Validation of The Fracture Mechanics Test METHOD EGF P1-87D (ESIS P1-901ESIS P1-92)

Fatigue Fracl. Engng Maler. Struct. Val. 16, No. 11, pp. 1231-1284, 1993 8756-758X193 $6.00 + 0.
00
Printed in Great Britain. All rights reserved Copyright 0 1993 Fatigue & Fracture of
Engineering Materials & Structures Ltd
VALIDATION OF THE FRACTURE MECHANICS TEST

METHOD EGF P1-87D (ESIS P1-901ESIS P1-92)
ANALYSIS OF AN EXPERIMENTAL ROUND ROBIN
K.-H. SCHWALBE,'B. HA YES,^ K. BAUSTIAN,'

A. CORNEC,'
R. GORDON,3 M. HOMAYUN4 and B. VOSS'
'GKSS-Forschungszentrum Geesthacht GmbH, Geesthacht, F.R.G.
'The Welding Institute, Abington, U.K.
)The Welding Institute, Abington, U.K., now with Edison Welding Institute, Columbus, Ohio, U.S.A.
4GKSS, now with MAN, Munchen, F.R.G.
5Fraunhofer Institut fur Werkstoffmechanik, (IWM) Freiburg, F.R.G.
(Received in Jinal form 15 July 1993)
Abstract-The elastic-plastic fracture mechanics test method EGF P1-87 I): EGF Recommendations for
Determining the Fracture Resistance of Ductile Materials has been validated by an extensive round robin,
covering fracture tests and their evaluation as well as scanning electron fractography. The results
confirmed the suitability of the procedure, but suggested some modificalions. The resulting procedure
ESIS PI-90 was then further modified and its actual designation is ESIS PI-92.
The present report provides comprehensive information on the determination of crack growth resistance
curves and initiation values. Particular emphasis is on single specimen techniques and on determining the
stretch zone width. Beyond the validation of the ESIS procedure the information given is pertinent to
elastic-plastic fracture testing in general.
CONTENTS
NOMENCLATURE
I INTRODUCTION
2 OUTLINE OF ROUND ROBIN
2.1 Draft procedure P1-87D
2.2 Round Robin
3 MULTIPLE SPECIMEN METHOD
3.1 First evaluation
3.2 Conclusions from the first evaluation
3.3 Second evaluation
3.4 Conclusions from the second evaluation
4 SINGLE SPECIMEN METHODS
4.1 Introduction
4.2 Evaluation procedure
4.3 Aa-prediction capability of the methods
4.4 Resulting shape of R-curves
4.5 Analysis of specific results
1231
1232 K.-H. SCHWALBE
et a1
4.6 Regression curves

4.7 Fracture toughness parameters
4.8 Conclusions
5 BLUNTING, STRETCH ZONE WIDTH, AND INITIATION

5.1 Introduction
5.2 Definition of task
5.3 Blunting
5.4 Stretch zone width
5.5 Initiation
6 CONCLUSIONS FOR THE FINAL PROCEDURE-ESIS P1-92

6.1 Measurement of crack growth on fracture surface
6.2 Blunting line
6.3 Data spacing
6.4 Curve fitting
6.5 Fracture parameters
6.6 Single specimen methods
Acknowledgements
7 REFERENCES
NOMENCLATURE
a = crack length
a, = initial crack length
a, = estimated initial crack length (when using single specimen method)
Aa = crack growth
AaB = crack growth due to blunting
Aa,,, = crack growth at termination of test
Aamax= valid crack growth range
AaSm = critical stretch zone width determined on one specimen, representing the average of 9 Aa,,,,, values
Aaszwi = refers to Aaszw for individual specimens as determined by the participants
&is,, = Aasm for a given material, averaged over at least three specimens
Zszwl = refers to &isZwfor a given material as determined by individual participants
A U , ~=, local
~ stretch zone measurement taken at 9 locations on the fracture surface
Aa$, = refers to the Aa,,, for a given material as determined in Action 4 in Section 5
B = specimen thickness
B, = net thickness of specimen
d, = factor relating Aa, to J as defined by Eq (5.1)
d,* = factor relating AaB to J as defined by Eq (5.2)
E = Young’s modulus
F = applied load
F, = maximum fatigue load applied during fatigue cracking
J = J-integral
J, = J uncorrected for crack growth
J,,,, = J corrected for crack growth
Jg = J at upper limit of J-controlled crack growth
J, = J at initiation of stable crack growth
JiBLiR = J, as determined by the intersection of the blunting line with the crack growth resistance curve
JisZwiBL = Ji as determined by the intersection of Au,,, with the blunting line
JisZwiR= Ji as determined by the intersection of Aa,, with the crack growth resistance curve
J,,% = validity limit for J
Validation of the fracture mechanics test method 1233
Jo,2= J at 0.2 mm crack growth including blunting

= J at 0.2 mm crack growth offset to the blunting line
K,= maximum stress intensity factor during fatigue cracking
R = stress ratio during fatigue cracking
V = mouth opening displacement
W = specimen width
6 =crack tip opening displacement: the subscripts have the same meaning as those of J
uo= reference stress as defined in Appendix 7 of PI-87D
uy = yield strength
uu = ultimate tensile strength
rp = dectrical potential drop
Abbreviations
ACPD = alternating current potential drop technique
BL, =blunting line based on formula given in PI-87D
EL, =blunting line based on diagram given in PI-87D
CTOD = crack tip opening displacement
DCPD = direct current potential drop
PUC = partial unloading compliance
S Z W = stretch zone width
1. INTRODUCTION
In 1984 the EGF (now ESIS) Task Group on Elastic-Plastic Fracture Mechanics set up the
Working Party on Fracture Mechanics Testing Standards for developing a fracture mechanics test
method based on European experience. A unified test method was the final goal, to cover as many
aspects as possible instead of having a number of independent standards. As a first step, the
Working Party developed a procedure covering resistance against initiation and crack growth, both
in terms of J integral and crack tip opening displacement (CTOD).
The discussion within the Working Party was first published at the 6th European Conference
on Fracture[l]. The first complete draft was ready in 1987[2] and was named
EGF P1-87D: EGF Recommendations for Determining the Fracture Resistance of

Ductile Materials.
An extensive round robin was then defined aimed at verifying the Procedure and which had 21
laboratories from various European countries as participants. The management and evaluation of
the round robin was jointly carried out by The Welding Institute (TWI), Cambridge (management,
analysis of multiple specimen method), GKSS-Forschungszentrum Geesthacht, Geesthacht (analy-
sis of initiation related work), and FhG-Institut fur Werkstoffmechanik, Freiburg (analysis of single
specimen methods). It was financially supported by the Community Bureau of Reference (BCR)
of the European Commission.
The first evaluation of the round robin ended up with a number of unsolved problems which
where treated in a second evaluation.
This round robin analysis identified areas where the Procedure could be improved. Some of these
require further study, while others have been incorporated into the finalised version of the
Procedure, published under the designation ESIS P1-90, followed by the updated version ESIS
P1-92 [3].
This report, which is a condensed version of the project report presented to BCR[4], gives
the important details of the work undertaken to validate the Procedure and describe the
1234 K.-H. SCHWALBE
et al.
concIusions and recommendations drawn from this work. The report is structured in the following
sections:
Outline of round robin
Multiple specimen method
Single specimen methods
Blunting, Stretch Zone Width and Initiation
Conclusions and Recommendations for ESIS P1-92.
The authors of each section are identified separately, to aid in future communications, and all
references are given at the end of the report.
2. OUTLINE OF ROUND ROBIN (by B. Hayes)
2.1 Draft Procedure PI-87D

The draft Procedure has already been described in detail [2]. In this Section, therefore, only a
brief description of the main items will be given in order to provide a basis for the discussion of
the round robin.
Both compact (CT) and single edge notched bend (SENB) specimens are allowed, with a
preferred thickness of 0.5 W, where W is the specimen width.
A valid crack growth resistance curve (R-curve) is determined from which fracture parameters
can be derived. Since indirect techniques for measuring crack growth are still in a development stage,
Measure crack g r M at pcsitions 1 - 9

1 2 3 4 5 6 7 8 9 tronttaceot
I I I I I I
1'
t .-
t load line of
CT specimen
fatigue precrack
stretch zone
stable crack growth
~ ~~~
Fig. 2.1. Measurement of crack growth.

the multiple specimen method is considered the reference method. It requires the loading of several
nominally identical specimens to different displacements. The extent of ductile crack growth, Aa,
is measured on the fracture surface, as shown in Fig. 2.1. It should be mentioned here, however,
that the outline of the measurement positions gave rise to ambiguities in the round robin because
fractured specimens can exhibit a pronounced thickness reduction as a consequence of the
prefailure deformation.
Indirect techniques for measuring crack growth can be used if the difference between crack
extension measured on the fracture surface and the predicted value is no greater than 15% or
0.1 mm, whichever is the greater. In order to provide some guidance for this, an unloading
compliance and several electrical potential methods are described in an appendix to
P1-87D. However, the user is warned that those methods may not give accurate estimates of Ji
or di.
The evaluation of the J-integral or the CTOD, 6, is done using standard formulae, see Ref. [2],
and if the resulting J-Aa or 6-Aa data points are located within the validity box shown in Fig. 2.2,
they are regarded as characterising the crack growth behaviour of the material independent of
specimen size.
A curve of the type
J = A(Aa + C)" (2.1)
or
6 = F(Aa + G ) H (2.2)
is then fitted to the valid data points. The resulting curve is valid, if the data spacing requirements
illustrated by Figs. 2.3 and 2.4 are met. The abscissa value of that point where the curve leaves
the validity box is used to define four equal growth regions, each of which shall contain at least
one data point.
A further condition is required for data points in the first crack growth interval, namely only
data exhibiting ductile crack growth shall be used. This can be achieved by excluding data
J.6
Vaiidw box for crack i l l

growth resistance
0 0.25 0.5 015 1OAamu
Crack Growth Aa
Aa
CrackGrowth M
Crack Growth - M
Fig. 2.2. Validity box forming the area for valid crack Fig. 2.3. Jvalidity limits a) controlled by Aumaxb) controlled
growth resistance data. by Jm,,.
1236 K.-H. SCHWALBEet nl.
J.6
Crack Growth Aa
* -
CrackGrowth Aa
Fig. 2.4. 6 validity limits a) controlled by Aumaxb) controlled Fig. 2.5. Schematic of differently defined crack initiation
by J,,,. parameters.
from specimens exhibiting less than 0.2 mm crack growth and/or obtaining a more accurate
estimate of the blunting behaviour of the material, where relevant. The weak formulations for
blunting line determination in P 1-87D were replaced by strengthened and more accurate provisions
in the final version of the procedure, [3].
The following three fracture parameters can be determined with EGF P1-87D:
Ji/Si corresponding to the fracture resistance at initiation of ductile crack growth.
Two alternatives are offered for determining 4 and Si, Fig. 2.5:
The critical stretch zone width, Aa,,,, is measured in a scanning electron microscope (SEM),
and averaged for at least three specimens. The point on the crack growth resistance curve
corresponding to Aaszw defines initiation.
If the critical stretch zone width cannot be determined by this method then the intersection of
the blunting line with the crack growth resistance curve defines initiation. (Throughout the report,
this parameter is referred to as J i B L to avoid confusion.)
In the draft document it was believed that these ways of determining initiation would give the
same values. But, as it was demonstrated in the round robin, this is not the case.
corresponding to the fracture resistance at 0.2 mm of ductile crack growth
Jo.l,eL/SO,Z/BL
These engineering approximations to the initiation values are determined by the intersection of
the R-curve with the parallel to the blunting line offset by 0.2mm, Fig. 2.5.
corresponding to the fracture resistance at 0.2 mm of total crack growth

Jo,l/&o.2
These parameters are defined by the intersection of the line corresponding to constant total crack
growth of 0.2mm with the R-curve, Fig. 2.5.
The three parameters described above have to meet certain requirements in order to be “valid”,
however, these requirements did not play a significant r81e in the round robin and are therefore
not discussed further.
During the round robin and its evaluation the following points were found in particular to give
rise to problems. These were discussed extensively leading to some modifications in the final
procedure:
R-curve fit equation.

R-curve fit to data points within the validity box, instead of imposing the J,,,,,/S,,,,, limit after
the fitting procedure.
Measuring crack growth on the fracture surface of a deformed specimen.
Measuring the critical stretch zone width.
Data point distribution.
Definition of initiation.
2.2 Round robin

The round robin to validate the Procedure involved the testing of three materials with different
toughness behaviour. Therefore, the initial steps were to select three suitable materials, to fully
characterize their properties and to check their homogeneity. The materials chosen were two C-Mn
microalloyed steels, BS 1501-224-490B LT50 and BS 4360-50E, and an Al-Mg alloy, 5083-0. These
were labelled M3, M1 and M2, respectively.
The materials characterization tests showed the BS 1501 steel (M3) to be high toughness, with
the measured R-curves rising steeply and leaving the validity box after little crack extension. The
BS 4360 (Ml) steel was found to be a suitable medium toughness material, that is with its R-curves
leaving the validity box near the corner, while the 5083-0 alloy (M2) showed low toughness
behaviour, the R -curve being Aamaxlimited.
Another purpose of the round robin was to establish a reference material for the Community
Bureau of Reference (BCR). Such a material would be available for future use for verifying R-curve
testing facilities. The BS 1501 steel was purchased for this r61e and as such was tested by the
majority of participants in the round robin.
Twenty-one laboratories participated in the study; these are listed below:
Admiralty Research Establishment, Holton Heath, UK.

