM and having H / W = 0 . 6 0 0 was used. However, to take advantage of
a larger crack length to obtain a greater number of test points per speci-
men, testing was switched to a compact tension specimen having an
H/W=0.486[7]. The dimensions of the ASTM specimen were changed
only in the width dimension and crack lengths. The formulas for the
stress intensities for both specimens are as follows
AK = Ap y
BW ~ (2)
For H / W - - 0.600
Y - = f ( a / W ) = 29.6( a / W ) 1/2- 185.5( a / W ) % + 655.7 ( a / W ) %
- lO17.0(a/W)%+638.9(a/W)%
University of Washington (University of Washington) pursuant to License Agreement. No furth
For H / W = 0.486
Y = f( a / W ) = 30.96( a / W ) ]/2- 195.00( a/W)~-' + 730.60( a / W ) ~ -
_ l186.30(a/W)Z/,-
Precracking of all specimens was accomplished in an Amsler Vibraphore

fatigue machine. The fatigue crack growth rate specimens were precracked
at a load less than the intended cyclic test load. Sufficient precracking was
performed ( a / W > 0 . 3 5 ) to establish an equilibrium crack shape and avoid
the effects of pin loading on the validity of the ASTM equation[8].
Fatigue crack growth rate tests were performed in specially constructed
closed-loop, electrohydraulic test machines at frequencies of 1 or 6 Hz.
The cracks were followed on both sides of the specimen by sighting through
transit sighting scopes. Clear plastic scales were bonded to the specimen
allowing accurate crack length determination. The crack length measure-
ments were made at intervals of 0.050 in. and recorded, along with the
cumulative number of cycles at each reading. The fatigue crack growth
rate tests were performed in a low humidity air environment with R =0.05,
0.30, 0.50, 0.70, and 0.80. The low humidity air environment was attained
by enclosing the specimen in a plastic bag filled with freshly dried dessicant.
FIG. I--Test coupons.
I,O
O,
CO
T A B L E 1--Chemistries.
Material
Material No. Heat No. Si Cr Mn Cu Fe Zn Mg Ti Zr V C N Al H O~
2219 4 7150190 0.27 0.01 0.31 6.76 0.08 0.04 0.01 0.08 0.15 0.08 Balance C
2219 7 7150189 0.22 0.01 0.31 6.68 0.07 0.04 0.01 0.07 0.17 0.9 Balance O
-r
Ti-6AV-4V 67 K 8294 0.12 Balance 4.0 0.022 0.011 6.0 0.005 0.08 Z
z
o
TABLE 2--Heat Treatment and Mechanical Properties. o~
5
Ulti-
mate Reduc-
Ten- tion
sile Yield Elonga- of K~o/
Material Plate Size Strength Strength tion Area (KQ) ~
Material No. (in.) Heat Treatment (ksi) (ksi) (%) (%) (ksi)
z
2219 Al 4 2 by 36 by 100 T851 69 59 13 28 39 o
(Reynolds) 7 1% by 48 by 144 T851 66 50 12 25 (45)
Ti-6A1-4V 67 11/2 by 24 by 124 recrystallize annealed, 1700 ~ F,
(TIMET) 4 h FC to 1000 ~ F, A C 135 121 13 33 89
In accordance with ASTM requirements.
Closure Tests
Crack tip closure measurements were obtained for 2219-T851 aluminum

and RA Ti-6A1-4V, titanium on compact tension specimens with two
types of crack tip strain gages with approximately 0.050-in. gage length.
Each gage had two hardened steel attaching points. Both gages were
spring-loaded under sufficient pressure to prevent unwanted movement.
The gages were located such that the points bridged the crack. A side view
would show an imaginary line between the points to be approximately
0.050 in. behind the crack tip and perpendicular to the crack. Closure
load measurements were obtained from compact tension specimens having
H / W : 0 . 6 0 0 and achieving crack growth rates between 10-5 and 10-6
in./cycle.
The crack tip strain gages were plugged into a signal conditioner and
X-Y recorder where load-displacement records were obtained. Sample
load-displacement records are shown in Fig. 2. The closure loads were
obtained from curves as originally suggested by Elber. The rising and
declining portions are represented by parallel arrows pointing in opposite
directions. Additional arrows point to the beginning of crack opening
(rising curve) and crack closing (declining curve). Closure loads were
FIG. 2--Load-displacement records [or R A Ti-6A1-4V obtained with use of Elber

crack tip gage at a : 0 . 9 3 1 in.
averaged from these points. Most curves obtained during this program
displayed the hysteresis shown in Fig. 2.
Notched S-N Fatigue Tests

Tension-tension fatigue tests were performed in a 33 000-1b Amsler
Vibraphore fatigue machine at 125 Hz. Dimensions for the circumferen-
tially notched, round bar specimens used to obtain S-N fatigue data are
shown in Fig. 1. Tests were performed only on the 2219-T851 aluminum
alloy. Specimens were machined from Material 4 in the L T or longitudinal
direction with a stress concentration factor Kt = 2.0[9].
Results
Crack Growth Rate
The results of the constant amplitude fatigue crack growth rate tests for
the 2219-T851 aluminum alloy are shown in Fig. 3. It can be seen from
F I G . 3 - - C r a c k g r o w t h rates f o r 2219-T851 at positive stress ratios.
these data that the effect of R-factor between 0.08 and 0.30 is significant.
At a given • the increase in crack growth rate between these R-factors
was between 2 and 3 times. This figure also shows that as the R-factor
increases beyond 0.3 ( R = 0 . 5 , 0.7), the fatigue crack growth rate at a
constant ~ K was unaffected. The R-factor wherein no further shift in C G R
curves occurs with increasing R-factor is termed R-factor cutoff or Reut.
Due to a tack of d a / d N data at more intermediate values of R and also
some scatter in the data, one can only infer that the Reut occurs between
R = 0 . 0 8 and the vicinity of R = 0 . 3 . Crack tip closure data to be reported
later will actually show the Reut for the 2219-T851 aluminum alloy to be
about R = 0.32.
The results of the constant amplitude fatigue crack growth rate tests for
the R A T i - 6 A 1 - 4 V alloy are shown in Fig. 4. The crack growth rate
increased between 2 and 3 times as the R-factor shifted from 0.08 to 0.30,
and again increased between 1.5 and 2 times as the R-factor shifted from
0.30 to 0.50. As the R-factor increased to 0.7 and 0.8, no further increase
in crack growth rate was noted. Again due to a lack of C G R data at more
intermediate values of R and also some scatter in the data, one can only
FIG. 4--Crack growth rates for RA Ti-6A1-4V at positive stress ratios.
infer the Reut to have a value between R = 0 . 3 and the vicinity of R---0.5.
Here again, crack tip closure measurements reported later will show the
R-factor cutoff to occur at about R ---0.35.
Crack Closure
Elber was able to predict the shift in d a / d N data for 2024-T3 aluminum
through a relationship between stress ratio and closure loads. If one could
predict the R-factor shift in the d a / d N data of other alloys using the same
technique, then the nature of the shift would be explained. In this program,
measurements of the crack opening loads (closure loads) were made for
2219-T851 aluminum and R A T i - 6 A 1 - 4 V titanium at increasing positive
stress ratios. An L V D T strain gage was used to record closure loads for
2219-T851 (Table 3), and an Elber-type strain gage was used to record
closure loads for T i - 6 A 1 - 4 V , (Table 4).
For 2219-T851, tests were performed at constant maximum load while
increasing the minimum load in order to vary the stress ratio. Due to the
nature of the compact tension specimen, when the maximum load is kept
constant, the m a x i m u m stress intensity Kma~ will increase as the crack
grows. During this program, Kmax was increased from 7.4 to 16.3 ksiVin.
For the titanium alloy, tests were performed in two ways. In one case, the
stress ratio was varied under quasi-constant A K = 1 4 . 1 ksiVin, and, in
the second case, the stress ratio was varied under quasi-constant Kma~
= 17.3 ksiVin. "Quasi" herein refers to an attempt to maintain constant
stress intensity conditions by lowering the applied loads on the compact
tension specimen as the crack length increases. Results of this program
showed the closure loads to remain essentially constant with inceasing
crack lengths and increasing stress ratios. The closure stress intensity
Kel, however, increases because the crack length increases.
To graphically present the concept of crack growth rate inhibition
resulting from cracks that are closed during part of the fatigue cycle, Elber
TABLE 3--Closure loads [or 2219-T851 W--6.005 in., B=0.997 in.).
a Pet Pmax Pmln AP APe~f

(in.) (lb) (Ib) (lb) R (lb) (lb) U
2.730
2.943 580 2000 160 0.08 1840 1420 0.772
3.196 671 2000 600 0.3 1400 1329 0.949
3.495 701 2000 1000 0.5 1000 1299 1.299
3.696 647 2000 1400 0.7 600 1353 2.255
3.951 483 1108 89 0.08 1019 625 0.613
4.150 601 1987 159 0.08 1828 1386 0.758
T A B L E 4--Closure loads/or Ti-6A1-4V, R.4 ( W = 1 . 9 9 7 in., B : 0 . 7 1 9 in.).
a P~l P .... Pml~ P AP~f U