BAM, Berlin, FRG.
CEGB, UK.
CENIM, Madrid, Spain.
CISE spa, Milan, Italy.
Creusot-Loire Industrie, Le Creusot, France.
David Taylor Naval Ship R&D Center, Annapolis, USA.
ECN, Petten, The Netherlands.
EDF, Moret-sur-Loing, France.
Fraunhofer IWM, FRG.
GKSS, Geesthacht, FRG.
MPA, Stuttgart, FRG.
NII (HSE), UK.
Universidade do Porto, Portugal.
SCK/CEN, Belgium.
Statens Provningsanstait, Boras, Sweden.
Sulzer, Winterthur, Switzerland.
Technische Universiteit Delft, The Netherlands.
TWI, Cambridge, UK.
UKAEA, Harwell, UK.
UKAEA, Risley, UK.
1238 K.-H. SCHWALBE
et al.
Two types of specimen were tested: compact (CT) and single edge notched bend (SENB)
specimens. The specimen thicknesses were 25 mm, with 2.5 mm deep sidegrooves on each face. The
initial crack length to specimen width ratio, a,/W, was specified as 0.6 and all testing was performed
at room temperature.
As the principal testing method in the Procedure is the multiple specimen technique, participants
were supplied with sets of eight or ten specimens. Laboratories with the necessary facilities to
perform single specimen testing using unloading compliance or potential drop methods were asked
to test individual specimens for the multiple specimen R-curve also as single specimens. In this way,
100 single specimen R-curves were produced in addition to the multiple specimen curves. The
multiple and single specimen results were analysed by TWI and Fraunhofer IWM, respectively.
Most participants tested two of the materials, depending on the availability of specimens. The
table below shows the number of sets of specimens for each material.
M3, High Toughness MI, Medium Toughness M2, Low Toughness

CT Specimens 8 7 3
SENB Specimens 6 5 4
All the specimens in any one set were machined from one area of the original plate. The specimen
numbering system used was devised to relate the specimen to its location in the plate, thus
maintaining traceability.
In order to reduce possible sources of variability due to, for example, differences in initial
a,/ W ratios or sidegroove depths, the participants were asked to follow the guidelines, summarized
here:
1. Due to the different Aa spacing requirements for J-Aa curves and CTOD-Aa curves, it might
be difficult to produce valid R-curves in terms of both J and CTOD. Therefore, in the first
place produce a valid J R-curve and then, only if possible, derive a CTOD R-curve from the
data.
2. All specimens must be sidegrooved to a depth of 2.5mm on both sides.
3. After fatigue precracking, a,/ W should be 0.6.
4. The loading span for SENB specimens should be 200mm.
5 . For the high toughness steel (M3), test to the Aamueven though the J,,, limit would be exceeded.
6 . The tests should be carried out at room temperature which should be measured and reported.
7. The following values of material properties should be used:
High Medium Low

Toughness (M3) Toughness (M 1) Toughness (M2)
(BS 1501-224) (BS 4360 Grade 50E) (5983-0)
OY 358 N/mm2 353 N/mm2 171 N/mm2
0" 560 N/mm2 5 18 N/mm2 321 N/mm2
E 207000 N/mm2 207000 N/mm2 70000 N/mm2
V 0.3 0.3 0.3
8. For the determination of initiation parameters using the analytical blunting line approach, the
values of d, and ( T ~(eqn 5.1) to be used are as follows:
High toughness 0.320 309 N/mm2

Medium toughness 0.347 3 10 N/mm2
Low toughness 0.254 145 N/mm2
Finally, in order to evaluate the approach to determine initiation, eight laboratories carried out
stretch zone width (SZW) measurements on their tested specimens using a scanning electron
microscope (SEM). These results were supplemented by additional SZW measurements by GKSS,
who were responsible for the analysis of this part of the study.
3. MULTIPLE SPECIMEN METHOD (by B. Hayes and R. Gordon)
3.I First evaluation

The data sets were initially checked for obvious deviations from the procedure [2]. These checks
included specimen dimensions, fatigue precracking loads, a,/ W ratios, sidegroove depths, fatigue
crack curvature and crack growth measurements.
It should be noted that one laboratory who machined their own specimens, produced them with
a thickness of 20mm. These results have not been included in the pooled analyses due to the
reduced validity limit obtained with this size of specimen, however, the results fall within the scatter
bands for the materials M1 and M3 and are included in the relevant figures.
It was found that most of the a,/ W ratios actually achieved were greater than the 0.60 specified
in the instructions to participants. The range was greatest for the low toughness materials, i.e. 0.53
to 0.72. Initial data checks also revealed that it was virtually impossible to meet the crack growth
measurement requirements set out in the June 1987 draft of the procedure. These state that the
difference between the minimum and maximum measurement of crack growth on a specimen
should be less than 20% of the average crack growth, Au. For small crack extensions this
requirement is very onerous and only one specimen was within this specification. The requirements
were therefore subsequently relaxed.
There was some concern over how Aa had been measured by the participants. The June 1987
draft of the test procedure indicated that the position of the nine measurement points should be
relative to the root of the sidegroove and made no allowance for any lateral contraction of the
specimen in the region of interest. It became apparent at the participants’ meeting in Brussels in
February 1989 that different approaches to measuring Aa had been taken. Following the meeting,
all participants were asked to report how they had performed the measurements and approximately
half of the laboratories responded. Most had taken B, as the specimen thickness corresponding
to the greatest lateral contraction, and based their measurements on this. One laboratory had
avoided the problem by using the centreline of the specimen as a reference.
The crack growth, toughness and specimen dimension data for each set of results were stored
on computer. Each set of data was re-analysed according to the procedure using a computer
program specifically written for this purpose. The program allowed either individual sets of data
or groups of sets of data to be analysed in four ways. The simplest analysis produced an R-curve
using only those data with a crack growth greater than 0.2mm. The other three analyses gave
1240 K.-H. SCHWALBE
et al.
R-curves which were based on data with a crack growth greater than the "stretch zone width"
determined either by measurements or by one of the analytical blunting lines based on d, or d,*
(equation 5.1, 5.2).
For each of the materials, the J and CTOD data were re-analysed as complete groups. Following
initial studies of the results, some errors were found in the data. These were corrected by
participants and the re-calculated R-curves based on data with a crack growth greater than 0.2 mm
have been plotted (solid line) along with the data points in Fig. 3.1 to 3.3. The validity limits for
each of these curves have been included in these figures. For the high toughness material, M 3 , the
J,,, limit was ignored and the curve fitted to all points below the Aamaxlimit.
The coefficients for the offset power law fit to each of the groups of data are included in the
captions. It may be noted that the offsets in the offset power law are either small or zero.
From these figures it can be seen that the scatter of the data is very low. Although the scatter
of the CTOD data for M2 appears at first sight to be large, it is in fact comparable with the scatter
in the CTOD data for high and medium toughness materials once the difference in scales has been
taken into account. The difficulties in measuring very small values of CTOD should, however, be
appreciated.
The data obtained from SENB specimens are indicated in the figures by solid symbols. There
appears to be no systematic difference between the data obtained from SENB specimens and CT
500 , , , 1 , , , 1 ~ 1 , 1 , 1 ~ 1 , ,
- c T SEN8
- OLlb 3 .Lab 1
Vbb4 *Lib5
Abb8 Ahb7
E "1 :::::: 0
, c <-ff
. '*-<-K-- 0-
- Obb16
xbb19
J max p :------
(a) -
1 ;
I , 1 , 1 , , , 1 , 1 , 1 , 1 , 1 ,
0 02 0.4 0.6 0.8 1.0 12 11 16 1.8 2.0
i
l
-
l
I c T
l ~ l
SENB
~ l ~ l l l l i ~ l l l / L l CT l SEN8
.
1.4 OLab3 OGbi
10 Lab2 0 Lab1 I vLib4 *Lab5
' 0 Lab4 Lab2
1.2 - I0 Lab8
A
Lab8
Lab7
I +Lab12 .Lib18
I A Lab9
'E 1.0- 0.8 c

'b'
-1
0.8 -
0.6 -
g-
. o " -
0.4 + 0.2
7%
(bl - I
0 0 0.2 0.4 0.6 0.8 1.0 1.2 14 1.8 1.1) 2.0

0.01' I I a 1 ' " 1 % 1 , I a I 1 ,
0.0 0.2 OI( 0.8 0.8 1.0 1.2 1.4 1.6 1.8 2.0
CRACK GROWTH Aa rnm
CRACK GROWTH Aa rnrn
Fig. 3.2. Offset power law fit to data for the medium
Fig. 3.1. Offset power law fit to data for the high toughness toughness material, (a) Jo = 3 1 1.3(Aa)O440 (b)
material, (a) Jo = 6 8 0 . 7 ( A ~ )(b)
~ . So
~ ~= 0 . 8 3 8 ( A ~ ) ~ . ~ " 6, = 0 . 4 3 0 ( A ~ ) ~ . ~ ~ ~
specimens. It is also pertinent to note that despite the various degrees of experience in this type
of testing on the part of the participants, low scatter was obtained with three materials of very
different toughness behaviour.
Although the data for each particular material lie within a fairly narrow band, the shapes of the
R-curves returned by the participants for their sets of data do vary from a straight line to a power
law with or without an offset. To give a feel for the variation in these curves they have been drawn
in Fig. 3.4 for J . These curves are in reasonably good agreement over crack growth ranges of about
tT SEN8
OLab3 mbb6
ObbS *Lab13
AbbB *Lab21
.Lab 18
I I I I I , I , I I I 1 1 I I I
04 0.6 08 1.0 1.2 14 1.6 1.8 2.0
CRACK GROWTH Aa mm
CRACK GROWTH rnrn
Fig. 3.4. Offsetpower law fits determined by participants,(a)
Fig. 3.3. Offset power law fit to data for the low toughness for high toughness, (b) for medium toughness, (c) for low
material, (a) J, = 45.8(Aa 0.009)0.503 +
(b) 6, = 0.136(Aa)0.457 toughness material.
1242 K.-H. SCHWALBE
et a1
0.2 mm to 1 mm. At crack growths of greater than 1 mm, the curves tend to diverge considerably
as they do at the lower end of the crack growth range.
In order to determine whether this variability was due to the particular fitting programmes used,
the data were re-analysed set by set at TWI. The curves calculated by TWI were found to be very
similar to the returned fits. The variability in the returned fits would therefore appear to be a
function of the approach used for curve fitting and not due to differences between the laboratories.
The TWI analysis program also checked the validity of the R-curves for each data set and in many
cases the curves were invalid due to incorrect data point spacing or too few points within the
validity box. This was a particular problem with the high toughness material for which the first
two quartiles were to the left of Au = 0.2 mm and hence excluded from the curve fitting range.
Furthermore, as anticipated, it was generally not possible to obtain valid curves for both J and
CTOD from the same set of specimens.
As there was a lack of consensus on the best type of fit to use for R-curve data, it was decided
also to consider a linear fit, using the ordinary least squares approach, and a fit of the type
J = C,/(Aa). The C,/(Aa) fit had been put forward in the final report on the Nordic Round Robin
on J-R-curve testing [5].
Linear, and C , / ( A a ) fits to the J data for all three materials have been plotted in Fig. 3.5. These
figures show both the fitted lines and the original data. For the low toughness material, the linear
and C , / ( A a ) fits correspond very closely over the fitting range Au = 0.2 to Aa = Aumax.The offset
power law calculated for the same range is virtually the same as the C,/(Aa) curve (cf. Fig. 3.3(a)).
The coincidence of the linear and C J ( A a ) fits for the medium toughness materials is less close
but still quite good, with the offset power law falling between the two. With the high toughness
material, the limitations of the linear and C J ( A a ) fits start to become evident. The fixed exponent
of the C,/(Aa) fit does not allow a good description of the curvature, tending to give higher
toughness values a t low crack growth and lower values closer to Aamax. The linear fit is better in
the region of Aa,,, where it agrees well with the offset power law, but at low crack growth it cannot
model the curvature shown by the data.
Although the C , / ( A a ) fit gave the best reproducibility in the initiation parameter (see report
to BCR [4]), this type of fit is restricted by its fixed exponent. For larger specimens and greater
crack growth, the limitations of the fit would probably become more apparent.
The linear fit has the advantage over the offset power law of being easy to apply and more
statistically correct when data sets are as small as four points, however, for small crack extensions,
the model may be poor.
The offset power law would appear to model the true R-curve behaviour better, allowing a range
of curves from a power law to a straight line. The disadvantages of this type of fit is that three
parameters are being determined from as few as four points. The offset power law is a nonlinear
model which, strictly speaking, requires sophisticated fitting routines. The fitting method proposed
in the ESIS procedure is approximate and hence may be open to criticism.
3.2 Conclusions from the first evaluation

This study revealed a number of problems with the first draft of the procedure for determining
resistance curves [2] and also raised some basic questions about the approach. Before summarising
these, it should, however, be appreciated that the intercomparison programme produced an
unprecedented amount of resistance curve data from a wide range of organizations. Despite the
differing degrees of experience of the participants, the toughness measurements obtained showed
a remarkably low amount of scatter. This is very encouraging for the further development and
application of resistance curve testing.
Validation of the fracture mechanics test method I243
V h b 4 *Lab2
OhbB a h b e
AhbS Abb7
OLablO * L i b 1 7
Fig. 3.5. Linear and C& fits to J, data for (a) high toughness, (b) medium toughness, (c) low toughness
material.
The results of the comparison showed that it was very difficult to achieve a strictly valid R-curve.
Firstly, the validity requirements for crack growth measurements were excessively rigorous for
the small crack extensions studied here. Secondly, the data point spacing requirements were
impossible to meet for the high toughness material. This led to the data spacing quartiles being
modified and in ESIS P1-92 [3] they are now based on the material’s analytical blunting line, see
Section 6 . In addition, the exclusion of points with toughness values greater than J,,, or 6,, from
the curve fitting within the valid Au range caused unnecessary difficulties in obtaining sufficient
points. This was also changed in P1-92 where the J,,, or d,,, limit is applied after the curve has
been fitted.
1244 K.-H. SCHWALBE
et al.
3.3 Second eoaluation

After the first evaluation of the round robin results, several modifications were made to the draft
procedure[2] (see Section 6 ) prior to final issue as ESIS P1-92[3]. With respect to the multiple
specimen technique, the changes having the greatest potential effect were the new data exclusion
and data spacing requirements.
The spacing requirements still call for a minimum of one point in each of four quartiles, but
the new quartiles are based on the 0.1 mm offset to the blunting line which is also the lower
exclusion limit. Thus, the contradictory situation of defining data spacing quartiles below the lower
exclusion limit could no longer arise.
Secondly, the removal of an exclusion limit based on toughness and the imposition of the
maximum toughness limit, Jmax or h,,,, only after the fitting of a curve to the data greatly
increased the number of points available for regression in the case of high toughness materials.
This indirectly raised the question once again of data set size which had been d e v t e d many
times during the drafting of this procedure and others. The problem is one of being able to
keep testing to a minimum when necessary and yet still be confident of deriving meaningful
results.
A second question was also raised by this change to data selection for curve fitting: was it possible
to extend the regression range beyond Aamax as well as beyond J,,,? If the data were available, could
they not be usefully included in the derivation of an R-curve?
It had been intended to also look at the effect of allowing a degree of overlap of the inner data
spacing quartiles with the view to making the requirements more flexible. The results of the data
set size study indicated, however, a general lack of sensitivity of the derived fracture parameters
to the number of points in each fit. This, in turn, indicated the very probable lack of sensitivity
to the exact positioning of the spacing requirement quartiles. Thus it was decided to look more
closely at the issue of type of regression curve by investigating the “goodness of fit” of a series
of different equations applied to simulated data sets.
Finally, the individual data sets returned by participants were re-analysed according to the
modified procedure, P1-92. These results were then compared to those obtained from a re-analysis
using an ordinary power law fit instead of the offset power law.
Due to the limited space only the results obtained from the simulation studies will be presented
here.
The five different regression equations considered in the simulation studies were as follows:
J= A + CAa (3.11
J = CJAa (3.2)
J = C(Aa)” (3.3)
J=A(Aa+C)D (3.4)
J =A + C ,Aa + C,(Aa)*+ C,(Aa), (3.5)
The linear fit was chosen for its simplicity as was the CJAa fit which had originally been put
forward in the Nordic J-R-curve round robin test programme [5]. The ordinary power law was
felt to be a strong contender to replace the offset power law in the ESIS Procedure as the first
evaluation of the round robin had shown the value of the offset to be zero or very small. The
third order polynomial was included as polynomial fits are allowed in the DVM procedure
002, [6].
A suite of five computer programs was written to produce and analyse the simulated data sets
using the different proposed regression equations. The creation of the simulated data sets was done
by pooling all the data for the material of interest and sorting them into the four spacing quartiles
required in ESIS P1-92. One point would then be selected at random from each quartile to form
a simulated data set. For data sets with more than four points, additional points would be randomly
chosen from anywhere. In accordance with P1-92, the J,,, limit was ignored at this stage. Data
sets of 4, 5, 6, 7, 10 and 20 points were studied.
For each data set size 500 simulated data sets were created. The fracture parameters, Jo,2,Jo.z,eL
and Jg were calculated from each simulated set. The J0.2,BL values were determined using the
analytical blunting line. In accordance with P1-92, the slope of the blunting line was determined
using d,*.
J, was evaluated as this was a parameter of interest to some of the participants. Although strictly
J,,, varied from data set to data set, the calculation of J,,, for each set would have been excessively
complicated. Therefore a fixed value of J,,, was used based on the nominal dimensions of the
specimens (B = 25, a / W = 0.6). These values were:
high toughness, M3 J,,, = 367 N/mm

medium toughness, M , J,,, = 349 N/mm
low toughness, M2 J,,, = 197 N/mm
All of the simulations were done using J, data as it was assumed that the findings would be equally
valid for J,,,,, 6, and 6,,,,.
For each material, the means and standard deviations of the fracture parameter, JO,Z,BLhave been
plotted for the different data set sizes for Aumax= 1.2 mm, Fig. 3.6. For the high toughness material
the mean values are fairly insensitive to the number of points per data set, however, they do depend
on the type of fit. As severe problems were encountered with the polynomial fit for this material
1 B H"%,
9
9
a
8 a h &&n
P
A 0 0
m
rMMpar!w.**
v)
0 4 6 6 7 10 ?o
NUMBER OF POINTS IN SIMULATED DATA SETS

Fig. 3.6. Means and standard deviations of J0,2,BL
values determined from simulated data sets; (a) for high
toughness material (b) for medium toughness material (c) for low toughness material
1246 K.-H. SCHWALBE
el al.
and the results from the other two lower toughness materials showed this type of fit to have
serious drawbacks, it was decided to only consider the linear, C,/Aa, power, and offset power law
fits.
Figure 3.6 also shows the standard deviations of the simulated fracture parameter. There is both
an effect of type of fit and of data set size. The linear and offset power law fits produce the greatest
scatter in the fracture parameter and the CJAa fit the lowest. The standard deviations for the
C J A a are fairly constant for data sets up to 10 points. The other fits show a reduction in scatter
from 4 to 5 to 6 points. Increasing the number of points per set to 20 greatly reduces the scatter
for all fits.
With the medium toughness material, the mean fracture parameter values are fairly constant with
data set size. All five fits give very similar mean values. The ordinary power law shows an
unexpected result in that the standard deviations rise as the data sets are increased to 10 points.
However, as the standard deviations for 7 and 10 points are similar to those from the offset power
law, perhaps the unusual feature is the low scatter for sets of 4 and 5 points. Looking at the
coefficients of the power law fits, these are fairly constant with size of data set but the exponents
do change. The spread of exponents increases with the number of points and this will give an
increasing spread in fracture parameters.
The mean fracture parameter values for the low toughness materials also shows insensitivity to
the number of points per data set. Except for the polynomial fit, the standard deviations are
reasonably constant irrespective of data set size and insensitive to type of fit.
3.4 Conclusionsfrom the second evaluation

The third order polynomial fit is not suitable for multiple specimen R-curves as it is sensitive
to data set size and produces highly scattered fracture parameter values. The linear fit is
more predictable than the polynomial fit and is easy to perform, however, the scatter in parameters
is high and mean values found in the simulations tend to be greater than for the power law
fits.
The C J A a is very insentive to the number of points per data set and the calculated parameters
show remarkably low scatter. However, for the high toughness material, the mean fracture
parameters from this fit are very much greater than the other power law fits.
The power law and offset power law tend to give fairly consistent results, the offset power law
often producing greater scatter than the ordinary power law.
From the point of view of derived fracture parameters, the ordinary power law is, therefore, the
best of the types of fit considered in this simulation study.
4. SINGLE SPECIMEN METHODS (by B. Voss)
4. I Introduction
The first goal for all participants was to produce a series of tests following the multiple speci-
men method which would result in a valid J-Aa curve and, if possible, a valid 6-Aa
curve also. Participants were allowed to test individual specimens within the multiple speci-
men series using single specimen methods. For this it was proposed to test two specimens up
to Aamax( M 1 mm, depending on specimen size) and to stop one additional test in the crack
growth range 0 < Aa < 0.3mm in order to validate the R-curves for evaluation of initiation
parameters.
The three different single specimen methods, used by the participants, are abbreviated in the
following by
PUC Partial Unloading Compliance
DCPD Direct Current Potential Drop
ACPD Alternating Current Potential Drop
MSU Multiple Specimen Unloading
A total of 100 single specimen R-curves for the three different materials were delivered for this
evaluation:
8 1 CT-specimen tests
19 SENB-specimen tests
56 PUC-evaluations
44 PD-evaluations (ACPD or DCPD)
Some of these tests were performed using partial unloading compliance and a potential drop
method in parallel, thus resulting in two R-curves. The main effort was done on the reference
material M3 using CT-specimens.
The R-curves were defined by sets of J,-Aa points. The number of measured points varied from
about 10 to about 60, depending on the amount of total crack extension, the material, the method
and the laboratory. This resulted in a significantly greater number of valid data points for each
R-curve than the minimum number of tests specified for multiple specimen evaluations.
4.2 Evaluation procedure
The first goal of the evaluation was to unify the data handling after the J-Aa points had been
defined by the participants. Then the data were re-evaluated as defined in the draft procedure [2]
and in addition with a number of modifications including the final specifications of P1-92[3].
Part of the evaluations done for the multiple specimen series was repeated for those single
specimen tests loaded up to a sufficient amount of crack extension. Though this discussion is limited
to J-Aa curves using Jo without crack growth correction, the results will be valid for the
qualitatively equivalent R-curves in terms of other parameters as well.
Some of the participants established their individual blunting line from SZW measurements,
others calculated the slope of the blunting line following APPENDIX 7 of the procedure.
The influence of the blunting line definition on critical J-values is beyond the scope of this
chapter. Therefore the calculated blunting line slopes following APPENDIX 7 of the procedure [3]
were used for these evaluations.
Selected “extreme” R-curves were specially investigated and partly re-evaluated based on more
complete data sets than originally returned in order to identify problems that might be specific for
the type of material
the toughness level
the experimental method for monitoring crack growth
the laboratory.
4.3 Aa-prediction capability of the methods
The most important criterion for the applicability of any single specimen method is the capability
to predict the crack extension during the test. The only direct check is the comparison of the last
estimate of the indirect method (Aa (predicted)) with the crack extension measured on the fracture
surface after the test (Aa(measured)).
1248 K.-H. et al.
SCHWALBE
The correlation of measured and predicted Aa-values is shown in Figs 4.l(a) to (c) for the three
materials. The lines mark the ideal correlation and the upper and lower limits of errors accepted
by the procedure PI-92, i.e. the greater of 0.15 mm or 15% of the measured Aa-value. Different
methods and specimen types are plotted by different symbols.
For the medium toughness material (MI, BS 4360) positive and negative errors are nearly evenly
distributed. Positive errors dominate slightly, mainly due to the DCPD-results. With only one
exception all errors are within the tolerable error band.
‘1 M 1 / BS 4360 Gr.5OE //
- (a) s
2-
CT DCPO
1- + CT ACPO
7 M 2 / A15083 -0
E
1 -
3- Ibl
4
8 -
2 2-
3
3 - CT DCPD
P CT ACPD
u+n 1-
W
a -
n.
0
0 1 2 3
M 3 / BS1501-224 Gr.4
CT DCPD
MEASURED Aa (mm)
Fig. 4.1. Correlation of measured and predicted crack growth; (a) M1 steel (b) M2 Aluminium alloy (c)
M3 steel
For the low toughness material (M2, A1 5083) all errors are negative with only one exception.
Especially for small amounts of crack extension up to about 1 mm nearly 50% of the predicted
values are significantly smaller than the fracture surface measurements done by optical microscopy.
About 30% of the measurements are outside the tolerable error band.
In contrast to this, for the high toughness material (M3, BS 1501) the majority of the errors are
positive; negative errors appear only for crack extensions greater than 1 mm. Nearly 30% of the
measurements are outside the tolerable error band.
Some participants delivered a series of 5 to 10 single specimen tests, covering the crack growth
range from less than 0.5 mm to more than Aumax.Laboratory specific results of unloading
compliance tests for the materials M2 and M3 are plotted in Figs 4.2(a) and 4.2(b), respectively.
For the material M2 both series are underestimating the crack extension by about 0.15 mm for
the CT-specimens of laboratory 9 and by about 0.25 mm for the SENB-specimens of laboratory
6. Post examination of some fracture surfaces of M2 specimens by scanning electron microscopy
(SEM) for SZW measurements (see Section 5 ) indicated that the standard light microscope
measurements as done by the participants tend to overestimate at least small amounts of crack
extension by about 0.15 mm. Though this difference was not checked for larger crack extensions
it is likely that a correction of this order has to be applied to all the tests compared in Fig. 4.2(a).
Then the agreement of predicted and measured crack extensions would be nearly perfect for the
CT-specimens and acceptable for the SENB-specimens.
Fig. 4.2. PUC results, (a) from 2 laboratories from Fig. 4.l(b), and (b) from 3 laboratories from Fig. 4.l(c)
1250 K.-H. SCHWALBE
et al.
For the material M3 the three series in Fig. 4.2(b) differ by less than 0.15 mm. But below 1 mm
of crack extension all predictions are overestimating. Only the CT-specimen tests of laboratory 9
are underestimating beyond 1 mm. At least the first finding seems to be systematic and independent
of the specimen type. It is not known whether this is due to compliance evaluation problems for
large deformations and small crack extensions or to the optical measurement of crack extensions
as found for M2.
These results show that the prediction capabilities of the single specimen methods are sufficient
to meet the requirements of the procedure in the majority of tests.
The quality of PUC-evaluations may depend on the material toughness, but apparent inaccura-
cies may be due to the light optical measurement as applied in the multiple specimen evaluation
as well. The agreement is best for the medium toughness material. There is a tendency to
underestimate crack extension for lower toughness and to overestimate for higher toughness,
0 1.0 20
1W
E "
E
t w
40
20
0 10 20
x
x
200
0 0.5 1.0
/ I I .
1.5
, , , , ,
2.0
CRACK GROWTH Aa mm
Fig. 4.3. Final J-An points of all single specimen tests, power law regressions for these points (solid lines)
and MSU regressions (dashed lines): (a) MI, (b) M2, (c) M3.
respectively, as compared to the light optical measurements. If this trend is true critical J-values
derived from these R -curves may be overestimated for low toughness and underestimated for high
toughness materials as compared to multiple specimen results.
Similar checks for the potential drop evaluations are not possible, because a large number of
tests from some of the laboratories were calibrated using the measured final crack extension of each
individual test.
..
. - .. .
0 05 10 1.5 2.0
CRACK GROWTH Aa rnrn
Fig. 4.4. All data points of all single specimen tests with valid ha prediction error: (a) M1, (b) M2,
(c) M3.
1252 K.-H. SCHWALBE
et al.
4.4 Resulting shape of R-curi1es