(in.) (lb) (lb) (lb) R (lb) (lb) U (avg)
Constant K .... = 1 7 . 3 k~'Cr~.
0.751 907 3535 283 0.08 3252 2628 0.808
0.813 1057 3285 263 0.08 3022 2228 0.737 0.773
0.821 1058 3245 974 0.3 2271 2187 0.963
0.864 1070 3204 961 0.3 2243 2134 0.951 0.946
0.881 1060 3004 901 0.3 2103 1944 0.924
0.891 1001 2964 1186 0.4 1778 1963 1.104
0.900 1038 2925 1170 0.4 1755 1887 1.075 1.074
0.913 1081 2885 1154 0.4 1731 1804 1.042
0.923 1113 2839 1420 0.5 1419 1726 1.216
0.931 1005 2808 1404 0.5 1404 1803 1.284 1.250
0.940 ... 2761 1381 0.5 1381 . . . . . .
0.951 1050 2731 1639 0.6 1092 1681 1.539
0.960 1012 2693 1616 0.6 1077 1681 1.561 1.527
0.971 1083 2655 1593 0.6 1062 1572 1.480
Constant A K : 1 4 . 1 ksil'CTm.
0.985 . . . . . . . . . . . . . . . . . . . . .
0.991 805 2101 168 0.08 1933 1296 0.67
1.000 844 2071 166 0.08 1905 1227 0.64
1.011 848 2071 166 0.08 1905 1223 0.64
1.021 852 2010 161 0.08 1844 1158 0.63 0.65
1.030 845 1980 158 0.08 1822 1135 0.63
1.039 765 1950 156 0.08 1794 1185 0.66
1.050 984 2522 757 0.3 1765 1538 0.87
1.061 881 2483 745 0.3 1738 1602 0.92
1.077 1033 2434 733 0.3 1710 1410 0.82 0.87
1.083 981 2403 721 0.3 1682 1422 0.85
1.091 925 2363 709 0.3 1654 1438 0.87
1.099 946 2713 1085 0.4 1628 1767 1.09
1.1215 1003 2667 1067 0.4 1600 1664 1.04
1.130 923 2575 1030 0.4 1545 1652 1.07 1.06
1.140 981 2529 1012 0.4 1517 1548 1.02
1.145 936 2982 1491 0.5 1491 2046 1.37
1.162 936 2928 1464 0.5 1464 1992 1.36
1.172 1021 2874 1437 0.5 1437 1853 1.29 1.34
1.185 947 2820 1410 0.5 1410 1873 1.33
1.213 1012 3325 1995 0.6 1330 2313 1.74
1.2215 873 3259 1955 0.6 1304 2386 1.83
1.230 1121 3193 1916 0.6 1277 2072 1.62 1.72
1.240 1015 3127 1876 0.6 1251 2112 1.69
1.250 1169 4083 2858 0.7 1225 2914 2.38
1.285 1190 3826 2678 0.7 1148 2636 2.30 2.35
1.289 1082 3742 26t9 0.7 1123 2660 2.37
1.343 929 4998 3998 0.8 1000 4069 4.07 4.07
chose to show how the applied stress intensity range • may be converted
to the effective stress intensity range • as a function of the stress
ratio. Therefore, he plotted U = • as a function of R. The data in
Tables 3 and 4 were plotted in this fashion in Figs. 5 and 6. Note the stress
ratio at the point U = 1 for both alloys. These stress ratios, R = 0 . 3 2 for
2219-T851 and R = 0 . 3 5 for RA, T i - 6 A 1 - 4 V , are within the range of
values indicated by CGR data to contain the Reut. In fact all d a / d N versus
AK data obtained at higher stress ratios coincided. This particular stress
ratio appears to be critical for both alloys, because it represents the condi-
tion wherein the applied fatigue loads maintain the crack tip fully open
with time. This means that the crack is always open during the rising as
well as during the declining portions of each successive cycle during
constant load amplitude cycling. This is true at the Reut as well as at
higher stress ratios. Figure 7 presents this schematically. Here one can see
how the relative magnitude of the closure loads and the applied minimum
loads interact.
The experiments with T i - 6 A 1 - 4 V were performed at quasi-constant
AK and quasi-constant Kmax in order to verify the influence of Kmax on U.
The stress ratios were varied between R = 0 . 0 8 and R--0.8. Crack length
was not kept constant to facilitate testing procedures and also, because
aKEFF I r ] F I l
U = ~ O = 2219-T851 ALUMINUH, KMAX
TM 7.4 TO 16.3 K S I ~ .
3.0
I
2.5
2219-T851
2.0
A KEFF
- - = 0.68 + 0,oI"
1.5
1.0 J 'k-
2024-T3 (ELBER)
0.5
~KEF F
= 0.5 + 0.4R
AK
0 I I
0 0. I 0.2 0.3 0.4 0.5 0.6 0. 0.8 0.9 1.0
STRESSRATIO (R)
FIG. 5--Stress ratio effects on crack closure ]or 2219-T851 and 2024-T3 aluminum
alloys.
4.0
E]
3.5
3.0
U:AKEFFI aK = CONSTANTKMAx : 17.3KSl\riN.
Q = CONSTANT/"K = ~.I KSI ~,~-IN.
2.5
2. ,.,
].0
0.5
.I .2 .3 .4 .5 .6 .7 .8 .9 1.0
STRESS RATIO (R)
FIG, 6--Stress ratio efjects on crack closure ]or RA Ti-6A1-4V titanium alloy.
R=.OB
R=.3 R=.5 R=.7
I PMAX PMAX
LOAD
i__ PMIN
P _
OPEN POPEN P
OPEN
PMIN
TIME
FIG. 7--Schematic diagram showing fatigue cycling with residual crack tip loads.
Elber reported U to be independent of crack length. Elber reported U to be

independent of AK as well; therefore a curve of U versus R was generated
at various AK values and used as a standard. Figure 6 presents U as a
function of R at constant AK and at constant Km~x. When Km~x was held
constant and R increased, AK was reduced. When AK was held constant
and R increased, K .... increased. While K .... was held constant for one
curve and varied for the other curve, U as a function of R remained
unchanged. Therefore, the term U is independent of K .... at stress ratios
between R = 0.08 and R = 0.8.
N o t c h e d S - N Fatigue
The R-factor cutoff was also observed on an S-N fatigue life plot of
2219-T851 data shown in Fig. 8. Material for the fatigue tests came from
a heat of metal different from the heat used for C G R tests. Comparison of
results, however, is still possible because the mechanical properties and
chemistries were similar. To observe the R~ut, S-N data represented by
open symbols were replotted as AS-N data, represented by closed symbols.
Note the coincidence of curves for R-factors of R = 0 . 5 and R = 0 . 7 5 at the
fatigue limit stress range. Considering the scatter typical of S-N data, the
deviations which exist herein may be considered slight by comparison.
Note the separation of these AS-N curves from the curve plotted for
lO
%
^~
60
s
OR 50 R:O. 75
AS
(X1000, PS 1)
4O
STRESS POINTS
R -- O
30 "~ R=0.5
O 0.05
00.5
o 0.75
R=O.O5
DELTA STRESS POINTS
20 R
9 .05 It $
0.5
9 .75
IO
104 105 106 107 108

CYCLESTO FAI LURE
FIG. 8--Notched fatigue S-N and AS-N data [or 2219-T51 at K~=2.0.
R = 0.05 data. In the latter case, the fatigue limit stress range is about 50
percent higher. It is most interesting that crack growth rate data shown
in Fig. 4 for 2219-T851 exhibited a similar coincidence for several differ-
ent R-factor curves at R = 0 . 3 , R=0.5, and R = 0 . 7 . Again, the lower
R-factor curve, in this case at R = 0 . 0 8 , was displaced to significantly
slower rates of crack growth. Having both slower crack growth rates and
a higher fatigue limit stress range would together imply a higher fatigue
resistance at low stress ratios. The foregoing results show the similarity of
behavior during both notched fatigue and CGR tests for 2219-T851. The
compatibility of these fatigue data with closure measurements lends much
weight to the idea that an effective stress concept similar to that found in
crack growth may be a major influence on notched bar fatigue.
Discussion
Closure and Crack Growth Rates
The interpretation of fatigue crack growth phenomena is a complex
undertaking. Many interactive parameters are involved. Elber showed
fatigue crack growth rates to depend on an effective stress intensity range;
Paris, before him, showed them to depend on an applied stress intensity
range, and Walker showed them to depend on stress range and maximum
stress (or stress intensity range and maximum stress intensity). The results
of this program showed all of the foregoing investigators to be correct if
their conclusions are restricted to certain magnitudes of crack growth rate.
Elber was the first investigator to relate residual crack tip stresses in
terms of closure loads and to provide a direct measurement of such loads.
In addition, he showed these residual loads to act in a regular manner to
inhibit crack growth. When properly accounted for, fatigue crack growth
rates at different R-factors could be resolved into a single curve of da/dN
versus aKew Cheng and Brunner[10] at Boeing and Pitoniak[ll] at the
Air Force Materials Laboratory, using plastic specimens, have verified that
crack tip closure is a real phenomena. They loaded fatigue cracked com-
pact tension specimens in tension and measured the interferometric stress
patterns when the crack closed upon unloading. Cheng and Brunner
corroborated the existence of a compression kernel across the crack front
through observation of a high stress region in two places--ahead of the
crack tip in base material and behind the crack tip compressing the
fracture surfaces together.
Pitoniak noted that, upon unloading, the crack was fully closed before
zero load was obtained and that the outside edges of the specimen dosed
first. The fact that the outside edges closed first may have been due to a
nonuniform thumbnail-type crack shape and the increased plasticity of
2.78 FRACTURETOUGHNESS AND SLOW-STABLE CRACKING
plane stress conditions near the surface. This effect may vary the closure
loads obtained from load-displacement records as a result of the inside
crack front closing or opening at a slightly different load than the outside
crack front. All crack fronts in this program displayed a slight arc be-
tween surfaces with the center length being longer than at the surface.
Some of the scatter in the closure load measurements may have been
caused by this phenomenon.
Evidence of the relationship between crack growth rates and closure is
presented herein for 2219-T851 aluminum and RA T i - 6 A 1 - 4 V titanium.
Figures 5 and 6 display closure measurements in the form of U = AKeff/AK
as a function of R for both alloys. The curves shown exceed U--1 at a
stress ratio R = 0 . 3 2 for 2219-T851 and R = 0 . 3 5 for T i - 6 A 1 - 4 V . At
U = I , the applied minimum load equals the residual closure load. At
U > 1, the applied minimum load exceeds the closure load and the crack
tip is open during each entire cycle. This explanation follows from the
closure results found by Elber and shown schematically in Fig. 7 for the
alloys investigated in this program. When U is less than unity, the data in
Figs. 5 and 6 may be adequately replaced by a linear function of R. The
following equations were developed. For 2219-T85 l
U----0.68+0.91R, where 0 . 0 8 < R < 0 . 3 2 (3)