Because the number of SENB-specimens is too small both specimen types are discussed together.
Figures 4.3(a) to (c) show the final J-Aa points of all single specimen tests together with the power
law regression curves of the type
J =AAaD
through the valid data points (full line) and the respective regression curves calculated for all
multiple specimen tests discussed in Section 3 (dashed line). As a basis for comparison of individual
R-curves or groups of R-curves these two multiple specimen regression curves for the respective
material are plotted together.
All R-curves were replotted and evaluated using the J-Aa data points returned by the
participants. In the first step R-curves not meeting the Aa-prediction accuracy, i.e. lying outside
the new tolerance bands in Figs 4.l(a) to (c), were excluded. All data points of all tests with a valid
Aa prediction are plotted in Figs 4.4(a) to (c). Only part of these tests having at least one data point
beyond the 3rd validity box and a valid data point distribution resulted in valid power law
regression curves shown in Figs 4.5(a) to (c), respectively.
When comparing single specimen results with the multiple specimen regression curves it should
be considered that
(i) the individual data points of the multiple specimen tests are distributed in scatter bands on
both sides of the regression curves with a width of about:
MI k 25 N/mm
M2 k 5 N/mm
M3 k 60 N/mm
(ii) points outside the regression range should not be compared with the regression curves fitted
to data within the range only
(iii) deviations of single specimen data points from regression curves do not only reflect errors
of the single specimen test method, but in addition contributions from e.g. material scatter
and errors in optical crack length measurement.
Scatter of data points
Especially in the initial region the scatter of single specimen data points is significantly larger
than the widths of scatter bands mentioned above. This is not due to tests with small amounts of
crack extension with the relatively mild validity condition for the Aa prediction. For the three
materials the highest and lowest data points of Figures 4.4(a) to (c) are close to or even identical
with those of fully valid R-curves.
Considering the scatter along the Au (instead of the J ) axis, i.e. approx. k0.1 mm, this is within
the expected experimental error of individual data points of single specimen evaluations; and of
the optical crack length measurements for multiple specimen series. Neglecting the occurrence of
negative Aa -points the resulting R -curves do not contradict the multiple specimen regressions.
Scatter of R-curves
Some of the R-curves showed deviations from the behaviour normally expected. Problems were
caused by, for example:
(i) scatter, crack growth steps forward or backward in the initial part of PUC-tests causing
problems of how to define ai (point of Aa = 0)
0 0.5 1.0 1.5 2.0
0 0.5 10 15 20
CRACK GROWTH ha rnrn
Fig. 4.5. Valid power law regressions for single specimen tests: (a) M1, (b) M2, (c) M3.
(ii) “negative crack growth” in the initial part of PUC tests and DCPD tests causing problems
of how to define Aa = O .
(iii) S-shape of the initial part of R-curves from DCPD-tests
Only part of these curves were invalidated because of unacceptable Aa prediction errors, but some
others remained valid, e.g. PD-curves calibrated with the individual final crack extension, thus not
allowing any accuracy check for an individual specimen.
An additional criterion to assess tests encountering these problems, added in the final procedure
P1-92, requires two specimens to be unloaded in the initial and intermediate range of ductile crack
extension of each series of single specimen tests. Not all of the participants met this condition, as
it was not required by the original specification of the round robin.
1254 K.-H. SCHWALBE
el al.
4.5 Analysis of specijk results

All sets of two or more R-curves with valid Aa predictions from the same participant for the
same material were separately analyzed for trends that might be specific for the material, the
method or its special implementation in the laboratory concerned. This resulted in 8 PUC-series
and 7 PD-series, respectively. Only some examples are presented here, Figs 4.6 to 4.9.
In each plot the two multiple specimen regression lines are added as a reference. Vertical lines
with a length of about 1/10 th of the J-range mark the positions of the final J-Aa points of the
set of specimens, thus showing the respective multiple specimen R-curve. All lines defined in the
procedure are plotted, as are the blunting line, different offset lines and JmaX.
Their positions are
always calculated for a nominal specimen with a, = 30 mm, ( W - a,) = B, = 20 mm, independent
of the actual dimensions of the respective specimens.
4.5.1. Partial unloading method.
Definition of ai.
Most of the tests showed scatter of Aa measurements in the blunting region in the order of
f O . l mm. Though this scatter was expected, it influences the definition of the initial crack length
ai (Au = 0) and of the derived fracture toughness parameters. The three options described in the
procedure [3] cause different problems:
(i) From 1st estimate of crack length.
In many cases the first estimate of crack length is not a good estimate. Though it should be the
average of at least three unloading cycles it may be influenced by scatter. Positive and negative
errors within this scatter range are possible. In addition it may be influenced by crack closure effects
because it should be measured at low load levels.
One participant measured a, from the first unloading cycle in the load range of the last fatigue
cycles. This gave a smaller crack size than unloading cycles in the load range recommended in the
procedure, at higher levels of J, as shown in Fig. 4.6. If such an effect is obtained, use of the second,
or a succeeding point or group of points, according to option (iii) is preferable.
(ii) From the minimum crack length obtained in the test.
This neglects that this point includes scatter and consequently handles the effect of scatter in the
0
05 1.0 1.5
CRACK GROWTH Aa rnm
Fig. 4.6. Results for PUC method on MI-2CT speciment test.

most conservative but normally unrealistic way. For cases of pronounced “negative crack growth”
the improvement of this definition compared to (i) is not quite clear.
A test showing negative crack growth beyond scatter and an invalid under-estimation of the final
crack growth was re-evaluated based on a mechanical model for a possible experimental effect [7].
Figure 4.7 shows the original R-curve (0-symbols) and the final correction (.-symbols) on the
reference curves. The correction of the measured compliance is based on the geometry effects
obtained from the rolling of the loading pins (8= 12.3 mm) in round holes of the specimen
(8= 12.5 mm) and of the clevis. Instead of a flat clevis surface an effective clevis diameter of 13 mm
caused by plastic indentation of the pin was assumed. This result does not prove that the model
and the assumed geometry conditions are correct, but it indicates that the effect of negative crack
growth may cause an underestimation of the final crack growth and consequently an overestimation
of derived fracture toughness parameters. As any correction needs knowledge of the actual effect
causing the negative crack growth problem, the schematical application of this option cannot be
supported.
(iii) From the apparent blunting line.
Defining a, in such a way that the early stages of the crack growth behaviour follow the apparent
blunting line seems to be the best if option (i) does not work.
In fact this option is normally used to “check by eye” whether a, is properly defined. In many
tests the exactly defined option (i) did work when the initial data points scattered around the
calculated blunting line. In other cases the positions of the first data points indicated that a, was
not properly defined by this method and the R-curve might be improved by shifting all data points
along the Aa axis.
Figure 4.8 shows one of the R-curves of Fig. 4.6. The 0-symbols are J-Aa points as
returned by the participant adjusted following option (i). The curve meets the condition of
the procedure requiring “at least five data points. . . within 0.2 mm of u,”.In fact there are ten
points in this region, and they show that a shift by about -0.06 mm might improve the definition
of a,. The valid regression through these points is nearly straight (dashed line with one point),
resulting in low fracture toughness parameters, while the shifted data (.-symbols) and the
respective regression curve (dashed line with two points) agree much better with the reference
curves.
0 0.5 10 15
CRACK GROWTH Aa mm
Fig. 4.7. Effect of negative crack growth (open symbols) and pin rolling correction (closed symbols).
FEMS 16111-E
1256 K.-H. SCHWALBE
et ai.
Fig. 4.8. Effect of shift in Au of -0.06 mm on regression result, (open symbols and curve A-original data,
closed symbols and curve B-shifted data).
It seems possible to demonstrate the sufficiency of quality by five points within 0.2 mm of ai,
but it is obvious that more points will be needed to measure the improvement of quality obtained
by option (iii) or any other more precisely defined correction procedure that may be recommended
in the future for cases with a badly defined ai. Ten points in the blunting region seem to be more
appropriate and should be recommended in order to facilitate subsequent corrections with only
little additional test effort. After more experience with the procedure a proper algorithm to check
and improve the ai definition should be appended.
Selection of regression points
Scatter in Aa in the low Aa-region may cause problems in the selection of data points for the
regression. The example R-curve in Fig. 4.9 is not valid because the final crack extension is not
CRACK GROWTH Aa mm
Fig. 4.9. Effect of point selection in first quartile of validity box on resulting regression.
sufficient. Nevertheless it is selected because the initial part allows several different selections of
regression points all of which could be valid according to the procedure for determination of Jo.z,
JO.ZIBL and even Ji. Here a, was defined by option (i), and this definition is supported by the following
four data points with low scatter.
With increasing J , the positions of the following points seem to be influenced by scatter in Aa.
The program originally used for producing the regression curve selected points No. 11 to 19.
The rule applied was that only points which fell within the validity boxes in the order in which
they were obtained could be included in the analysis. Thus although points 7 , 8 and 9 are themselves
valid, as point 10 is not, they were excluded. The effect of this was tested by calculating a number
of regression curves, each excluding the invalid point 10 but starting from point 7, 8, etc. The
respective regression curves are plotted in Fig. 4.9, identified by the respective starting point. It
is apparent that using starting points which sequentially follow the highest invalid point does not
influence the regression result significantly, while including the other nominally valid points
irrespective of order increases the slope.
The regressions starting from points 9 or 11 are closest to the reference curves. This alone
however, cannot validate the rule on excluding out-of-order points. Nevertheless, a procedure such
as this should be defined in order to better standardise the selection of points and to help distinguish
between scatter and real crack extension.
Material specific interpretation
M1
The R -curves are generally in good agreement with the multiple specimen regressions. Possible
corrections would even diminish the resulting scatter band of regression curves.
M2
The R-curves and the resulting regression curves (Fig. 4.5(b)) are very close to the reference curves,
but there is a trend to overestimate J or, more appropriate, to underestimate Aa in the low Aa range.
The deviations are in the range of the overestimation of crack extension by the optical
measurement, mentioned above (see also Section 5). Possibly the multiple specimen reference curves
are not better at low Aa than the single specimen results.
M3
Many of the R-curves reflected the trend shown in Figs. 4.l(c) and 4.2(b), namely underestimation
of crack extension in the blunting region (including “negative crack growth”), overestimation of crack
extension up to about 1 mm and a change back to underestimation for larger amounts of crack
extension. As a consequence the regression curves tend to be more straight than the reference curves
and deliver lower initiation toughness parameters than the multiple specimen series. The lowest J-Aa
points in Fig. 4.4(c) and the lowest regression curve in Figure 4.5(c) are due to a test not rejected
by the present criteria but showing obvious errors in the definition of ai by option (i) because of
significant scatter. Rejection of this test would diminish the scatter of the regression curves
significantly. Further investigations are necessary to clarify whether this behaviour is typical for
other high toughness materials. Improvement of the compliance evaluation, e.g. by improving the
rotation correction for the large geometry change by plastic crack opening may be needed.
Suggestions for improving the reliability of the partial unloading method
( i ) Measure more J-Aa-points in the blunting region than specified at present (e.g. 10 instead
of 5) in order to estimate the quality and improve the definition of a,.
K.-H. SCHWALBE
et al.
Exclude low load effects like crack closure from the compliance evaluation.
A better measure of the quality of the a,-definition and a procedure for improvement or
rejection is needed.
The option (ii) of the procedure seems not to be adequate to define a, if “negative crack
growth” is found in a test.
The influence of the significant geometry change of specimens on the measured compliance
when testing tough materials needs special consideration, e.g. by an improved rotation
correction.
Do not select all potentially valid J-Aa points in the first validity box. Where there are a
large quantity of data, consider including only one or two points which are in the upper
limit of the data range within the box, and which are below a Au = 0.2 mm offset. Note
that such a selection can only be made if there are a sufficient number of data points
obtained at low crack extensions.
4.5.2 Potential drop methods.
Most of the results were consistent within the respective set of curves of each participant.
Therefore differences should be mainIy caused by differences of the experimental techniques and
of the evaluation procedures.
As for the P.U.C. method, three methods for the definition of ai and the evaluation of Aa from
potential drop measurements are described in the procedure [3]:
(i) at the minimum of the potential; linear calibration between Aa = SZW/Acp = 0 and
Aaend /AVend.
(ii) at an abrupt change in slope of the potential displacement record; linear calibration as in (i)
(iii) from the definition of a reference line by a linear part of the potential vs. load curve in the
range of elastic load; calculation of Au from Acp by a given formula.
Each of the three different methods was used by some of the participants. Some of them used
two different methods for the two different materials tested. The results showed that the selection
of the evaluation method is not a free choice and does not depend only on the method (AC or
DC) or on the position of contacts. Groups of specimens tested in the same laboratory by the same
method with identical contact positions showed different behaviour; the steel specimens needing
0 70 20 30 60
APPLIED FORCE F kN
Fig. 4.10. Example of V , # vs. F curves for M3 ACPD: reference lines for minimum method (O),real
elastic slope dq5/dF ( + ) and apparent elastic slope ( x ) .
I""' ' ' " '1' ' ' ' I 94
0 5 10 15
APPLIED FORCE F kN
Fig. 4.1 1. Example of V , 4 vs. Fcurves for M2 ACPD: real elastic slope dd/dF(dashed line) and apparent
elastic slope (full line).
YI-
5 40-
u -
Lu-
0
u 30-
B
a -
-
W
-1
P 20-
0.
a
/
/
10 - /
/
/
J ' ' I I I I
08
I I I I
12
I I I I
1.6
I I I ~
2.0
I
1260 K.-H. SCHWALBE
el al.
,
I
/ /
CRACK GROWTH Aa mm
Fig. 4.13. J-Aa curve evaluated using reference line of Fig. 4.12 ( x ) , and the curve returned by the
participant (open symbols) for comparison.
The possible improvement by this evaluation procedure is demonstrated for M1 in comparison

to the original evaluation following method (ii). The reference line was calculated by a linear
regression through the 3 “x”-points in Fig. 4.12. The resulting R-curve is shown in Fig. 4.13
together with the curve originally returned by the participant (0)and the two reference regressions.
The examples re-evaluated showed that the method (iii) did not always improve the resulting
R-curves, but it seems to be a good basis for a unified evaluation procedure independent of the
material, the method (AC or D C ) and the contact position. In addition it has a physical background
in separating elastic effects from plastic deformation and crack growth.
Suggestions for improving the reliability of potential drop methods
The quality of the derived R-curves close to initiation is highly dependent on the proper
definition of the reference potential or reference line. This is very important for derived fracture
toughness parameters as well.
The analysis of several test results indicate that a linearly load-dependent reference line similar to
method (iii) may be a basis for a unified evaluation procedure, independent of the mat-
erial, the method (AC or D C ) and of the contact positions. The calibration curve will depend on
the exact positions, but a linear interpolation will suffice at least for small amounts of crack extension.
Based on the differences seen in the slopes of the reference line according to (iii) and of the
unloading cycles, an improved reference line may be found by recording cp vs. F i n some unloading
cycles in the initial elastic load range (above crack closure) even in tests that are not intended for
unloading compliance evaluation.
4.6 Regression curves
For all R-curves with sufficient data points regression lines were calculated. Based on the first
results a change of the definition of the regression ranges and the regression function was proposed
and checked. For the single specimen data the variation of regression functions was limited to an
offset power law and a linear fit in addition to the power law without offset now prefered in the
procedure [3].
The single specimen curve fits were mainly consistent with those from the multiple specimen
series as could be expected from the consistency of single specimen data points with the total
scatter of all multiple specimen data. Figures 4.5(a) to (c) show the valid non-offset power law
regressions.
The scatter band covered by the regression curves is diminished by the new regressions especially
in the low Aa range. Considering that the extremes of the regressions are due to tests that could
be improved by re-evaluations proposed above or tests that should be rejected by improved validity
checks to be specified, the single specimen results are comparable to the multiple specimen
evaluations.
4.7. Fracture toughness parameters
For all valid R-curves critical fracture toughness parameters Jo,2,Jo.z,BL and Ji were calculated.
Despite the great number of specimens included in the evaluation, the number of valid fracture
toughness parameters is too small for a statistically significant assessment of their quality related
to specimen type, experimental method or laboratory. Some general conclusions may be derived.
The broadening of the regression curve range below 0.2mm of crack growth causes a great
scatter of derived fracture toughness parameters, especially for J , , the value of the curve at a small
Aa value in the range of 0.1 mm. This effect was even more pronounced for the offset power law
regressions proposed in the draft procedure, and was one reason for the change to the non-offset
power law in the final procedure. Additional critical checks, not specified in the present procedure,
possibly causing improvements or rejection of the candidate tests, can reduce the scatter band
significantly.
As a consequence of the greater scatter of the single specimen regression curves when compared
to reference curves the scatter ranges of derived fracture toughness parameters are greater as well.
Nevertheless the mean values of the single specimen tests are comparable with the reference results
within one standard deviation.
The deviations between results of different methods are smallest for M1 (medium toughness)
while they tend to be greater for the other materials (with changing signs).
Each fracture toughness parameter derived for a reference multi-specimen series reflects one
average number for the properties of a set of specimens while the scatter band of single specimen
results shows the scatter within the set of specimens. This band reflects scatter (and possibly sys-
tematic errors) caused by the methods but in addition it contains information on the material scatter.
4.8 Conclusions
The preliminary conclusions from the first step of this evaluation gave rise to some proposed
improvements. Some of them are already included in the final procedure after additional more
specific analyses in the second step.
The results presented show that the capabilities of the single specimen methods applied to predict
Aa and to derive R-curves is sufficient to meet the requirements of the procedure in the majority
of tests. To ensure this, specific experimental problems of the single specimen methods (e.g.
“negative crack growth”, calibration for potential drop) contributing to crack growth prediction
errors and badly defined initial parts of R-curves have to be handled carefully. When comparing
single specimen results to multiple specimen series it should be considered that the Aa-prediction
error allowed by the procedure is comparable to the accuracy that can be expected for the optical
fracture surface measurement itself. Thus differences slightly greater than the allowed error do not
necessarily indicate serious deficiencies of the applied indirect methods. The errors seem to depend
on toughness systematically, but possible errors in the light optical measurement as indicated,
especially for the aluminium alloy M2, should be considered as possible sources of error for
multiple specimen series as well.
1262 K.-H. SCHWALBE
et al.
Because of the impact of such problems in the initial part of R-curves on derived fracture
parameters the procedure requires one test unloaded at a small amount of crack extension. This
ensures that the R -curves resulting from single specimen evaluations are comparable with multiple
specimen results and the scatter bands are of similar width, if R-curves with obvious problems in
the initial part are consequently excluded from further evaluation. Improved criteria for acceptance
or rejection should be developed for appending to the procedure after more experience.
The J fracture toughness parameters derived from the single specimen R-curves show more
scatter than the reference multi-specimen results, but the mean values are comparable. To some
extent this is due to the fact that each multi-specimen result reflects an average property of a set
of specimens while the single specimen results show the scatter within this set of specimens.
Some minor changes may be recommended on the basis of the present evaluation:
(i) Measure more J-Aa points in the blunting region than specified at present (e.g. 10 instead
of 5 ) in order to measure the quality and improve the definition of a, (mainly for PUC).
(ii) Exclude low load effects like crack closure from the evaluation (PUC and PD).
(iii) A better measure of quality of the ai definition and a procedure for improvement is needed
(PUC).
(iv) The option (ii) of the procedure for PUC does not seem to be adequate to define ai if
“negative crack growth” occurs.
(v) The influence of the significant geometry change of specimens on the measured compliance
when testing tough materials needs special consideration, e.g. by an improved rotation
correction (PUC).
(vi) Do not select all possibly valid J-Aa points in the first quartile of the validity box: exclude
those which are out of sequence with the first one, in order to reject points that might reflect
only scatter instead of real crack extension (PUC and PD).
(vii) The quality of the derived R-curves for the PD method close to initiation is highly
dependent on the proper definition of the reference potential or reference line. This is very
important for derived fracture toughness parameters as well. The results presented indicate
that a linearly load dependent reference line similar to method (iii) in the procedure, may
be a basis for a unified evaluation procedure, independent of the material, the method (AC
or DC) and of the contact positions (PD).
(viii) Based on the differences seen in the slopes of the reference line according to (iii) in the
PD method and of the unloading cycles, an improved reference line may be found by
recording cp vs. F in some unloading cycles in the initial elastic load range (above crack
closure) even in tests that are not intended for unloading compliance evaluation.
5. BLUNTING, STRETCH ZONE WIDTH, AND INITIATION

(by K-H Schwalbe, K. Baustian, A. Cornec, and M. Homayun)
5.1. Introduction
Although numerous fundamental investigations, e.g. [8-201 have led to reasonable understanding
of the events in the neighbourhood of initiation of stable crack growth, it has turned out to be difficult
to establish a technical definition for a reproducible determination of the fracture resistance at
initiation. The reason for this is that initiation of stable crack growth is not a sudden event, but rather
a gradual transition from crack tip blunting to material separation by ductile tearing.
In the draft version, PI-87D [2] the Procedure provides two different methods for determining
initiation, see Fig. 2.5:
Stretch zone width method: the crack growth resistance curve is intersected with the critical stretch
zone width, GSzw.
Blunting line method: the R-curve is intersected with the analytical blunting line.
In the draft document it was believed that the two ways of determining initiation would give
the same values. However, it was demonstrated in the round robin that this is not the case.
Three elements are needed for determining the fracture resistance at initiation:
Determination of the analytical blunting line.
Measuring GsZw in a scanning electron microscope.
Intersecting the R-curve with either the blunting line or with Gszw.
These were checked with regard to their physical background and to their reproducible determi-
nation.
5.2 Dejinition of task
According to the above mentioned three elements the task consisted of the following items:
1. Check of the blunting line predictions.
2. Analysis of the stretch zone width data returned by the participants.
3. Additional SZW measurements on the fracture surfaces using a scanning electron microscope
(SEM).
4. Determination of the initiation parameters Ji/Si according to Fig. 2.5, using the R-curve fits
provided by the Welding Institute, and comparison of the two definitions of initiation.
5. Re-formulation of the blunting line procedure, if necessary.
6 . Re-formulation of the SZW determination, if necessary.
7. Re-formulation of the J, /hi determination, if necessary.
A particular point of interest was the expected scatter in stretch zone width data.
Items 1 and 4 required tests in the blunting and near initiation phases which were not included
in the round robin. Therefore, additional experiments were carried out, which will be referred to
in the following as “subsidiary tests”.
5.3 Blunting
5.3.1 Determination according to EGF P1-87D [2].
The slope of the blunting line in J terms is determined from either
J
Au=
~ O.4dn-
60
or
J
AaB = d,* -
E
In CTOD terms the blunting line slope is given by either
(5.3)
or
(5.4)
1264 K.-H. SCHWALBE
et al.
Table 5.1. Pre-fatigue parameters maximum load, F,,,,,maximum

stress intensity factor during fatigue, K,, R-ratio and frequency, f.
Material F,,, , kN Kf,N/mm3” R .LHz
MI 10 773 0.1 100
M2 2 154 0.1 50
where d, and d,* are functions of the strain hardening exponent, n, and oolE (ao is the reference
stress); ay is the yield strength.
The factor d , is evaluated from
d,, =
rca,(l
1.185E
+ n ) ( l - v’)
2a0(1+v)(1+n)
1 Itn
(5.5)
5.3.2 Experimental work.

Eight specimens of material MI in the T-L orientation, six specimens of Material M2 in the T-L
orientation, and eleven specimens of material M2 in the L -T orientation were pre-fatigued,
subsequently loaded to varying load levels, and then re-fatigued. All specimens
were 25 mm thick, side-grooved CT specimens. High R-ratios were chosen for re-fatigueing,
because previous experience showed that this is essential for a re-fatigue fracture surface to be safely
distinguished from the fracture surface created by ductile tearing. Tables 5.1 and 5.2 list the
parameters used for pre-fatigueing and re-fatigueing. The loads for pre-cracking had to be very
low to avoid plasticity effects on those specimens which were loaded only in the blunting phase.
SZW and crack growth measurements were done in accordance with the 9 point average method
described in EGF PI -87D [2], Fig. 2.1. The experimental techniques applied were as follows:
Material M 1
Ductile tearing did not initiate in a uniform manner along the crack front. Initiation rather
occurred in the form of localised peaks as shown in Fig. 5.1. For crack growth Aa > 0.1 mm a light
microscope (LM) was used. Heat tinting helped to improve distinction of the boundary between
stable crack growth and re-fatigue fracture surfaces. From all fracture surfaces photographs were
taken along the complete crack front in order to determine crack growth by the area method: the
area covered by stable crack growth divided by the specimen’s net thickness was taken as the
average amount of crack growth. Crack growth GO.1 mm was determined in an SEM.
Material M 2
Oxidation by heat tinting is not a suitable method to determine crack growth in aluminium
specimens. Therefore, high R ratios and low loads during re-fatigueing were used to generate a
fatigue fracture surface with a very fine structure in order to obtain a well marked boundary
between stable crack growth and re-fatigue crack growth. However, a safe distinction in the light
microscope turned out to be very difficult so that all measurements were done in the SEM. At each
Table 5.2. Re-fatigue parameters F,,, in terms of the final

R-ratio, and frequency, f.
load of the static test, FEND.
M1 0.5 FEND 0.6 100

M2 0.4 FEND 0.8 50
Fig. 5. I . Fracture surface of material MI. Dark area = fatigue pre-crack and localised stable crack growth.
Bright area = re-fatigue crack surface.
of the nine measurement locations the local amount of crack growth was determined by the area
method, in order to obtain very accurate results.
The material data needed for determining the blunting line are compiled in Table 5.3. The data
for material M3 are also included because they will be used for the determination of initiation in
Section 5.5.
5.3.3Results
Material M 1
Table 5.4 contains the measured Aa values which refer to either pure blunting or blunting plus
some ductile tearing, and the related J and 6 values. These data represent the very beginning
Table 5.3. Material properties needed for the blunting line