For RA, T i - 6 A 1 - 4 V ,
U = 0 . 7 3 +0.82R, where 0 . 0 8 < R < 0 . 3 5 (4)
No statement can be made concerning the validity of Eqs 3 and 4 at nega-

tive stress ratios because closure measurements made during this in-
vestigation were from fatigue cracks grown at positive stress ratios. Note
that these equations differ from the one presented by Elber for 2024-T3,
as follows
U=0.5 +0.4R where -0.1 < R < 0 . 7 (5)
In addition, the plot of U versus R for 2219-T851 and R A T i - 6 A 1 - 4 V

displays a slope which increases more rapidly in parabolic fashion above
U = 1. Another difference between Elber's data and the closure data re-
ported herein is that U values for 2024-T3 never exceeded U = 1 at stress
ratios up to R = 0 . 7 and, when extrapolated linearly up to R----1.0, the
value of U never does exceed U = 1.
Constant amplitude crack growth rate data were obtained from the
same heats of metals from which closure measurements were made and
showed certain interesting characteristics in common at rates between
10 -7 and 10-5 in./cycle (see Figs. 3 and 4). The stress ratio for which
closure measurements displayed U = 1 is also the stress ratio wherein

da/dN data as a function of • coincide. At stress ratios greater than
about R = 0 . 3 for 2219-T851 and R = 0 . 5 for R A T i - 6 A 1 - 4 V , the con-
stant amplitude, fatigue crack growth rate curves fail to shift as the stress
ratio is increased. This suggests that above these stress ratios crack growth
rates are dependent on the stress intensity factor range and independent of
the stress ratio. This would also suggest the inadequacy of many hereto-
fore accepted equations and models of crack growth behavior which pre-
dict a continuous shift in crack growth rates with increasing stress ratio.
The stress ratio above which da/dN data no longer shifts has been termed
Rout. The da/dN versus ~K curve that represents the Rout signifies the
curve describing the crack growth rates induced by a cyclic effective
stress intensity range. This was postulated by both Elber and Walker,
but never observed from the data of actual test specimens.
A principal objective of this program was to develop an equation that
would accurately replace C G R data for use as source data in predictive
computer programs. Because variable amplitude spectrums repeatedly
jump from one stress ratio to another, such a source equation of necessity
has to accurately account for the dependency of C G R on stress ratio
changes. An attempt was therefore made in this program to develop
equations that would enable calculation of da/dN from ~xK and R data.
By using the Elber form where AKe,f=U aK, the following equations
were found successful in replacing R = 0.08 data and R = 0.3 data.
For 2219-T851
da/dN= 7.377 X 10 -2~ [U • 4~ (6)
For R A T i - 6 A 1 - 4 V
da/dN= 5.353 X 10 -33 [U AK] 649 (7)
when ~K has units of psi "v/in. and da/dN has units of in./cycle.
The constants in these equations differ from those used by Elber for
2024-T3. Figures 9 and 10 show these equations plotted as solid lines
overlaying actual test points. Note the inability of Eqs 6 and 7 to ac-
curately describe C G R data at rates above about 10 -~ in./cycle. Such
results imply that the mode of cracking during fatigue loading changes
with the magnitude of the applied stress intensity range.
Notched S-N Fatigue Tests

Closure observation of the notched fatigue • curve reveals addi-
tional facts about the influence of AS and Sm,.~ in fatigue life. Since crack
growth rate data and notched fatigue data display the R,.ut phenomena,
FIG. 9--EIber equation ]or fitting crack growth rate data points for 2219-T851.
one can postulate that a significant portion of the life of 2219-T851

during notched fatigue testing is spent in crack growth rather than crack
initiation and at stress intensities where closure loads are effective. This
appears to be true at stress range levels at or near the fatigue limit. How-
ever, insufficient data are presented in Fig. 8 to make the same postulate
at higher stress levels. Unpublished data, however, have shown the pre-
dominance of crack growth during the life of notched fatigue specimens
of 2219-T851 at stress levels above the fatigue limit[12].
Conclusions
1. Fatigue cracks may be closed during a part of the tensile load cycle
at low positive stress ratios and open for the entire cycle at high positive
stress ratios.
2. An Reut exists for 2219-T851 and RA T i - 6 A 1 - 4 V , wherein da/dN
versus AK data does not shift to faster rates with increasing stress ratios
above R = 0 . 3 2 and 0.35, respectively. The physical basis for the Reut can
be adequately explained in terms of crack tip closure.
3. The crack tip closure model as postulated by Elber is valid for
FIG. lO--Elber equation for fitting crack growth rate data points for RA Ti-6A1-4V.
2219-T851 aluminum alloy and R A T i - 6 A 1 - 4 V titanium alloy below a

C G R of approximately 10 -~ in./cycle and at positive stress ratios below
the Reut.
4. Notched fatigue life data when plotted as the stress range versus life
cycles exhibit the Rout phenomenon.
Acknowledgment
The authors wish to thank A. W. Sommer and M. J. Harrigan for their
support and stimulating discussion on the subject of fracture and fatigue.
The important contribution of R. Fugazzi in the area of testing and in-
strumentation is also most appreciated.
References
[1] Forman, R. G., Kearney, V. E., and Engle, R. M., Journal of Basic Engineering,
Sept. 1967, pp. 459--464.
[2] Paris, P. C. in Proceedings, 10th Sagamore Conference, Syracuse University
Press, 1964, p. 125.
[3] Katcher, M., "Crack Growth Retardation Under Aircraft Spectrum Loads," Los
Angeles Aircraft Division/Rockwell International, NA-72-374, to be published
in Engineering Fracture Mechanics.
[4] Walker, E. K. in Effects of Environment and Complex Load History on Fatigue

Life, A S T M STP 462, American Society for Testing and Materials, 1970, pp. 1-
14.
[5] Elber, W., Engineering Fracture Mechanics, Vol. 2, 1970, pp. 37-45.
[6] Elber, W. in Damage Tolerance In Aircraft Structures, A S T M STP 486, Ameri-
[7] Communication with W. F. Clark, ASTM E-24 Committee on Fracture Testing.
[8] Newman, J. C., Jr., in this publication.
[9] Petersen, R. E., Stress Concentration Design Factors, Wiley, New York, 2nd ed.,
1959.
[10] Cheng, Brunner, International Iournal o[ Fracture Mechanics, Vol. 6, 1970,
pp. 931-434.
[11] Pitoniak, F. J., Experimental Study of Fatigue Crack Propagation and Retarda-
tion Using Polymethyl Methacrylate, AFML-TR-72-235, Nov. 1972.
[12] Martin, G., unpublished data, Rockwell International, LAAD.
A . F. G r a n d t , Jr. 1 a n d J. P. G al l aghe r 2
Proposed Fracture Mechanics Criteria

to Select Mechanical Fasteners
for Long Service Lives
REFERENCE: Grandt, A. F., Jr. and Gallagher, J. P., "Proposed Fracture

Mechanics Criteria to Select Mechanical Fasteners for Long Service Lives,"
ABSTRACT: It is proposed that long service lives may be attained by select-
ing a fastener system which keeps the stress intensity range during service
below the threshold stress intensity range. Using this criterion one may spec-
ify the maximum flaw size (from prior damage or initiated by fretting)
which will be permanently retained by the residual stress field surrounding
the fastener.
An analytical procedure which can be employed to obtain stress intensity
factors for practical fastener crack configurations is demonstrated with re-
suits for open holes, cold-worked holes, interference fit fasteners, and pin-
loaded holes. In addition, threshold stress intensity factor data are taken
from the literature and presented in a form convenient for use by the de-
sign engineer.
KEY WORDS" fatigue (materials), residual stress, mechanical properties,
mechanical fasteners, fracture properties
Since the o p e r a t i o n a l lives of aircraft c a n be limited by cracks which

initiate at fastener (bolt or rivet) holes, recent A i r F o r c e e x p e r i m e n t s [ / ] ~
have b e e n c o n d u c t e d to d e t e r m i n e procedures for reducing the severity
of these flaws. C o l d - w o r k i n g precracked holes with a n oversized m a n d r e l ,
for example, was shown to arrest s u b s e q u e n t crack growth i n some cases
1Air Force Materials Laboratory, Wright-Patterson AFB, Ohio 45433.

Air Force Flight Dynamics Laboratory, Wright-Patterson AFB, Ohio 45433.
paper.
283
Copyright9 1974 by
for over 10 6 additional cycles of loading[2]. It would appear that residual

stresses generated by the cold-working process were directly responsible
for arresting the initial cracks. These results suggest that it may be possible
to permanently arrest fatigue cracks growing from fastener holes under
certain conditions. The purpose of this paper is to detail the conditions
which may arrest hole cracks and to formulate a preliminary design
procedure which could be utilized to achieve long service lives for
mechanical joints.
Approach
The current fracture mechanics approach to safe life design requires
the knowledge of:
1. crack growth behavior of the structural material under service con-
ditions,
2. initial flaw size and shape,
3. the stress intensity factor (KI) as a function of crack position and
shape, and
4. the allowable or terminal crack length.
With this information it is possible to estimate the service life of flawed
components and, thus, set reasonable inspection intervals.
It is proposed here that when the allowable crack size around fastener
holes is set by the value of the fatigue threshold stress intensity factor[3],
mechanical joints can possibly be designed for an "infinite" fatigue life.
The goal of long service life may be achieved by selecting a fastener sys-
tem which keeps the alternating component of stress intensity factor (AK)
during service loading below the threshold stress intensity (AKth). Thus,
crack growth is prevented.
In order to predict ~,K for fastener cracks, one must know the stress
intensity factor calibration for the fastener geometry of interest. Thru-
cracks extending from open holes have been examined by Bowie[4], while
Cartwright and Ratcliffe[5] determined the strain energy release rate for
two equal radial cracks emanating from a pin-loaded hole. In addition,
Crews and White[6] recently developed stress intensity factor calibrations
for cracks growing from open holes into high prior load induced residual
stress fields.
To date, analytical consideration of holes which have been subjected to
some type of fatigue life extension process, namely holes either containing
an interference fit fastener or subjected to mandrel enlargement, has been
limited to determining the residual stress field surrounding the hole[7-12].
Corresponding stress intensity factor solutions are presently unavailable
for "fatigue improvement" fasteners.
GRANDT AND GALLAGHER ON MECHANICAL FASTENERS 285
Stress Intensity Factor Analysis

A rational procedure to determine stress intensity factor solutions for
cracks propagating from various types of fastener holes is described below.
This procedure utilizes linear superposition, a technique which has seen
frequent use in the past (see for examples Refs. 13-19) to obtain the
stress intensity factor directly from the hoop stress surrounding an unflawed
hole. The unflawed stress distribution may result from remote loading or
be introduced either by cold working or by interference fit fasteners.
In order to apply the linear superposition method (see Ref 13 for de-
tails) one needs stress intensity factor solutions for flaws subjected to
arbitrary crack face pressure. Such solutions are available in the literature
for central flaws in wide plates[13] and for edge-cracked strips[14-17].
In addition, the central crack solution has been modified to estimate K~
for two diametrically opposed radial cracks extending from a circular
hole[6, 18].
In this paper, a stress intensity factor solution was developed directly
for the single hole crack loaded with arbitrary crack face pressure as
shown in Fig. I. This solution is based on work by Rice[20] who has
shown that once the displacement field and stress intensity factor are
known for one geometry and loading (Case 1), K~ may be obtained for
any other symmetric loading (Case 2) on the same geometry. If one
chooses a single radial crack emanating from a hole in a large plate as the
flaw geometry, and specifies the Case 1 loading as uniform tension per-
FIG. 1--Open hole containing a radial crack subjected to pressure p(x).
pendicular to the crack, the desired Case 2 stress intensity factor solution
for the loading shown in Fig. 1 becomes
K~= ( l _ vE2 ) K j ~L " p ( x ) ~~ a (1)
Here E is the elastic modulus, v is Poisson's ratio, KB is the known Case

1 Bowie stress intensity factor solution[4] for a single crack of length a,
and ~ is the y-component of the Bowie crack surface displacement. Crack
surface displacements were determined for several crack lengths using
finite element analyses of the Bowie problem and then were fit with a
conic section equation described by Orange[21]. The conic section pro-
~v.
vided a convenient method for obtaining the partial derivative ~-d in Eq 1
(see Ref 22 for a detailed derivation of Eq 1 ).
In the present work it is assumed that p ( x ) is given by
Tix ~for p < x _ < a

ki=O
Here x is measured from the edge of the hole, while p and a specify the
g A
i=O
w, x
-el"
FIG. 2--Schematic of hoop stress variation from edge of fastener hole (unflawed).
range over which Eq 2 is valid (see Fig. 2). It is clear that any arbitrary
pressure distribution may be expressed to suitable accuracy with poly-
nomials of degree m and n by choosing the coefficients S, and T, by
standard curve fitting techniques. When the pressure p(x) is chosen as
the hoop stress surrounding an unflawed fastener hole, the superposition
method enables one to calculate stress intensity factors directly from Eq 1.
This procedure is demonstrated with several examples of specific fastener
problems in the next section.
Application of the Linear Superposition Method to the Fastener Hole

Problem
Both Emery[14-16] and Bueckner[17] have shown that stress intensity
factors calculated by linear superposition with their edge crack formula-
tions agree well with other published results for edge cracked strips in
tension and bending. Emery also demonstrated the technique for radial
cracks emanating from holes in plates subjected to remote tension.
The encouragement provided by Emery's success in determining stress
intensity factors for open holes and the simplicity and versatility of the
linear superposition approach lead the present authors to propose this
method for determining stress intensity factors for cracks emanating from
cold-worked holes or from holes containing interference fit fasteners. Four
examples will be discussed to demonstrate the feasibility of applying Eq
1 to calculate K~ for fastener hole crack problems.
Open Holes
For the case of an unworked open hole of radius r located in a large
plate which is remotely loaded in one direction with a stress of ~o~, the
hoop stress (~n) along the radial line perpendicular to the loading direc-
tion[23] is
r (3)
where x is measured from the edge of the hole.

Taking this expression for p(x), Eq 1 gives the stress intensity factor
calibration shown in Fig. 3. Note that the hole crack solution agrees within
7 percent of the Bowie analysis[4] throughout the crack length range in-
vestigated. An earlier application of the Emery-Bueckner edge-crack solu-
tions to the Bowie problem[19] gave results which agreed within 10 per-
cent for small cracks (a/r< O.15) but which exceeded the Bowie analysis
by as much as 23 percent for the larger flaws as shown in Fig. 3.
4- ~o
K I =o,=,"t/'-ffa F(a/r)
/ "~ --BOWIE ANALYSIS
9 I ~ \ - - - HOLE-CRACK SOLUTION
3 _,0o, sow ,o,,
A
" *-~.
~ % . %. 0"_
u
Z
0
u,~l'
z
i 9 i i i i
0.5 1.0 1.5 2.0 2.5 3.0

DIMENSIONLESS CRACK LENGTH air
FIG. 3--Comparison of stress intensity [actor solutions for the Bowie problem.
Mandrel Hole Enlargement

Estimates of the residual stress field produced by two values of mandrel
hole enlargement in aluminum plates are presented in Ref 9. This analysis
shows that the cold-working operation causes hoop compression next to
the edge of the hole, and residual tension in the far field. The dramatic
reduction in stress intensity factor due to the residual compression is
readily determined by the linear superposition method as shown in Fig. 4.
Here the mandrel enlargement stresses were fit with polynomials and
added to the hoop stress caused by a 40 ksi remote load (see Eq 3).
This remote stress was chosen to ensure that the resultant stresses re-
mained tensile over the crack length, but did not exceed the elastic limit.
In actual applications, of course, the cold-worked hole subjected to remote
tensile forces or load transfer through a fastener could be studied with
more sophisticated stress analysis procedures (for example, finite ele-
ments). Nevertheless, the linear addition approximation demonstrates
the potential fatigue improvement resulting from cold-working fastener
holes. The 0.008 in. mandrel enlargement curve in Fig. 4, for example,
shows that K~ for small cracks ( a / r about 0.25) is reduced to the thresh-
old AK for fatigue, preventing crack propagation.
50
40
f s" "" "~" " " " 7075-T6 A,

30 /r ./" ,"_ ~r * ~ p 5 IN.
I/ Ks,
."|| KS, --Bow,E (NO
--BOW,E (No ccoL~
o l t , WORK1
./ /~ ~.,L.... .-.o,oo6 in. interference

20 / s
/ 9 --0.008 in. interference
9 i I
/ 9
10
1" ts
i
i i i i i i
0.5 1.0 1.5 2.0 2.5 3.0

DIMENSIONLESS CRACKLENGTHalr
FIG. 4--Stress intensity factor solutions for cold-worked holes.
Interference Fit Fasteners

If the hoop stress is known around an unflawed interference fit fastener,
stress intensity factors may again be predicted from the present solution.
A stress analysis based on the distortion energy theory of yielding is
given in Ref 8 for a titanium interference fit fastener in a titanium plate.
Representing these results by a polynomial expansion and calculating K1
as before, stress intensity calibrations are shown in Fig. 5 for the installed
fastener before remote loading, with a superimposed remote tensile load,
and after the remote stress has been removed. The difference in stress
intensity between the loaded and unloaded cases provides the ~K which
is related to the fatigue crack growth rate. Results of this type may be
used to estimate the cyclic life of cracked interference fit fastener holes.
Pin-Loaded Holes
The circumferential stress field around a 0.25-in.-diameter hole loaded
through a rigid pin in a finite aluminum sheet is given in Ref 24. In this
instance the maximum principal stresses were found to occur at an 81-deg
angle with the loading axis as shown in Fig. 6. Expressing the analysis for
two values of pin clearance (based on diameters) by polynomial expan-
60
50
40
30
20
SHEET (0.002 IN.
n ..., .... .~,~-. . . . -
10
0.2 0.4 06 0.8 1.0 1.2 1.4

DIMENSIONLESS CRACK LENGTH air
FIG. 5--Stress intensity factor solutions [or interference fit fastener insulation.
sions and using the general solution of Eq 1, one obtains the stress
intensity factor predictions of Fig. 6. The fact that the radial crack occurs
at an angle causes no additional complication, since the original stresses
were found for this angle.
Note in Fig. 6 that KI increases with pin clearance and initially exceeds
the Bowie analysis[4] for an open hole containing a radial crack perpen-
dicular to the same remote tension, but that KI decreases as the crack ex-
tends. The strain energy release rate found by compliance techniques for a
pair of radial cracks emanating from pin-loaded holes[5] confirms the
initial magnification in stress intensity factor for pin-loaded holes in wide
plates. The fact that KI eventually falls below the remote loaded open
hole (Bowie) analysis for longer cracks is also expected from the well-
known result that crack line loaded flaws show a decreasing stress inten-
sity factor[25].
Achieving a Fatigue Stress Intensity Threshold

As just noted, the stress intensity factor for a crack propagating from a
cold-worked hole is substantially lower than that from an unworked hole.
Consequently, due to the basic relationship between the range in alternat-
ing stress intensity factor (• and cyclic crack growth rate ( d a / d n ) ,
fatigue crack growth rates from cold-worked holes will be significantly
70-
60
50
ii ......... -. "'-....
ii//p.s KIP ......... _._.:--:::::::
N 311
' /--~ 2r I--~ 2r~0.25 IN.
I I/I~xl I 9~ -DIRECTION --BOWIE (NO PiN)
/ ~ - L TOMAXIMUM -- 2.O~PINCLEARANCE
~k. p ~ 7 TENSILESIRESS .--0.4~ PIN CLEARANCE
10
i ! i i i
0.5 1.0 1.5 2,0 7.5 3.0

DIMENSIONLESSCRACKLENGTHalr
FIG. 6--Stress intensity [actor solutions for pin-loaded holes.
slower for the same remote loading. The cyclic crack growth rate behavior
as a function of AK is shown schematically in Fig. 7. Figure 7 also shows
that if • drops below a given level (AKth) cracking rates are immeasur-
able. This critical level of stress intensity for no observed crack growth has
been referred to as the fatigue crack growth threshold.
While several investigators[26-32[ have recently explored the influence
of changing material, environment, and frequency on the fatigue crack
growth threshold (AK,,), the concept of a nonpropagating fatigue crack
has been discussed in the literature for at least 20 years[33[. Until 1969,
however, the parameters used to describe nonpropagating cracks were not
of the fracture mechanics variety and, therefore, not directly useful in
present design practice. Harrison[34[ reanalyzed data taken from several
Frost papers[35--46[ into a fracture mechanics format and found that, for
a stress intensity ratio R (R=Kmi,/K,,,,,.O of zero, the stress intensity
threshold was normally greater than the AKth value given by
AKthfR=0= 1.5 X 10-' • E\/in. (4)

where E is the elastic modulus. For all materials with the exception of
pure aluminum, Harrison found that cracks will not propagate if
• < 10--~~ .
University of Washington (University of Washington) pursuant to License Agreement. No furt
K~ ~K K!ax
I0-2
Kn,dn
10-4
o
.S
10-6
I0-8 ~ : CAKn
CRACKS DO NOTGROWBYFATIGUE
I0-I0
i I I
2 5 ,o ~o ;o ,oo
A K,,lk'~ in
FIG. 7---Schematic o] ]atigue crack growth behavior of aluminum alloy ]or zero-
tension loading.
Aluminum alloy threshold data as a function of the stress intensity

ratio was taken from several sources[31,47-49] and plotted in Fig. 8.
The data shown can be adequately described for 0 < R _< 0.8 with
AKth -= AKthln=0( 1 -- R ~') (5)