MI M2, T - L M2, L - T M3
uy,MPa 353 171 177 358
0,.MPa 518 32 1 323 560
u,,. MPa 310 145 151 309
E, MPa 207000 70 000 69 231 207 000
V, 0.3 0.3 0.3 0.3
d" 3 0.347 0.254 0.266 0.32
1266 K.-H. SCHWALBE
et al.
of the R-curve, and they are plotted in Figs. 5.2 and 5.3, together with the data points contributed
by GKSS to the round robin, and the R-curve fits through the data of all participants of the round
robin. These fits were forced through the origin, i.e. a non-offset formulation was applied.
It can be seen that the R-curve fits which are based on data Aa > 0.2 mm match perfectly the
data points below 0.2 mm.
The area method, which uses the whole crack front contour for averaging crack growth, seems
to provide a smoother slope of the R-curve than the 9 point method. The blunting line was
evaluated using d,* from Fig. A7.2 of EGF P1-87D [2] (see equations (5.2,5.4)), resulting in the
slope BL, in Figs 5.2 and 5.3, and using equation (5.5), resulting in the slightly higher slope BL,.
Looking at the diagrams, a slight preference may be given to BL,, so that it is recommended to
drop equation (5.5). It is seen:-
( i ) that the blunting line is a very good representation of the SEM data points from the blunting
phase, and
(ii) that the blunting line in its CTOD formulation (Fig. 5.3) has the same degree of accuracy
as on the J level (Fig. 5.2).
It should be noted that this is the first validation of the ESIS CTOD blunting line, whereas the
formulation on the J level has already been verified previously [18-201. It can thus be concluded
that the ESIS blunting line provides a realistic description of the blunting phase.
200
of this specimen
175
150
125
5E 100
7-
75
50
25 f S E M i
I
0 9 point average
0 areamethod
0 GKSS data points from round robin
0 0.10 0.20 0.30
Aa, mm
Fig. 5.2. J-R-curve of material MI close to initiation.
0.275
0.250
0.225
+v
0.2w
0.175
Y I
0.150 , .....
E Fitted by TWI LhrOWh ! hdata of
E participants aceording m W o n 6
L 6 of the EGF-pmcadurn.
0.125
0.100
'!ME; I I
0.075
0.050
0 areamethod
0 GKSS data points from round robin
V Bluntingwntribution to average Aa
0.025
,
I -
t,aw -0.055
I
/
I
Aa. m m
Fig. 5.3. 6-R-curve of material MI close to initiation
Material M 2
Tables 5.5 and 5.6, and Figs. 5.4 and 5.5 contain the relevant data for material M2. When
comparing the data of the round robin with those from the subsidiary tests, it is immediately
obvious that there is a problem: the round robin curve does not match the data points of the
subsidiary tests. Since the generation of these data points in the SEM did not give rise to any
Table 5.4. Exuerimental results for material MI

~
Suecimen Aa, mm J , N/mm 6. mm

M185A 0.609* 215.1 0.328
M185B 0.145* 113.9 0.191
0.170+
M185C 0.0590 87.1 0.128
M185D 0.0320 62.5 0.092
M195A 0.090* 103.9 0.148
0. I08 +
M195B 0.020. 42.5 0.064
M195C 0.0450 74.7 0. I08
M195D 0.152* 122.6 0.172
0. I48 +
* 9 point average using light microscope + area method
0 9 point average using SEM
1268 K.-H. SCHWALBE
et al.
45
6=0.210 Aa
- BLD Curvefit of GKSS in me
40
So.ml=
0.120
35 ._
I
=32.8 /
./ I
I n
30
25 curvefit 01 GKSS in me
E range0 < M S O . 2 6 mm
E
t ................67%...... . . . 6796.
7- E
20 Lc-
I x?
15 JiBuB= 12.5 S = 0.137'Aa

/ J = 44.708 * Aa' a
' Fined by TWI through me data of me
partkipants (EGF-round&in) amrding
paroc~pants(EGF-roundmbm) m r d l n g to Section 6 of me EGF-pocadure.
to W m n 6 of me EGF-procedure
II I
10 0 T-L orientation (subsidiary tests)
0 T-L orientation (subsidiary tests) L-T orientation (subsidiary tests)
L-T orientation (subsiiiary tests) W GKSS data point from round robin
W GKSS data point from round robin 0.02 M-detemination wim LM
M-determination wim LM 0 new aa-delerminatmn with SEM
5
0 new Aa- determination wim SEM
1 j ,G,, = 0.023 I +) first visible dimples on me SEM-

picturesofmisswimen
+) flnt mible dimpies on me SEM-
picturesof m s specimen
I/ I
I 1
I
I I
0 0.1 0.2 0.3 0 0.1 0.2 0.3
&, rnm &a, rnm
Fig. 5.4. J-R-curve of material M2 close to initiation. Fig. 5.5. 6-R-curve of material M2 close to initiation.
Table 5.5. Experimental results for material M2, orientation T-L

Specimen Aa, mm J , N/mm 6,mm
M2124A 0.258 32.7 0.100
M2124B 0.162 25.4 0.079
M2124C 0.014 17.3 0.054
M2 134A 0.081 21.4 0.067
M2134B 0.010 11.3 0.036
M2134C 1.699 61.7 0.173
Table 5.6. Experimental results for material M2, orientation L-T.

Specimen Aa, mm J , N/mm 6,mm
M21240 1.870 88.7 0.247
M21340 1.267 75.7 0.212
M2143A 0.603 53.2 0.154
M2143B 0.013 15.8 0.051
M2143C 0.072 25.2 0.078
M21430 0.144 33.3 0.102
M2144A 0.024 18.0 0.058
M2144B 0.005 6.3 0.020
M2144C 0.010 11.9 0.038
M21440 0.059 23.6 0.074
M2144E 0.109 29.2 0.090
ambiguities, it was suspected that the light microscope measurements carried out in the round robin
were erroneous.
In order to verify this statement, the specimen M2123D, which was investigated by GKSS in
the round robin and which then gave a light microscope value of Aa = 0.19 mm, was again looked
at in the SEM: the resulting value was 67% lower and fell now right into the scatter band of the
specimens from the second evaluation (which were only investigated in the SEM), see Figs. 5.4 and
5.5.
These other specimens were then investigated in a light microscope to reproduce the conditions
of the round robin. The light microscope revealed a dark seam at the beginning of the re-fatigue
fracture surface which had the same colour as the ductile tearing area and which could not be seen
in the SEM. This seam was interpreted as an oxidation product on the initial re-fatigue area. It
is thus concluded, that the optical investigations overestimated the amount of ductile tearing by
the width of the seam described. The SEM showed without any doubt, that the seam was not on
the ductile tearing surface. The crack orientation has no effect on the blunting phase, but ductile
tearing is significantly affected.
The findings and conclusions relevant to the blunting line are the same as those for material M 1.
A further conclusion-which is not directly related to blunting, but to the early stages of ductile
tearing-is that there is the danger of overestimating ductile tearing in aluminium alloys. It is
therefore recommended to measure the early stages of crack growth (Aa < 1 mm) in an SEM.
5.4. Stretch zone width

5.4.I . Determination according to EGF PI-87D [2].
Due to the variation of the criticaktretch zone width, Aa,,,, from specimen to specimen, the
required critical stretch zone width, AasZw,has to be determined as the average of at least three
specimens:
(5.6)
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
r
I l l l l l l l l
0.01 B
(No
machined notch
initial crack front
find crack front
Fig. 5.6. Measurement of crack lengths on single edge notch bend specimens: a) plain sided specimens
b) sidegrooved specimens.
1270 K.-H. SCHWALBE
et al.
r----l
fatigue preuac&
-stretch zone
--stable crack growth
Beginning of stretch zone
i=l i=k
to fatigue surface
Fig. 5.7. Determination of the critical stretch zone width.
Furthermore, the stretch zone formation along the crack front exhibits substantial scatter;
therefore, the following averaging procedure is required:
A local Aaszw,Lis determined at the nine positions used for measuring the crack length, Fig. 5.6:
1 9
=-
A ~ z w C Aaszw,L,i
9,,,
(5.7)
Each of these nine local values has to be determined by averaging at least five individual
measurements, Fig. 5.7.
5.4.2 Work performed.
The SZW data returned by the participants were analysed with respect to their scatter. As a first
step, a compilation of the data and the illustration of their variability were performed, Section 5.4.4.
Detailed examination of the marked photos returned by the participants was then carried out to
reveal the sources of scatter. In order to obtain information on a possible non-symmetry of SZW
development, the two halves of one specimen of each material were examined. Using the experience
Table 5.7. Names of participants involved in the stretch zone width investigation
~ ~~~~~
Prof. Dr. Ing. A. Bakker, Technische Hogeschool Delft; Dr. E. Roos Staatl. Material-
priifungsanstalt Stuttgart; Dr. B. K. Neale CEGB, Berkeley Nuclear Laboratories,
Berkeley; Dr. A. Cornec, GKSS-Forschungszentrum Geesthacht GmbH Geesthacht;
Dr. R. L. Jones, A R E , Poole; Dr. S . C. Druce, Harwell Laboratory, UKAEA, Harwell,
Dr. K. Wobst, BAM, Berlin; Dr. Ing. J. M. Amo, CENIM, Madrid, E. Morland, Risky
Nuclear Laboratories, United Kingdom Atomic Energy Authority, Risky
Table 5.8. Average GsZw

(in mm) and the standard deviation u (in mm) of each participant’s Gszw,
MI M2 M3
BS
-
4360 Grade 50E - 5083-0 BS-1501-224-490 B LT.50
Laboratory Specimen Aa,,,, U AffP7W I7 Aac7wC U
SENB 0.0893 0.0 160 0.1247 0.0 I90

CT 0.0869 0.0120 - -
CT - - 0.1269 0.0023
SENBjCT 0.0602 0.0097 - -
SENB - - 0.1318 0.03 I7
SENB 0.0685 0.0291 0.0763 0.0390
CT 0.0720 0.0049 0.1030 0.0183
SENB - - 0.1024 0.0210
CT 0.1067 0.0150
All laboratories 0.0754 0.0123 0.0310 0.0028 0.1 103 0.0193
gained during the analysis of SZW data, a “final” bas,, for each material was derived. This work
will be described in Section 5.4.5.
The experience gained in the foregoing item was supposed to give some hints on generating more
consistent SZW data, see Section 5.6.
5.4.3 Round robin results.
The laboratories participating in the round robin did the testing using single or multiple specimen
methods according to their facilities. Nine of these laboratories (see Table 5.7) took SEM pictures
of the broken specimens and provided the respective Aaszw and baszwi values. All three materials
were evaluated this way. Table 5.8 shows the Gszwi values of the individual laboratories. From
these aszwi values, i.e. those values which were returned by the participants as their ~,,,,
the
average ZsZw was calculated which is supposed to be a characteristic value of the material under
consideration, see Table 5.8. Fig. 5.8 shows representative pictures of the stretch zone of the three
materials investigated.
5.4.4 Analysis of the SZW data.

The analysis was performed in the following four actions:
Action 1: Analysis of the original data provided by the participants.
Action 2: The micrographs returned by the participants were re-evaluated with subsequent
analysis.
Action 3: The broken specimens provided by the participants were used for independent stretch
zone width measurements by GKSS.
Action 4: Some of the measurements of Action 3 were done on identical photographs by seven
GKSS staff members. This mini round robin was supposed to provide additional insight
into the problem of individual interpretation of SZW pictures.
For each individual action standard deviations, 6,from average values, as,,,were evaluated by
(5.8)
with N being the number of Gszwi values used for the analysis. In addition the standard deviations
of the AasZwivalues of each participant from his own average, daszwi were determined the same
way.
Fig. 5.8. Typical stretch zone formation at crack tips for steels ( M l , M3) and aluminium alloy (M2). The
micron markers shown on each photograph are 100 microns in length.
1372
Round Robin Data Final Values

0.14
J
0.12-
R
0.10-
0.08 --
0.06.
0.04 -
0.02 -
I I
1 0.01 J
3 5
1 Participant
-251
-50 J
3 5
m
Laboratory numbel
Fig. 5.10. Round robin results of M2 data in comparison

with the final value, together with the SEW scatter for each
.
- 0 -
participant in relation to Aa,,, .
5
I2
1 3 4 5 7 8
Laboratory number
Fig. 5.9. Round robin results of MI data in comparison with

the final value, together with the SZW_ scatter for each
participant in relation to Aa,,, .
Figs. 5.9(a), 5.10(a) and 5.1 l(a) show the individual Aaszwimeasurements (each value represent-
ing one specimen) in relation to the total average, zsZw,
for the three materials, as extracted in
Action 1. The diagrams also show the standard deviations obtained by each participant as well
as the standard deviations of the baszwivalues from the total average, by the error bars. zsZw,
The relative deviation -
- -- Aaszw * 100 (in percent)
Aas~wi
Aaszw
1274 K.-H. SCHWALBE
et al.
0 40 Round Robin D m Final valuer
0.35
0.30
2 0.25
$. 0.20
2 0 I5
0.10
0.05
Participant
One specimen from Lab. 6

P with + 230 %is out of the range. '
$
-2
-501
-100
,
CT
;
,[y(, SEN8
@I
1 2 2 4 6 7 8
Laboratory number
Fig. 5.11. Round robin results of M3 data in comparison with the f i g 1 value, together with the SZW
scatter for each participant in relation to Aa,,, .
of the measurements of each specimen from zszw

is illustrated by Figs. 5.9(b), 5.10(b),
and 5.11(b).
Since only two participants generated values for material M2 it does not make much sense
to derive any scatter statements. However, for the sake of comparison, the measured values
and the derived standard deviations are included in the Figures and Tables. As far as the two
steels are concerned the extreme deviations of individual Aaszwi values from the total
average amount to about 45% for material M1 and up to roughly 60% for M3, the latter with
one outlyer at +230%, Fig. 5.11b. The largest standard deviation of 17.5% was found for M3.
Table 5.9. zszw,

values (in mm) as determined in Actions 1. 2 and 3.
- MI - M3
Laboratory Specimen Action Aa,,,, U Aaszw, U
I 0.0893 0.0160 0.I247 0.0190

1 SENB 2 0.0657 0.0020 0.1005 0.0130
3 0.0743 0.0050 0.0848 0.01 57
I - - 0.1269 0.0023
4 CT 2 - - 0.0730 0.0050
3 - - 0.0850 0.0050
1 - - 0.1318 0.0317
6 SENB 2 - - 0.0933 0.0150
3 - - 0.0921 0.0090
1 0.0685 0.02 10 0.0763 0.0390
7 SENB 2 0.0594 0.0080 0.0501 0.01 10
3 0.0796 0.0050 0.0897 0.0040
1 - - 0.1067 0.0150
2 CT 2 - - 0.0907 0.0130
3 - - - -
1 - 0.1024 0.02 10
-
7 SENB 2 -
-
-
- 0.077 0.0090
3 - - -
I: Action I : Contributed b y participants.