It also appears that Eq 5 provides a useful approximation to some avail-
able steel data[27,30,49] shown in Fig. 9. While other functions may be
chosen to correlate AKth and stress ratio for specific materials, Eq 5 pro-
vides a convenient relationship which gives a much more reasonable ap-
proximation to available data than an overly conservative linear equation.
In the event that threshold data do not exist for a material of interest,
approximations might be made utilizing Eqs 4 and 5. When limited data
are available for 0.1 < R < 0.4, Eq 5 can be used to represent the data
for estimates of threshold values outside this range.
For a crack being propagated from a hole through a residual com-
pressive stress zone by a remotely applied cyclic stress, the stress intensity
(I-R 2)
_•2.2
FORCE FIT THRU
.t
o 7075-T6511,(47) ( 1.5 .-R'~J ~ ~
9 2~4-1"3, (32) ~
nx BS L654.5%Cu-AIAIIoy,(491
Pure AI, (49)
I I I I
o .2 .4 .6 .8 I.O
DIMENSIONLESS STRESS RATIO R
FIG. 8--Stress intensity factor threshold data for aluminum alloys in laboratory
air.
7 9
6 ~ R 2 ~ /--FORCEOFITT.ROUG.
x.
9 INCONEL iUTS- 95 ksi)(49) ~_~"
-r 2 m L0W ALLOY STEEL (UTS= g6 ksi) (49) ~
V AS)) GRADE B CLASS I STEEL (30) ~
< 9 ~ cLAss 2 s ~ (~} \\
0 i i i i T
.2 .4 .6 .8 1.0
DIMENSIONLESS SIRESS RATIO R
FIG. 9--Stress intensity Jactor threshold data for steel alloys in laboratory air.
minimum will normally be zero except for high remote stress ratios. To
illustrate the usefulness of the stress intensity threshold concept, consider
the case where the remote stress ratio is sufficiently low, such that Kmin
is zero. When Kmi. is zero, Kma,---• and if Km~.~-(AKthlR=0,cyclic
crack growth should not occur. The data given in Fig. 8 for example, in-
dicates that AKthIR=0 is approximately 2 ksi'x/in, for 7075-T6 aluminum.
This information implies that the tolerable infinite life crack length is ap-
proximately 0.040-in. for 0.008-in. mandrel interference on a 5/16-in.-
diameter hole under 40 ksi remote loading (see Fig. 4). For this specific
case, then, cracks smaller than 0.040 in. should be halted permanently by
the cold-work process.
Implications to Fastener Design

To achieve infinite life for a fastener joint during the design stage re-
quires the use of traditional fracture control tools, that is, NDI, a stress
intensity factor analysis, and a critical level of stress intensity. As can
be noted from Fig. 10, the object of an infinite life design procedure is to
A CRACK IN SHADED R E G I O N WILL NOT PROPAGATE

SINCE A M <: AKfh . CRACKS LONGER THAN o +
WILL EXTEND AND MUST BE FOUND DURING INSPEGTION.
f T
FIG. lO--Concept of crack retention by residual stress zone surrounding fastener

holes.
specify the distance a*, the maximum crack length that will be contained
by the residual stress zone for an "infinite" number of applied load cycles.
The amount and type of cold work as well as hole size and applied load
levels combine to specify the distance a* through the use of the supposi-
tion: • The approach allows for directly setting the infinite life
defect length that can be missed during an inspection prior to a cold-work
process. In the event that the residual stress field should change during
the life of the structure due to unexpected overloads or cyclic relaxation,
stress intensity factors could be calculated from the new stresses as before
to determine the severity of existing flaws. Also, it would appear feasible
that the approach might be equally valuable for specifying a sufficiently
large a* which would arrest fretting cracks nucleated in the hole during
the service life.
Summary
F r o m the results generated to date, it is clear that linear elastic fracture
mechanics tools provide the designer with the ability to achieve long
fatigue lives for fastener holes. The proposed approach states that cracks
around fastener holes will not extend during long service lives if the stress
field generated by the fastener keeps the range associated with service
loading below the threshold ~xK for fatigue.
Linear superposition was employed to determine stress intensity factor
calibrations for several practical fastener applications. This procedure
allows one to calculate KI directly from the unflawed stress distribution,
avoiding the more difficult problem of resolving crack tip stresses. Thresh-
old stress intensity range data obtained from the literature were pre-
sented in a form convenient for estimating this parameter for other mate-
rials and load ratios.
References
[1] Tiffany, C. F., Stewart, R. P., and Moore, T. K., "Fatigue and Stress-Corrosion
Test of Selected Fasteners/Hole Processes," Technical Report ASD-TR-72-111,
Wright-Patterson Air Force Base, Ohio, Jan. 1973.
[2] See Appendix I by G. J. Petrak and Appendix II by J. P. Gallagher and H. D.
Stalnaker of Ref 1.
[3] Paris, P. C., Weiss, W., and Wessel, E. T., "On the Threshold for Fatigue Crack
Growth," presented at the Fifth National Symposium on Fracture Mechanics,
1 Sept. 1971, Urbana, Ill.
[4] Bowie, O. L-, Journal of Mathematics and Physics, Vol. 35, 1956, pp. 60-71.
[5] Cartwright, D. J. and Ratcliffe, G. A., International Journal of Fracture Me-
chanics, Vol. 8, No. 2, June 1972, pp. 175-181.
[6] Crews, J. H., Jr. and White, N. H., "Fatigue Crack Growth From a Circular
Hole With and Without High Prior Loading," Technical Report NASA TN
D-6899, Langley Research Center, Hampton, Va., Sept. 1972.
[7] Regalbuto, J. A. and Wheeler, O. E., Experimental Mechanics, Vol. 10, No. 7,
July 1970, pp. 274-280.
[8] Siegel, A. E., "Stress Patterns in a Sheet Containing an Interference Fit Fas-
tener," Interoffice Memo No. 231-4-359, McDonnell Aircraft Co., 29 Dec. 1966.
[9] Siegel, A. E., "Residual Stresses Produced by Mandrel Hole Enlargement," Inter-
office Memo No. 237-33, McDonnell Aircraft Co., 20 May 1969.
[10] Crews, J. H., Jr., "Stresses in a Uniaxially Loaded Sheet Containing an Inter-
ference-Fit Bolt; An Elastic Analysis," NASA Langley Working Paper No. 1955,
8 June 1972.
[11] Brombolich, L. J., "Elastic-Plastic Analysis of Stresses Near Fastener Holes,"
presented at the 1lth Aerospace Sciences Meeting of the American Institute of
Aeronautics and Astronautics, Washington, D.C., 10-12, Jan. 1973.
[12] Allen, M. and Ellis, J. A., "Stress and Strain Distribution in the Vicinity of
Interference Fit Fasteners," Technical Report AFFDL-TR-72-153, Wright-
Patterson Air Force Base, Ohio, Jan. 1973.
[13] Paris, P. C., Gomez, M. P., and Anderson, W. E., The Trend in Engineering,
Vol. 13, No. 1, University of Washington, Jan. 1961.
[14] Emery, A. F., Journal o/ Basic Engineering, Transactions, Series D, March
1966, pp. 45-52.
[15] Emery, A. F. and Walker, G. E., Jr., "Stress Intensity Factors for Edge Cracks
in Rectangular Plates with Arbitrary Loadings," ASME Paper No. 68-WA/
MET-18.
[16] Emery, A. F., Walker, G. E., Jr., and Williams, J. A., Journal of Basic Engineer-
ing, Transactions, American Society of Mechanical Engineers, Vol. 91, Series D,
No. 4, Dec. 1969, pp. 618-624.
[17] Bueckner, I-I. F., Zeitschrifl [ur Angewandte Mathematik und Mechanik, Vol.
51, 1971, pp. 97-109.
[18] Schmidt, R. A., "An Approximate Technique for Obtaining Stress Intensity
Factors for Some Difficult Planar Problems," paper presented at Fracture and
Flaws Symposium, Albuquerque, N. M., 2 March 1973.
[19] Grandt, A. F., Jr. and Gallagher, J. P., "Developing An Infinite Life Design
Procedure for Fastener Holes Utilizing Fracture Mechanics," Technical Mem-
orandum AFML-LLP-72-3, Air Force Materials Laboratory, WPAFB, Ohio,
Sept. 1972.
[20] Rice, J. R., International Journal of Solids and Structures, Vol. 8, No. 6, June
1972, pp. 751-758.
[21] Orange, T. W. in Stress Analysis and Growth of Cracks, Proceedings of the
1971 National Symposium on Fracture Mechanics, Part 1, ASTM STP 513,
[22] Grandt, A. F., Jr., "A General Stress Intensity Factor Solution For Fastener
Holes," Technical Memorandum AFML/LLP 73-7, Air Force Materials Labora-
tory, Wright-Patterson AFB, Ohio, June 1973.
[23] Seely, F. B. and Smith, J. O., Advanced Mechanics o/ Materials, Wiley, New
York, 1952.
[24] Harris, H. G., Ojalvo, I. U., and Hooson, R. E., "Stress and Deflection Analysis
of Mechanically Fastened Joints," Technical Report AFFDL-TR-70--49, Wright
Patterson Air Force Base, Ohio, 1970.
[25] Paris, P. C. and Sih, G. C. in Fracture Toughness Testing and Its Applications,
[26] Pook, L. P. in Stress Analysis and Growth of Cracks, Proceedings of the 1971
National Symposium on Fracture Mechanics, Part 1, A S T M STP 513, American
[27] Bucci, R. J., Clark, W. G., and Paris, P. C. in Stress Analysis and Growth of
Cracks, Proceedings of the 1971 National Symposium on Fracture Mechanics,
Part 1, A S T M STP 513, American Society for Testing and Materials, 1972, pp.
177-195.
(3RANDT AND GALLAGHER ON MECHANICAl. FASTENERS 297
[28] Paris, P. C., Bucci, R. J., Wessel, E. T., Clark, W. G., and Mager, T. R., "Very
Low Fatigue Crack Growth Rates in A533 Steels," presented at the Fifth Na-
tional Symposium on Fracture Mechanics, 1 Sept. 1971, Urbana, Ill.
[29] Bucci, R. J., Paris, P. C., Hertzberg, R. W., Schmidt, R. A., and Anderson, A. F.
in Stress Analysis and Growth of Cracks, Proceedings of the 1971 National
Symposium on Fracture Mechanics, Part 1, A S T M STP 513, American Society
for Testing and Materials, 1972, pp. 125-140.
[30] Paris, P. C., Bucci, R. J., Wessel, E. T., Clark, W. G., and Mager, T. J. in Stress
Analysis and Growth of Cracks, Proceedings of the 1971 National Symposium
on Fracture Mechanics, Part 1, ASTM STP 513, American Society for Testing
[31] Vasquez, J. A., "A Quantitative Continuum Approach to Fatigue Crack Propaga-
tion," presented at the Fifth National Symposium on Fracture Mechanics, 1
Sept. 1971, Urbana, Ill.
[32] Sehmidt, R. A. and Paris, P. C. in Progress in Flaw Growth and Fracture Tough-
ness Testing, A S T M STP 536, American Society for Testing and Materials,
1973, pp. 79-94.
[33] Fenner, A. J., Owen, N. B. and Phillips, C. E., Engineering, Vol. 171, 1951,
p. 637.
[34] Harrison, J. D., "An Analysis of Data on Non-Propagating Fatigue Cracks on a
Fracture Mechanics Basis," The Welding Institute Research Report No. E20/
3/69, The Welding Institute, Cambridge, England, March 1969.
[35] Frost, N. E., Engineer, Vol. 200, No. 464, 1955, p. 501.
[36] Frost, N. E. and Dugdale, D. S., Journal o[ the Mechanics and Physics o/Solids,
Vol. 5, No. 3, 1957, pp. 182-192.
[37] Frost, N. E., Proceedings, Institute of Mechanical Engineers, Vot. 173, 1959,
pp. 811-827.
[38] Frost, N. E., Journal of Mechanical Engineering Science, Vol. 1, No. 2, Sept.
1959, pp. 151, 170.
[39] Frost, N. E., Journal of Mechanical Engineering Science, Vol. 2, No. 2, 1960,
p. 109.
[40] Frost, N. E., Journal of Mechanical Engineering Science, Vol. 5, No. 1, March
1963, pp. 15-22.
[41] Frost, N. E. and Greenan, A. F., Journal of Mechanical Engineering Science,
Vol. 6, No. 3, Sept. 1964, pp. 203-210.
[42] Frost, N. E. and Grenan, A. F., Journal of Mechanical Engineering Science,
Vol. 9, No. 3, June 1967, pp. 234-240.
[43] Frost, N. E. and Dugdale, D. S., Journal of the Mechanics and Physics o[
Solids, Vol. 6, 1958, pp. 92-110.
[44] Frost, N. E., Nature, Vol. 187, No. 4733, 1960, pp. 233-234.
[45] Frost, N. E., Journal of Mechanical Engineering Science, Vol. 3, No. 4, 1961,
pp. 299-302.
[46] Frost, N. E., Journal of the Mechanics and Physics of Solids, Wol. 9, No. 3,
1961, pp. 143-151.
[47] Paris, P. C.,-data on 7075-T6 generated for Air Force C5A Independent Review
Team.
[48] Linder, B. M., "Extremely Slow Crack Growth Rates in Aluminum Alloy
7075-T6," Master of Science thesis, Lehigh University, 1965.
[49] Frost, N. E., Pook, L. P. and Denton, K., "A Fracture Mechanics Analysis of
Fatigue Crack Growth Data for Various Materials," Ministry of Technology,
National Engineering Laboratory Report No. Z2/1/69, Sept. 1969; presented at
the Fourth National Symposium on Fracture Mechanics, 1970.
T. R . B r u s s a t ~
Rapid Calculation of Fatigue Crack

Growth by Integration
REFERENCE: Brussat, T. R., "Rapid Calculation of Fatigue Crack Growth

by Integration," Fracture Toughness and Stow-Stable Cracking, ASTM STP
559, American Society for Testing and Materials, 1974, pp. 298-311.
ABSTRACT: A procedure is described for drastically reducing the computa-
tion time in calculating crack growth for variable-amplitude fatigue loading
when the loading sequence is periodic. By the proposed procedure, the crack
growth, r, per loading period is approximated as a smooth function and its
reciprocal is integrated, rather than summing crack growth cycle by cycle.
The savings in computation time result since only a few pointwise values of
r must be computed to generate an accurate interpolation function for
numerical integration. Further time savings can be achieved by selecting the
stress intensity coeff• (stress intensity divided by load) as the argument
of r. Once r has been obtained as a function of stress intensity coefficient
for a given material, environment, and loading sequence, it applies to any
configuration of cracked structure. Any of a broad range of prediction
models (such as "retardation models") may be used in conjunction with the
proposed procedure. Agreement with results obtained using cycle by cycle
summation is demonstrated, for two retardation models.
KEY WORDS: fatigue (materials), crack propagation, cyclic loads, mathe-
matical prediction
In the past, calculation of crack growth for variable-amplitude fatigue

loading has been carried out by essentially a cycle-by-cycle summation.
Even for one-time-only calculations on today's high-speed computers, this
brute-force procedure already requires excessive computation time. If fa-
tigue crack growth analyses are to be included in the iterative computations
of automated design, more efficient procedures are crucially needed.
A procedure is described for streamlining the calculation of crack length,
a, as a function of time, t, under a periodic fatigue loading function, P ( t ) .
1 Research specialist, Fatigue and Fracture Mechanics, Science and Engineering,

Lockheed-California Company, Burbank, Calif. 91520.
298
Copyright* 1974 by
BRUSSAT ON RAPID CALCULATION BY INTEGRATION 299
Typically, one to three orders of magnitude reduction in computer time can

be achieved, compared to cycle-by-cycle summation.
Although not required for the proposed procedure, P(t) usually consists
of an ordered sequence P1, P2. . . . Pr of discrete loading cycles. Let (~a)~
denote the crack growth due to the jth loading cycle in this sequence. In the
proposed procedure, the crack growth per period, r, defined by
T
r= E (Aa)j (1)
j=l
is calculated as a smooth function of crack length, and its reciprocal is inte-

grated numerically to obtain the crack growth lifetime L in number of
periods:
L-= f0 L dt= fa "'~~ r(a)

da (2)
Initial
This integration replaces the cycle-by-cycle procedure of calculation, which

involves a summation over perhaps millions of intervals. Since many nu-
merical methods of integrating Eq 2 are well-known, attention here is
limited to estimation of the crack growth per period, r, as a function of
crack length.
The concept of using numerical integration in calculating crack growth
for variable-amplitude fatigue loading was used in an earlier paper ([1],2
p. 135). More recently, the integration concept proved useful in simulated
design and analysis of damage tolerant aircraft structure [2, 3]. In each of
these references the effects of previous loadings were neglected in the crack
growth prediction model used. In contrast, this paper addresses the case in
which crack growth per period r is calculated by prediction models such as
the "retardation models" of Wheeler[4] and of Willenborg et al[5], which
include effects of previous loadings.
The Existing Summation Procedure

Suppose that a structure containing an initial crack of length ai.i.a~ is to
be subjected to an infinite sequence P1, P2 . . . . of fatigue loading cycles.
Figure 1 diagrams the existing summation procedure for calculating the
growth of the crack.
Starting at the initial crack length, an increment of crack growth (~a)l
(caused by P~) is calculated in accord with some crack growth prediction
model such as that of Wheeler[4] or Willenborg et al[5] and added to
*"The italic numbers in brackets refer to the list of references appended to this
paper.
(START)
Let a 1 = ainitial; i = 1
!Aa)j = f~aj,Pj,...) [
~ Crack I
Growth
Life
= L=j/T
(END)
FIG. 1--Existing summation procedure.
ainttial. From this new crack length, a2, and the next loading cycle, P2, the
next crack-growth increment (aa)e is calculated. This iterative process is
repeated until the final or "critical" crack size is exceeded, which marks the
end of the crack growth life.
In the past, modifications have been developed to improve the efficiency
of this procedure. For example, when several identical load cycles of equal
magnitude occur, the increment of crack growth due to all of them com-
bined may be computed simultaneously and added to the crack length as a
single entity. Even with such a modification, the essence of the procedure
is a brute-force summation of small increments of crack growth taken over
the entire lifetime. The longer the life, the more time-consuming is the
computation.
The Proposed Integration Procedure

The Basic Concept
Drastic reductions in computation time are possible if the calculation
centers around a numerical integration rather than a summation. However,
in order to use integration in the manner described later it is necessary to
represent the loading sequence as a periodic function of time P(t), with
period, T, as illustrated schematically in Fig. 2, where the period T is small
compared to the crack-growth lifetime. In aircraft design, for example, T
h T 9
LOAD
P(t)
TIME
FIG. 2--Periodic loading sequence.
could be one simulated flight or, if necessary, a sequence of 10 or so repre-

sentative flights from the anticipated loading history.
The basic concept behind the proposed integration procedure for calcu-
lation of crack growth under a periodic loading sequence P(t) is depicted
in Fig. 3. The procedure is as follows.
1. A small number n of distinct crack lengths al . . . . a~ are selected such
that the smallest is no larger than the initial crack length, ainiual, and the
largest is no smaller than the final crack length, a~i,~l.
2. At each selected crack length, the increments of crack growth ( a a ) j
due to each cycle in P(t) are computed. This is done using the same pre-
diction models as used in the summation procedure. However, the crack
FIG. 3--Proposed integration procedure.
length is not increased, but is held constant throughout one entire period
T. At the end of one period the increments of crack growth per cycle are
summed to compute the crack growth per period, r, for the selected crack
length.
3. A curve-fitting technique is used to connect the computed points and
obtain 1/r (the inverse of crack growth per period) as a function of crack
length.
4. The crack growth lifetime L in number of periods is calculated by
numerical integration of this function between the desired initial and final
crack lengths.
The rapidity of this approach is apparent from the fact that only a few
periods of the periodic sequence need to be considered by direct calcula-
tion. For example, the life might consist of thousands of periods, whereas
in Fig. 3 (which is schematic only, but typical nonetheless) only five pe-
riods are directly considered in the integration procedure. Rapid numerical
techniques for integration are well known.
The number n of point values of crack length required for a good fit of
a versus 1/r can depend on the configuration of the cracked structure. For
example, the crack growth rate in reinforced structure does not change
monotonically. Instead, as shown schematically in Fig. 3, the crack slows
down as it approaches a reinforcement, and then accelerates as the rein-
forcing member is bypassed. Therefore, at least the five points shown in
Fig. 3, and perhaps more, are required for an accurate curve fit.
It is explained in the following how the growth per period r can be made
always monotonic, so that fewer points are required for curve fitting, and
how the same function r can be applicable to various configurations, with-
out recomputing its point values.
The Use of Stress Intensity Coefficient