2: Acrion 2: Evaluated by GKSS, micrographs from participants.
3: Action 3: Evaluated b y GKSS, micrographs from GKSS.
0.15
0.20
Laboratory 1 Laboratory6 *)
0
I V
V
8
.0--
V
B
2
100
0.05
O l
b Gszw
=0.120
,100 ’ 1 2
Action
+
3
Aam
1
centre line
Fig. 5.12. Aa,,, measurements in Actions 1, 2 and 3

2
Action
3
1 bl
Action
*) Two specimens of Action 1 with

**I
iism
Action
-
One specimen of Action 1 with +230% is out of the range.
Fig. 5.13. Aa,, measurements in Actions 1, 2, and 3

Action
0.216 and 0.396 are out of the range.
(material MI) and the scatterof each specimen in relation (material M3; laboratories I, 4 and 6) a@ the scatter of each
to Aaszw. specimen in relation to Aa,, .
1276 K.-H. SCHWALBE
et al.
Table 5.10. Bas, measurements in mm on the same micrographs by 7 GKSS staff members.
Specimen MI44 MI54 M2113B M2113D M3-4 M3-3
Evaluator
A 0.0538 0.0599 0.0286 0.0286 0.0675 0.0815
B - 0.0527 - 0.0206 - 0.0693
C - 0.0747 - 0.0250 - -
D - 0.0642 - 0.0229 - 0.0607
E 0.0690 - 0.0332 - 0.0871 -
F 0.0923 0.1044 0.0293 0.0392 0.0856 0.1190
G 0.0768 - 0.0352 - 0.0858 -
-
All k z w 0.0730 0.0712 0.0315 0.0273 0.0815 0.0257
Evaluators
U 0.0160 0.021 2 0.0032 0.0073 0.0094 0.0257
Fig. 5.11 shows that the standard deviation of Au,,,~, i.e. the specimen to specimen variation,
of a participant may be larger than the standard deviation of dasZwi. The relative deviation
of da,,,, provided by the individual participants from the total average, Zszw, is less than
f25%.
The total averages, z,,,,
for materials MI and M2, as determined in Actions 2 and 3, are listed
in Table 5.9 together with the values from Action 1. The table shows that in five out of six cases
the standard deviations of zsZwi
from Actions 2 and 3 are smaller than those from Action 1. This
may be the consequence of the learning process resulting from analysing many more cases than
each individual participant had the opportunity to look at. Consequently, this kind of training led
V
_-
0 $ 0
0
3 V
B
0
0.05 0
OJ
1w
l v o I
IN131
a)
100
8
0-
0
50 -
B
V .
I$
-
B O -
I$
?
0 2 -50-
V 0
.looJ I )
1 2 3 1 2 3 -100
A B C D E F G
Action Action
differentevaluators
Fig. 5.14. Aaszw measurements in Actions 1, 2 and 3
(material M3; laboratories 2 and 7) andthe scatter of each Fig. 5.15. Scatter of the GKSS staff member measurements
specimen in relation to Aa,,, . on all three materials in relation to &is,, .
the evaluators from GKSS to more consistent definitions of beginning and end of stretch zone.
Furthermore, some of the participants stated little experience in SZW determination prior to the
round robin.
The scatter of the Aaszwivalues supports the observation from Table 5.9, Figs 5.12 and 5.13:
the specimen-to-specimen variation in Action 1 is in all cases greater than that of Actions 2 and 3.
Action 4 was then undertaken in order to obtain the variability within a control group from one
laboratory. Seven GKSS staff members measured Aaszw on identical photographs. Each of the
three materials was represented by 18 photographs, hence 54 photographs were evaluated.
Table 5.10 lists the results obtained from the individual specimens (two specimens per material).
Again, as in Actions 2 and 3, the average stretch zone widths are smaller than those from Action 1,
see Table 5.8. A graphical representation is given by Fig. 5.15, which shows the relative deviation
of each GKSS staff member from da,,, as of Action 1 for all three materials.
Table 5.1 1 compares the average stretch zone widths of all four actions and presents also values
averaged over all four actions.
If, however, a comparison is made only among the evaluators of Action 4, i.e. if each evaluator’s
values are compared with the average, zszw*,
obtained exclusively from Action 4, the result shown
in Fig. 5.16 emerges. It can be clearly seen that six out of seven evaluators agreed very closely in
their interpretation of the stretch zone. But due to evaluator F, the scatter amounts up to about
50%, which is not much different from the result of Action 1.
5.4.5 Origin of scatter.
Analysis of photos
Considerable effort was spent on examining the marked photos returned by the participants in
order to reveal the sources of scatter. Two kinds of problems could thus be identified: technical
faults and individual interpretations. Due to the limited space, they can only be briefly summarised
here. For details the reader is referred to the final BCR report [4].
Technical faults
In some cases dirt on the fracture surface obscured the details of interest.
Some photos were improperly exposed or
loo I
50 I
1
a
a
Table 5.11. Results of the 4 Actions,

Action M1 M2
zsZw
in mm
M3
-46 -50
- 100
1, A
~
B C
.b,A
D
~ camre
I line
E
~
F G
I 0.0754 0.0310 0.1103
2 0.0626 - 0.0808 different evaluators
3 0.0770 - 0.0879 Fig. 5.16. Analysis of scatter of the GKSS staff member
4 0.0720 0.0294 o.0820 measurements (Action 4) in g a t i o n to newly calculated
All actions 0.0718 0.0302 0.0903 average Aa&, .
1278 K.-H. SCHWALBE
et al.
0.1 mm
Fig. 5.17. Fracture surface photo with SZW-marking of specimen M337 made by participant 6.
improperly focused, so that it was hard to distinguish details of interest.

Magnification was sometimes too high, so that beginning and end of the SZW was not on one
photo.
Table 5.12. Determination of final SZW for material M I , in mm

Specimen Measurement position
No. Participant 1 7 3 4 5 6 7 8 9 Average
M185A 0.060 0.043 0.073 0.051 0.043 0.039 0.065 0.077 0.066 0.057
0.058 0.057 0.064 0.055 0.061 0.070 0.074 0.067 0.047 0.061
M144 5 0.054 0.048 0.050 0.046 0.044 0.042 0.052 0.047 0.043 0.047
M147 1 0.052 0.046 0.065 0.066 0.066 0.062 0.060 0.058 0.055 0.059
M191 7 0.042 0.042 0.081 0.064 0.062 0.044 0.051 0.045 0.057 0.054
-~
An,,, = 0.055
Standard deviation = 0.006
Table 5.13. Determination of final SZW for material M2, in mm

M2134C 0.026 0.019 0.022 0.018 0.025 0.018 0.026 0.025 0.026 0.023
0.032 0.021 0.026 0.018 0.018 0.023 0.022 0.023 0.020 0.023
M2113B 5 0.027 0.024 0.029 0.020 0.020 0.020 0.025 0.016 0.025 0.023
M2113C 5 0.031 0.016 0.021 0.029 0.022 0.013 0.018 0.026 0.026 0.022
M2113D 5 0.016 0.020 0.033 0.019 0.017 0.030 0.029 0.018 0.016 0.022
-
Aaszw = 0.023
Interpretation
In many case details such as dimples and broken particles which belong clearly to ductile tearing,
were included in SZW measurements. Fig. 5.17 shows an example. But also the interpretation
already shown in Fig. 5.8 yields SZW values too high. These items were carefully avoided in the
determination of the final SZW values, which thus led to substantially lower SZW values, see below.
Final stretch zone width oalues

One specimen of each material was examined on both halves. The magnification factor was
adjusted to 400, enabling an accurate SZW determination. In order to perform the measurements
as accurately as possible, the SZW area on the SEM photos was determined with a planimeter
so that the area method could be applied. The resulting values are listed in Tables 5.12 to
5.14. Additional specimens returned by the round robin participants were analysed the same
way (one specimen half only), the results are also given in Tables 5.12 to 5.14. Graphical
representations are given in Figs 5.9, 5.10 and 5.11, so that round robin and final values can be
easily compared.
Due to carefully avoiding non-SZW features in the measurements, the final da,,, values are
substantially smaller than the average over the values returned during the round robin. Since one
specimen of each material provided two values, it is believed that the analysis of four to six
specimens is sufficient. This view is supported by the low scatter, expressed in the percentage
deviation of the extreme values and in the standard deviation.
It is seen from the tables that the difference in the measurements of two specimen halves is in
all three cases smaller than the width of the scatter band of all specimens. From this it is concluded
that the examination of one specimen half is sufficient. This view is supported by a previous round
robin of the ESIS Task Group on Micromechanisms, which also came to the conclusion that
the SZW development is symmetrical. It is also seen in Figs 5.9, 5.10 and 5.11 that the
Table 5.14. Determination of final SZW for material M3, in mm

M365A 0.029 0.074 0.058 0.107 0.101 0.089 0.090 0.077 0.046 0.075
0.064 0.097 0.104 0.069 0.097 0.094 0.104 0.068 0.057 0.084
M315A 4 0.074 0.040 0.058 0.087 0.066 0.077 0.082 0.083 0.066 0.070
M345B 2 0.034 0.058 0.052 0.062 0.069 0.078 0.047 0.067 0.033 0.056
M345C 2 0.057 0.077 0.071 0.064 0.059 0.100 0.063 0.065 0.065 0.069
M334 7 0.038 0.054 0.073 0.060 0.070 0.068 0.070 0.065 0.068 0.063
M347 6 0.024 0.055 0.061 0.086 0.053 0.058 0.057 0.055 0.056 0.056
-
Aaszw = 0.066
1280 K.-H. SCHWALBE
ef at.
individual final values are well within *25% which is supposed to describe an acceptable scatter
band.
The term “final” SZW means that these values are the result of just the laboratory analysing
the SZW part of the round robin. There is still enough room for individual interpretation, so that
other laboratories may come to different conclusions.
5.5 Initiation
The Aa measurements generated for checking the theoretical blunting line served also for
validating the initiation procedure.
Figures 5.2. to 5.5 contain all the information required for determining initiation: in addition
to the crack growth data and the blunting line the critical SZW is shown. These are the final values
as shown in Tables 5.12 and 5.13. The diagrams-in particular those for material M1-show clearly
the gradual transition from blunting to ductile tearing. It is also seen, that the blunting process
continues to follow the theoretical blunting line after localised initiation, up to the point where the
critical SZW, Zszw, has been reached.
In addition to the two definitions of initiation outlined in Fig. 2.5, a third definition may be
formulated: it is the definition of initiation used by JSME SOO1-1981 [21], see also Ref. [I], which
intersects the blunting line with daszw.For better distinction of the two SZW methods, two
different subscripts were used in Figs. 5.2 to 5.5: JiSZWIBLand JlsZwiR.Fig. 5.18 gives a schematical
representation of these three definitions.
The constructions show clearly that the intersection of zszw with the R-curve provides the best
description of the initiation process, and it is therefore recommended to use only this method.
5.6 Conclusions
Careful re-analysis of round robin data and additional experiments in the initial phase of crack
growth including comprehensive SEM examinations have confirmed that the ESIS Procedure is
based on very reasonable assumptions concerning the physical events and their modelling. As a
result of the work described in this report, the following recommendations are derived in order to
improve the accuracy and reproducibility of the determination of initiation parameters.
Blunting line
A theoretical blunting line is needed for defining the region of valid crack growth data and for
determining J0.21BL/80.2,BL.
The blunting process is very well described by the theoretical blunting
line. It is recommended to drop the formula in A7.4 of Pl-87D [2] (equation (5.5) of this report)
to avoid ambiguities.
Determination of the critical stretch zone width

The authors of this contribution must admit that the experience gained during the analysis of
the round robin has taught them to define the stretch zone with more reliability and accuracy than
they did previously, in other words, there has been a considerable learning effect. It was seen that
careful experimentation results in lower values of the SZW and smaller scatter than experienced
in the round robin.
Most of the recommendations derived from this study emerge straight from the criticised items
listed in Section 5.4.5:
The 9 point averaging method with five individual measurements at each of the nine locations
is adequate.
It is sufficient to examine one specimen half for determining the SZW.
Technical faults may unnecessarily lead to erroneous results. Thus
clean the fracture surface before examining the specimen in the SEM,
0 adjust the SEM magnification such that the SZW covers most of the photo and that its
beginning and end are visible on one photo,
0 use vertical view on the fracture surface,
focus SEM properly,
expose photos properly.
It is recommended to gain experience in interpreting SEM photos and fracture features before
stating specific SZW values.
The main problem seems to be the definition of the beginning and in particular the end of the
stretch zone. When fixing the transition line between the stretch zone and the tearing surface,
dimples, broken second phase particles, and similar features should be carefully identified and
left aside. This serves to make sure that only stretching or blunting contributes to what is
measured as the stretch zone width. Examples were shown in Fig. 5.8.
Whenever possible, more than one person should be asked to carry out the SZW determination.
Examining more specimens may help to balance lack of experience.
Meemre u a c k m at F&%:W 1-9
1 2 3 4 5 6 7 8 9 hontfaCeOf
i I
-
fatigue crack
front lnnial
uadr front
Anal crack
hOnt
I
I
-
Aa sm 0.2 rnm
Aa
Fig. 6.1. Measurement of crack growth on a deformed
Fig. 5.18. Schematic of three definitions of initiation. specimen.
1282 K.-H. SCHWALBE
et al.
Determination of initiation
The R-curve fit performed for data with Aa 2 0.2 mm is adequate for describing near-initiation
data.
A light microscope for measuring low Aa values may not be adequate, as it may give rise to
mis-interpretation of the features on the fracture surface which in turn results in erroneous Au
measurements. This was found for the aluminium alloy, so that this kind of material requires the
use of an SEM measuring small amounts of crack growth (Aa < 1 mm).
Of two alternatives for determining Ji/Si, the stretch zone width method-which intersects Zszw
with the R-curve-was found to correspond well with the SEM observations. It is therefore
recommended to use this method as the only method in the ESIS procedure.
6. CONCLUSIONS FOR THE FINAL PROCEDURE ESIS P1-92

The extensive experimental round robin demonstrated that-gross0 modo-the draft ESIS
Procedure provides a very useful method for determining elastic-plastic fracture parameters and
that the starting idea of incorporating both the J-integral and CTOD, and both initiation and the
variation of fracture resistance with crack growth into a single method was successfully realised.
However, the validation by the round robin revealed weaknesses which were removed in the final
version PI-92 [3].
The conclusions drawn from the round robin have been listed in the preceding sections and
detailed in the three Addenda to the BCR Report [4]. The most important ones will be mentioned
in the following sections in the context of the items revised for P1-92.
6.1 Measurement of crack growth on fracture surface
The measurement positions on the fracture surface are now related to the deformed specimen,
Fig. 6.1.
6.2 Blunting line
A theoretical blunting line is needed for defining the region of valid crack growth data and for
determining &jBL/80,2/BL. The blunting line formula in PI-87D [2] has been dropped.
6.3 Data spacing
Four equal crack growth regions are defined, as shown in Fig. 6.2. Use of the line parallel to
the theoretical blunting line at an offset of 0.1 mm, ensures that only data exhibiting ductile tearing
J.6 BlunUng Uno
&nu
Aa
OlmlLJ
Fig. 6.2. Data spacing and curve fit. Fig. 6.3. Single specimen crack growth termination require-
ment.
beyond blunting are used. At least one point is required in each of these regions. The crack growth
limits, Aamax,shown in Fig. 2.2 have not been altered.
6.4 Curue fitting

The data points validated by Fig. 6.2 are then fitted by equations (2.1) and (2.2). However, the
second evaluation recommended removal of the offset in these equations as the ordinary power
law fits the early crack growth data more appropriately. It is recommended that this is taken into
consideration in a future revision. The J,,,/S,,, limits are then imposed, and the curve enclosed
by the J,,,/S,, and the Aamaxlimits is regarded as a material property independent of specimen
size.
6.5 Fracture parameters

The fracture parameters described in P1-87D [2] have been retained, with the following
alterations:
(i) The blunting line method for determining J,/Si has been dropped.
(ii) The determination of Ji/Si based on the measured SZW has been shifted into an appendix,
. because it was felt that a method which requires the use of an SEM should not be part of the
main procedure which deals with a mechanical test method.
6.6 Single specimen methods

It is recognised that there is a fundamental difference between multiple and single specimen
methods. The former gives an average of the crack growth resistance, whereas the latter gives
individual results which may serve for verifying material inhomogeneity.
At least three specimens are needed for a first R-curve determined in a series of tests using a
single specimen method, two of which are needed to demonstrate the accuracy of the experimental
technique at smaller amounts of crack growth.
One test should be terminated near initiation, between 0.1 and 0.3 mm of ductile crack growth,
Fig. 6.3. A second test should be terminated midway between the valid crack growth range, Aamax,
Fig. 6.3. If the difference between predicted and measured crack growth exceeds 15% of the
measured value or 0.15 mm, whichever is the greater, then the test is invalid and the single specimen
technique requires improvement.
The data points from valid tests used to demonstrate the accuracy of the test technique should
be combined to generate a single R-curve.
Acknowledgements-The authors are indebted to the participants of the round robin who provided them with very valuable
results, serving as the pre-requisite of this work. The financial support of the Bureau Community of Reference is gratefully
acknowledged.
7. REFERENCES
K.-H. Schwalbe, B. K. Neale, and T. Ingham (1986) The development of a European Group on Fracture
procedure to measure elastic-plastic fracture parameters, in Fracture Control of Engineering Structures,
ECF6, ed. H. C . van Elst and A. Bakker; vol. 1, pp. 277-299.
K.-H. Schwalbe, B. K. Neale, and T. Ingham (1988) Draft EGF recommendations for determining the
fracture resistance of ductile materials: EGF Procedure EGF P1-87D, in Fatigue Fract. Eng. Mater. Strucr.
11, 409-420.
ESIS P1-92 (1992) ESIS Recommendations for determining the fracture resistance of ductile materials,
January 1992, available from ESIS office, University of Technology, Delft, PO Box 5025, 2600 GA Delft,
The Netherlands.
1284 K.-H. SCHWALBE
et al.
4. B. Hayes, N. V. Challenger, J. R. Gordon, K. Baustian, A. Cornec, M. Homayun, B. Voss, K.-H.

Schwalbe (1990) Intercomparison of fracture toughness measurements of ductile materials; final report
of a European round robin supported by Community Bureau of Reference (BCR) July, 1990.
5. K. Wallin (1987) Final report of nordic J-R-curve Round Robin Test Programme, Research Report 510,
Valtion Teknillinen Tutkimuskeskus, Espoo, Finland.
6. DVM-Merkblatt 002 (1987) Ermittlung von RiDinitiierungswerten und RiDwiderstandskurven bei An-
wendung des J-Integrals; Deutscher Verband fur Materialprufung e.V., 1000 Berlin 45.
7. B. Voss and R. A. Mayville (1985) The use of the partial unloading compliance method for the
determination of J-R-curves and J,, in Elastic PIastic Fracture Test Methods: The User's Experience,
ASTM STP 856, pp. 117-130 ed E. T. Wessel and F. J. Loss.
8. W. A. Spitzig (1986) A fractographic feature of plane strain fracture in 0.45 C-Ni-Cr-Mo steel, Trans.
ASM 61, 344.
9. I. Kirman (1971)The relation between microstructure and toughness in 7075 aluminium alloy. Met. Trans.
2, 1761.
10. J. A. Psioda and J. R. Low (1975) The effect of microstructure and strength on the fracture toughness
of an 18Ni, 300 Grade maraging steel, NASA Technical Report No. 7, NASA Grant NGR 39-087-003.
11. K.-H. Schwalbe (1974) Crack propagation in AlZnMgCuO.5 during static loading. Engineering Fracture
Mechanic 6, 415434.
12. R. M. McMeeking (1977) Finite Deformation Analysis of Crack Tip Opening in Elastic-Plastic Materials
and imlications for Fracture. J. Mech. Solids 25, 357-381.
13. C. F. Shih (1979) Relationship between the J-integral and crack opening displacement for stationary and
extending cracks. General Electric, Report No. 79CR0075.
14. W. J. Mills (1981) on the relationship between stretch zone formation and the J-integral for high strain
hardening materials. Journal of Testing and Eualuation 9(l), 56-62.
15. D. M. O'Brien and W. G . Ferguson (1982) On the J-integral blunting line for soft materials. Int. Journal
of Fracture 20, R 39.
16. S. W. Yin, R. A. Gerbrands, and M. Hartevelt (1983) An investigation of the blunting line. Engineering
Fracture Mechanics 18(5), 1025-1036.
17. K. Ohji, A. Otsuka, and H. Kobayashi (1983) Evaluation of several testing procedures recommended in
Japan. ASTM STP 803, C. F. Shih and J. P. Gudas, pp. 11-398-11-419.
18. K.-H. Schwalbe (1984) Crack tip opening displacement for work-hardening materials. International
Journal of Fracture 25, R49-R52.
19. A. Cornec, J. Heerens, and K.-H. Schwalbe (1986) Bestimmung der RiBspitzenaufweitung CTOD und
RiDabstumpfung SZW aus dem J-Integral, GKSS-Report 86/E/15, GKSS-Forschungszentrum
Geesthacht, 2054 Geesthacht, FRG.
20. J. Heerens, A. Cornec, and K-H. Schwalbe (1988) Results of a round robin on stretch zone width
determination. Fatigue Fract. Engng. Muter. Struct. 11, 19-29.
21. H. Kobayashi, H. Nakamura, and H. Nakazawa (1985) Comparison of J,, test methods recommended
by ASTM and JSME, in: ASTM STP 856, E. T. Wessel and F. Loss, Eds., American Society for Testing
and Materials, pp. 3-22.

Validation of The Fracture Mechanics Test METHOD EGF P1-87D (ESIS P1-901ESIS P1-92)

Uploaded by

Document Information

Original Title

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

Validation of The Fracture Mechanics Test METHOD EGF P1-87D (ESIS P1-901ESIS P1-92)

Uploaded by

Copyright:

Available Formats

Fatigue Fracl. Engng Maler. Struct. Val. 16, No. 11, pp. 1231-1284, 1993 8756-758X193 $6.00 + 0.

VALIDATION OF THE FRACTURE MECHANICS TEST

K.-H. SCHWALBE,'B. HA YES,^ K. BAUSTIAN,'

(Received in Jinal form 15 July 1993)

4.6 Regression curves

5 BLUNTING, STRETCH ZONE WIDTH, AND INITIATION

6 CONCLUSIONS FOR THE FINAL PROCEDURE-ESIS P1-92

Jo,2= J at 0.2 mm crack growth including blunting

EGF P1-87D: EGF Recommendations for Determining the Fracture Resistance of

2. OUTLINE OF ROUND ROBIN (by B. Hayes)

2.1 Draft Procedure PI-87D

Measure crack g r M at pcsitions 1 - 9

stable crack growth

Fig. 2.1. Measurement of crack growth.

J = A(Aa + C)" (2.1)

Vaiidw box for crack i l l

corresponding to the fracture resistance at 0.2 mm of total crack growth

R-curve fit equation.

2.2 Round robin

Admiralty Research Establishment, Holton Heath, UK.

M3, High Toughness MI, Medium Toughness M2, Low Toughness

High Medium Low

High toughness 0.320 309 N/mm2

3. MULTIPLE SPECIMEN METHOD (by B. Hayes and R. Gordon)

3.I First evaluation

'E 1.0- 0.8 c

0 0 0.2 0.4 0.6 0.8 1.0 1.2 14 1.8 1.1) 2.0

3.2 Conclusions from the first evaluation

3.3 Second eoaluation

high toughness, M3 J,,, = 367 N/mm

low toughness, M2 J,,, = 197 N/mm

NUMBER OF POINTS IN SIMULATED DATA SETS

3.4 Conclusionsfrom the second evaluation

4. SINGLE SPECIMEN METHODS (by B. Voss)

CRACK GROWTH Aa rnrn

4.4 Resulting shape of R-curi1es

0 0.5 1.0 1.5 2.0

CRACK GROWTH ha rnrn

4.5 Analysis of specijk results

CRACK GROWTH Aa rnm

Fig. 4.6. Results for PUC method on MI-2CT speciment test.

I""' ' ' " '1' ' ' ' I 94

The possible improvement by this evaluation procedure is demonstrated for M1 in comparison

5. BLUNTING, STRETCH ZONE WIDTH, AND INITIATION

Table 5.1. Pre-fatigue parameters maximum load, F,,,,,maximum

5.3.2 Experimental work.

Table 5.2. Re-fatigue parameters F,,, in terms of the final

M1 0.5 FEND 0.6 100

Table 5.3. Material properties needed for the blunting line

0 0.10 0.20 0.30

Table 5.4. Exuerimental results for material MI

Suecimen Aa, mm J , N/mm 6. mm

15 JiBuB= 12.5 S = 0.137'Aa

1 j ,G,, = 0.023 I +) first visible dimples on me SEM-

Table 5.5. Experimental results for material M2, orientation T-L

Table 5.6. Experimental results for material M2, orientation L-T.

5.4. Stretch zone width

initial crack front

find crack front

Beginning of stretch zone

Fig. 5.7. Determination of the critical stretch zone width.

Table 5.8. Average GsZw

SENB 0.0893 0.0 160 0.1247 0.0 I90