The integration procedure for crack growth calculation as just introduced
is general and can be used even when the crack growth rate does not de-
pend exclusively on the current and previous values of K, the crack tip
stress intensity.
However, significant further improvements in the computation efficiency
of the integration procedure are possible when used in conjunction with
"linear elastic fracture mechanics prediction models" (such as the predic-
tion models of Refs 4 and 5) in which exclusive dependence on K is as-
sumed. The term "exclusive dependence" here means that the fatigue
crack growth rate depends on the loading history P(t) and the macroscopic
configuration of the cracked structure only through the past and present
values of K. This assumption is valid[6] as long as the crack tip plastic zone
size Ry remains small compared to all macroscopic dimensions such as

crack size, a condition commonly referred to as "small-scale yielding."
In the small-scale yielding case the stress intensity can be written in the
form
K(t, a) = P ( t ) ~z(a) (3)
where 0t(a) is called the stress intensity coefficient. This factorization of K
is always possible since K, when not adjusted for crack-tip plasticity, is al-
ways proportional to the applied load P.
Since the stress intensity coefficient a is simply stress intensity K nor-
malized by load, it is often a known function of crack size. Otherwise, it
can be determined using classical elasticity. The schematic in the upper
right-hand corner of Fig. 4 depicts stress intensity coefficient as a function
of crack size for a crack in an unreinforced panel and in a reinforced panel.
Because values of stress intensity coefficient are monotonically related to
crack tip severity, the crack growth per period, r, is always a smooth,
monotonically-increasing function of ot. Thus, an adequate curve fit of
r(0t) can often be obtained with just two or three computed points, as the
schematic in the upper left-hand corner of Fig. 4 illustrates.
The improved integration procedure, using the stress intensity coefficient,
is depicted in Fig. 4. The crack growth per period r is computed at selected
values of stress intensity coefficient, and curve fitting is employed to con-
nect the computed points and express r as a continuous function of ~. This
function is transformed by means of the known relationship between stress
intensity coefficient and crack length for the configuration of interest. This
transformation produces i/r as a function of crack size, and numerical in-
tegration between the initial and final crack lengths results in the estimate
of the crack-growth life.
The curve at the lower left in Fig. 4 is the same as the curve in Fig. 3,
but this curve can be accurately obtained with fewer computed points us-
ing the approach depicted in Fig. 4, because crack growth per period varies
smoothly and monotonically with stress intensity coefficient.
There is another advantage of the approach depicted in Fig. 4. Once the
relationship between crack growth per period r and stress intensity coeffi-
cient ~ has been computed, it applies for any configuration of crack and
structure provided the same loading sequence and environment is applied,
the material is the same, and the same prediction model is used. Figure 4
illustrates this applicability of a single r(~) function to crack growth in an
unreinforced panel as well as a reinforced panel. As shown, the changes in
configuration affect the crack growth life only because the relationship be-
tween stress intensity coefficient and crack length has changed. Thus, the
pointwise calculation of crack growth per period at specific values of ~ (the
FIG. 4--Use of stress intensity coefficient in the integration procedure.
most laborious part of the computation) is done only once for all configu-
rations.
Beyond examining various crack configurations, a common exercise in
design is to proportionally change the loading sequence by a factor, search-
ing for improved sizing of the structure. Such proportional changes in P(t)
can also be handled without recomputing the pointwise values of r ( a ) . The
factor by which P(t) changes can be absorbed in the definition of stress in-
tensity coefficient. As a result, the a(a) functions shown in the upper right-
hand comer of Fig. 4 are merely shifted vertically by this factor prior to the
transformation and iteration steps.
It is worth noting that instead of using a prediction model, the function
r ( a ) can be experimentally determined for the given loading spectrum, en-
vironment, and material, using any simple specimen configuration. Then
this function, together with the applicable relationships between stress in-
tensity and crack length (determined from analysis), can be used in pre-
diction for configurations that would be more expensive and difficult to
test. This semi-empirical approach may prove to be extremely useful in
design substantiation. However, the emphasis of the present paper is on the
case in which pointwise values of r are calculated, rather than experimen-
tally determined.
Pointwise Calculations of Crack Growth per Period

The calculation of crack growth per period r at selected values of stress
intensity coefficient a is described in the following. This procedure, dia-
grammed in Fig. 5, is similar to the existing summation procedure (see
Fig. 1) and therefore is straightforward except for an initialization step
that precedes each pointwise computation.
The initialization is needed because the crack growth life does not, in
general, begin at the times when the values of stress intensity coefficient se-
lected for pointwise computation occur. Rather, each of these a-values will
occur at some intermediate time in the life. The growth of the crack at that
time can depend strongly on prior values of cyclic stress intensity, as is
recognized, for example, in "retardation"-type prediction models.
One way to consider effects of prior stress intensity values is in terms of
material state. At any moment during crack growth, past values of K have
determined the present material state at the crack tip. The present applied
(START) INITIALIZE OTHER

MATERIAL -- STATE VARIABLES
[ SELECT
'Otk
Rep I ....
AND LET j --- 1
I
v (,~ a)j = fe(Pi Otk, Repj,...)
j= .11 I'
I
RePi + 1 ==fl (Pj ak' Repj'"')
NO ~ %= ~(zxa)j
lEND)
FIG. 5--Crack growth during one period.
K carries the material from its present state to a new material state. Mate-
rial state is described by material state variables, one of which is the stress
intensity coefficient. Another, as in the models of Refs 4 and 5, may be
Rep, the distance from the crack tip to the elastic-plastic interface (the
point ahead of the crack tip separating material that has been plastically
deformed by present and prior loadings from material not yet plastically
deformed).
It is the material state variables other than stress intensity coefficient
that must be initialized at each selected value of a prior to computing crack
growth per period.
Initialization of Rep in the models of Wheeler [4] and Willenborg et al
[5] is straightforward. Let/'max be the loading cycle in the sequence P(t)
with the highest peak. Immediately after Pmax has been applied, Rev will
equal Rv~max~, the plastic zone size for that highest peak. Therefore by
starting the periodic loading at Pmax, Rep is equal to Rv~max),and the proper
initial conditions are established for the next loading cycle in the sequence.
For other prediction models, other material state variable values (such
as the minimum stress intensity required for the crack tip to open, for ex-
ample) may be operative that are not so easily initialized at the selected
values of stress intensity coefficient. For each of these material state vari-
ables it should be possible to assume an initial value and correct it by itera-
tion through a few periods. During the iteration the stress intensity coeffi-
cient would be held constant. As a result, the stress intensity K would be
periodic with period T during the iteration (because P(t) is periodic).
Since the material state variables are functions of K only, their correct
values would also be periodic with period T, as long as the stress intensity
coefficient tz were held constant. Thus, the iteration to correct the assumed
values of material state variables other than tz should converge to a set of
periodic functions. (Convergence properties of this initialization procedure
would have to be verified for the particular prediction model in question.)
The computation of crack growth per period r at selected values of stress
intensity coefficient 0~ is carried out as shown in Fig. 5. After selecting the
value of ~ to be considered, the other material state variables are initialized
as just discussed. Then for the first cycle in the loading sequence, an incre-
ment of crack growth (Aa)l is calculated according to the prediction model
being used. The material state variables other than ~ are adjusted in prepa-
ration for the next loading cycle. For the next cycle, the same procedure is
repeated to compute the growth increment (Aa)2 due to that cycle. When
every loading cycle in the period has been considered, the growth incre-
ments are summed to arrive at r, the crack growth per period, for the se-
lected value of ~.
Then a new value of tx is selected and the entire procedure is repeated.
This provides the points leading to the r(a) function plotted in the upper
left-hand comer of Fig. 4. The rest of the integration procedure for crack
growth calculation has already been discussed.
Application and Comparison
Two Existing Prediction Models
It should be possible to use the integration procedure in conjunction with
any crack growth prediction model. The choice of prediction model affects
only the equations used to compute (aa)j, the crack growth per cycle, for
use in Eq 1. As examples of how (Aa)j is calculated, the prediction models
of Wheeler[4] and Willenborg et all5] are summarized here. This provides
background for a comparison, described in the section on comparison, and
at the same time emphasizes the distinction between the calculation proce-
dure itself and the prediction model selected for use within the procedure.
Let Kmin~ and Kmaxj denote, respectively, the stress intensities caused by
the minimum and maximum values of load for the jth cycle in the ordered
sequence P(t) at the appropriate value of stress intensity coefficient a. If
there were no effects of prior loadings on material state variables other than
crack length, then the crack growth could be determined directly from the
experimentally-determined crack growth rate function for constant-ampli-
tude fatigue loading
da
dN const-ampl ~-~f (Kmin' Kma x ) (4)
In the prediction models of both Refs 4 and 5, the extent of crack-tip
plasticity is used to characterize the effects of prior loadings on material
state. Let ~ys denote the tensile yield strength of the material. For small-
scale yielding the plastic zone size due to the ]th cycle in P(t) is
__ [Kmaxi ~ z
) (5)
where c~ is a selected constant. Let Repj. denote the distance from the crack
tip to the elastic-plastic interface before the ]th cycle, and let (aa)~ denote
the increase in crack length during the ]th cycle. Then
Repj---- M A X ( Rep,j_ O - - (Aa)(j'-I ), Ru~) (6)

where the function MAX (x,y) takes the value of the larger of its two ar-
guments.
Wheeler[4] defines a retardation parameter c~j by
where m is an empirical constant. Crack growth is calculated from

(Aa)j:cp i f(Kminj, K .... j) (8)
The function f is crack growth rate for constant-amplitude loadng, Eq 4.
In the model of Willenborg et al[5], Kmll~and Kmax are replaced by "ef-
fective" stress intensity values K,nin<e) and Kmax<e>,where
Km,xi~ = M A X 0, K ..... .j-Km..~j R~

(9)
Kmin.i (e) = M A X 0, Kminj -- K .... j Ry--~-.
The crack growth due to the jth cycle is calculated by

(Aa)j:f(Kminj (e), K . . . . i(~ (10)
where again, f is defined by Eq 4.
Comparison
A calculation was made to compare predicted crack growth lives using
the integration procedure with results from the summation procedure. For
the selected example, values of the crack growth life L computed by the
summation procedure are available in Refs 4 and 5 for the Wheeler and
Willenborg et al prediction models, respectively.
In the example, a semicircular surface flaw of initial depth 0.14 in. in a
0.69-in. plate of D6AC steel (%.~=206 ksi) is subjected to a block-type
loading sequence with 13 stress levels and is grown until the crack is
tangential to the back surface of the plate. The constant-amplitude crack
growth rate function and plastic-zone-size formula are given in Ref. 4. The
stress history g(t) is plotted in Ref 4 and tabulated in Ref 5. The stress in-
tensity formula is
V 1.21 ~r a
K:M.~ 2.46_ 0.212(~r/c~ys)2
For the back surface coefficient M, Wheeler[4] uses
a 2
M = 0.899 + 0.184( 0~69 ) + 0.279 ( - 0 . ~ - ) +0.108 (0--~69) 3
whereas Willenborg et al[5] obtain M from Fig. 4 of Ref 7.

For the integration procedure, K is factored, as in Eq 3, into stress in-
tensity coefficient tz(a), a function only of crack length, times P(t), a func-
tion only of the stress sequence. For the example the sequence of plasticity-
corrected values of stress are used for P(t),
p= o-
V1 0.212
2.46 (~/~ys) ~
while ~ is given by
l / 1 ~ - 1 ~r a
~ ~..46
(For small-scale yielding ~ is approximately equal to P. Here, for example,
P and ~ always differed by less than 0.7 percent. Thus, for purposes other
than this careful comparison, this plasticity correction to ~ could be omit-
ted, and it could be assumed that P = ~.)
The total crack growth time L was calculated by the integration proce-
dure using both prediction models. The rate function r was computed for
each model at only three values of t~ (~--0.4410, 0.8000, and 1.5179).
The interpolation function consisted of segmented straight lines in log(~)
versus log (r), connecting the computed points.
The results are summarized in Table 1. For the Wheeler model ( m = 1.3)
the result was L = 2 6 . 2 4 blocks, compared to L = 2 6 blocks calculated by
summation and reported in Ref 4, a difference of just under 1 percent. For
the modal of Willenborg et al, the integration result was L = 31.45 blocks,
compared to L = 32 blocks calculated by summation and reported in Ref 5,
a difference of under 2 percent.
Discussion
The slight differences observed in Table 1 between crack growth life
computed by integration as opposed to summation result from differing as-
sumptions about the variation in stress intensity coefficient ~ during one
period. In the existing summation procedure the crack length is incremen-
tally increased after each loading cycle. In general, this causes incremental
changes in stress intensity coefficient within a period. In the integration pro-
TABLE 1--Computed crack growth lijetimes.
Computed Life in Blocks

Percent
Model of: by Summation by Integration Differences
Wheeler [4] 26 26.24 0.9%
WiUenborget al [5] 32 31.45 1.7%
cedure a is held constant ~ within the period while computing r. Clearly, if

the average change per period in the value of a is small enough, the lives
computed by the two procedures must agree.
Suppose, as in the prior example, that a increases monotonically with
crack length from a minimum value ~mi,. at a.=ai,fitiab to a maximum value
...... at a=a,~,,.~. Then the average factor increase per period in the stress
intensity coefficient is
/ ~ \~IL
A~=( .... ) (11)
\ t~,min]
For this example using the Wheeler prediction model, • 1.049; that is,
the average change per period in the value of a was 4.9 percent. The small-
ness of this "input error" somewhat explains the smallness of the "output
error" shown in Table 1.
Note from Eq 11 that the "input error" of the proposed integration pro-
cedure, ( ~ l ~ - 1 ) , decreases exponentially as L (the number of periods in
the life) increases. At the same time, it should be clear that the relative
rapidity of the integration procedure (in comparison to the summation pro-
cedure) increases in direct proportion to L. For many practical cases the
crack growth life L might be 1000 or even 10 000 periods. For such large
L, both the accuracy and the rapidity of the integration procedure would be
much greater than in the prior example.
Conclusions
Calculation of fatigue crack growth for a periodic loading sequence by
computing crack growth per period as a function and integrating its recipro-
cal (Fig. 3) can be orders of magnitude more rapid computationally than
summing crack growth cycle by cycle through the lifetime. If the selected
prediction model is based on fracture mechanics, the rapidity of this "inte-
gration procedure" can be further improved by using the stress intensity
coefficient a in the manner illustrated in Fig. 4.
The proposed integration procedure will give the same results as brute-
force summation for periodic loading sequences provided the crack growth
life L consists of many periods. An example calculation, using the "re-
tardation" models of Wheeler[4] and Willenborg et al[5], has shown that
excellent agreement can be obtained between the integration and summa-
tion procedures even when L is as short as 26 periods (Table 1 ).
3It should be emphasized that the constant value of tz during the loading period is
not the value at the start or end of the period, but is representative of the midpoint
of the period. This fact improves the agreement between the integration and summa-
tion procedures.
BRUSSAT ON RAPID CALCULATION BY INTEGRATION 31 1
Acknowledgment
The m a n y valuable suggestions of C. S. Davis are gratefully acknowl-
edged.
References
[1] Brussat, T. R. in Damage Tolerance in Aircraft Structures, A S T M STP 486,
[2] Ekvall, J. C., Brussat, T. R., Liu, A. F., and Creager, M., "Preliminary Design of
Aircraft Structures to Meet Structural Integrity Requirements," A I A A Journal
of Aircraft, Vol. 11, March 1974, pp. 136-143.
[3] Ekvall, J. C., Brussat, T. R., Liu, A. F., and Creager, M., "Engineering Criteria
and Analysis Methodology for the Appraisal of Potential Fracture Resistant Pri-
mary Aircraft Structure," AFFDL-TR-72-80, Air Force Flight Dynamics
Laboratory, Wright-Patterson AFB, Ohio, Sept. 1972.
[4] Wheeler, O. E., "Spectrum Loading and Crack Growth," ASME Transactions,
Journal of Basic Engineering, Vol. 94, March 1972, pp. 181-186.
[5] Willenborg, J., Engle, R. M., and Wood, H. A., "A Crack Growth Retardation
Model Using an Effective Stress Concept," AFFDL Tech. Memo 71-1-FBR, Air
Force Flight Dynamics Laboratory, Wright-Patterson AFB, Ohio, Jan. 1971.
[6] Rice, J. R. in Fatigue Crack Propagation, A S T M STP 415, American Society for
[7] Engle, R. M., Jr., "CRACKS, A Fortran IV Digital Computer Program for Crack
Propagation Analysis," AFFDL-TR-70-107, Air Force Flight Dynamics Labora-
tory, Wright-Patterson AFB, Ohio, Oct. 1970.

STP 559-1974

Uploaded by

Copyright:

Available Formats

STP 559-1974

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

STP 559-1974

Uploaded by

Copyright:

Available Formats

FRACTURE TOUGHNESS AND

Proceedings of the 1973

ASTM SPECIAL TECHNICAL PUBLICATION 559

List price $25.25

~(~l~ AMERICAN SOCIETY FOR TESTING AND MATERIALS

Printed in Baltimore, Md.

The 1973 National Symposium on Fracture Mechanics was held at the

Fracture Toughness Test Methods for Abrasion-Resistant White Cast

Development of Fast Fracture in a Low Alloy Steel--J. c. RADON AND

Acoustic Emission from 4340 Steel During Stress Corrosion Cracking--

Effects of Shot-Peening Residual Stresses on the Fracture and Crack-

Fracture Properties of a Cold-Worked Mild Steel--E, J. RIFLING 59

More on Specimen Size Effects in Fracture Toughness Testing~J. 6.

Correlation Between Fatigue Crack Propagation and Low Cycle Fatigue

Readers of this volume will not be disappointed with regard to novelties

Department of Mechanical Engineering,

Copyright*1974 by ASTMInternational www.astm.org

Fracture Toughness Test Methods for

REFERENCE: Diesburg, D. E., "Fracture Toughness Test Methods for

T h e mining i n d u s t r y uses a b r a s i o n - r e s i s t a n t m a t e r i a l s in large q u a n t i t y

1 Senior research associate, Ann Arbor Research Laboratory of Climax Molyb-

Copyright9 1974 by ASTM lntcrnational www.astm.org

2 Greenberg, H. D. and Clark, W. G., Jr., Metals Engineering Quarterly, American

TABLE 1--Chemical analysis of the irons.

Element, weight percent

L ~ - "CLIP GAGE SUPPORT

FIG. 1--Compact specimen.

A cantilever extensometer, calibrated to meet the requirements of

Results and Discussion

0.03 0.06 0.09

to meet all of the requirements outlined by ASTM E 399-72. The

5B, c, a, L, and W are definedin Fig. 1.

intensity required to propagate a fatigue crack was usually very close

I HT 1 0.013 0.765 (19.4) A 23.5 (25.8) 0.48 23.5 (25.8)

H T = heat treated; A C = as-cast and no stress relief. m

Toughness of Common Abrasion-Resistant White Cast lrons

FIG. 4--Microstructures in 20Cr-2Mo-lCu white cast iron after heat treating at

TABLE 3--Fracture toughness of common white cast irons.

A=austenite, M=martensite (listed in the order of decreasing volume percent).

TABLE 4--Description of Irons 1 through VI.

"M---martensite; A=austenite; P=pearlite (listed in the order of decreasing

Development of Fast Fracture

REFERENCE: Radon, J. C. and Pollock, A. A., "Development of Fast

T h e objective of this p r o j e c t was to c o m b i n e three techniques: f r a c t u r e

1Research fellow, Mechanical Engineering Dept., Imperial College, London,

TABLE 1 Material characteristics.

mation events that precede fracture and is used in engineering inspection.

Material, Specimen Geometry, and Fracture Mechanics Results

SAW 0.)1' STAR'rER P L O P A G ~ " r l N ~ ~ RACI~

FIG. 1--DCB specimen (straight edge type).

emission transducers were attached. The specimens were tested in a Tinius-

160 Initiation) 2 in. plate &

-~ .135~ ~ -55~ ~ -32~ g8 -19~ a -8~

FIG. 4--Emission ]rom first set of five specimens.

nz=number of emissions with amplitude exceeding 500 ~V

FIG. 5--Variation of amplitude distribution with temperature.

TABLE 2--Variation of amplitude distribution parameter during test at --IO~

Emission Count Amplitude

evidence that deformation events prior to fast fracture tend to be larger at

FIG. 6--Emission from Specimen D21 (--IO~