Charpy Impact Test - STP 1072

STP 1072
Charpy Impact Test:

Factors and Variables
John M. Holt, editor
ASTM
1916 Race Street
Philadelphia, PA 19103
Copyright by ASTM Int'l (all rights reserved); Tue Dec 15 12:59:52 EST 2015
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
ASTM Publication Code Number (PCN): 04-010720-23
ISBN: 0-8031-1295-5
Library of Congress No: 90-085687
Copyright 9 1990 by the American Society for Testing and Materials. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the publisher
NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication
Peer Review Policy
Each paper published in this volume was evaluated by three peer reviewers. The authors
addressed all of the reviewers' comments to the satisfaction of both the technical editor(s)
and the ASTM Committee on Publications.
The quality of the papers in this publication reflects not only the obvious efforts of the
authors and the technical editor(s), but also the work of these peer reviewers. The ASTM
Committee on Publications acknowledges with appreciation their dedication and contribution
of time and effort on behalf of ASTM.
Printed in Chelsea, Mich.

December 1990
Foreword
The Symposium on Charpy Impact Test: Factors and Variables, sponsored by ASTM
Committee E-28 on Mechanical Testing, was held in Lake Buena Vista, Florida, on 8-9
November 1989. John M. Holt, Alpha Consultants & Engineering, served as chairman and
has also edited this publication.
Contents
Introduction 1
T H E PENDULUM-IMPACT MACHINE
Impact Tester Compliance: Significance, Sensitivity, and Evaluation--F. PORRO,

R. TRIPPODO, R. BERTOZZI, AND G. GARAGNANI 7
Comparison of Metrological Techniques for Charpy Impact Machine Verification--

A. K. SCHMIEDER 20
Influence of Dimensional Parameter of an Impact Test Machine on the Results of a

Test--G. REVISE 35
Factors Influencing the Accuracy of Charpy Impact Test Data--A. L. LOWE, JR. 54
Effects of the Striking Edge Radius on the Charpy Impact Test--T. NANIWA,
M. SH1BAIKE, M. TANAKA, H. TANI, K. SHIOTA, N. HANAWA, AND T. SHIRAISHI 67
THE SPECIMEN: NOTCHES
Evaluation of Fabrication Method for Making Notches for Charpy V-Notch Impact
Specimens---R. D. KOESTER AND S. E. BARCUS 83
Quantitative Comparison and Evaluation of Various Notch Machining Methods and

How They Affect ASTM E 23 a n d I S O R 4 4 2 Testing Equipment Results--
D. A. FINK 94
The Effect of Fatigue Pre-Cracking versus V-Notching on Impact Testing of Charpy

Specimens---B. A. F1ELDS, S. R. LOW I11, AND J. (3. EARLY 120
Pre-Cracking and Strain Rate Effects on HSLA-100 Steel Charpy Specimens--

S. MIKALAC, M. G. VASSILAROS, AND H. C. ROGERS 134
Significance of Precracking Variables for Slow-Bend Charpy Tests---

C. G. 1NTERRANTE AND J. J. FILLIBEN 142
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
THE SPECIMEN: SIZE
Specimen Size Effects in Charpy I m p a c t Testing---D. J. ALEXANDER AND R. L. KLUEH 179
THE TEST TECHNIOUE
Influence of Thermal Conditioning Media on Charpy Specimen Test Temperature---

R. K. NANSTAD, R. L. SWAIN, AND R. G. BERGGREN 195
Author Index 211
Subject I n d e x 213
STP1072-EB/Dec. 1990
Introduction
INTRODUCTORY REMARKS
The Symposium on Charpy Impact Test--Factors and

Variables had its genesis at the second meeting of
S u b c o m m i t t e e 4 - - F r a c t u r e of ISO T e c h n i c a l C o m m i t t e e 164--
M e c h a n i c a l Testing. S u b c o m m i t t e e 4 had the a s s i g n m e n t of
reviewing ISO R e c o m m e n d e d Practice for V e r i f i c a t i o n of
P e n d u l u m Impact Testing M a c h i n e s for T e s t i n g Steel, ISO
D e s i g n a t i o n R442, and of revising where necessary. Although
ISO, as does ASTM, r e q u i r e s that d o c u m e n t s be reviewed at
i n t e r v a l s not e x c e e d i n g five years, this document had not
been reviewed since it was first p u b l i s h e d in 1965 under the
j u r i s d i c t i o n of ISO T e c h n i c a l C o m m i t t e e 17--Steel. There
were 15 r e p r e s e n t a t i v e s from seven member c o u n t r i e s and a
representative from the European Community Bureau of
Reference (BCR) at that meeting. The m e m b e r s of ISO
technical committees are the national standard writing
b o d i e s - - n o t individuals; n a t i o n a l - s t a n d a r d s writing bodies
are o r g a n i z a t i o n s such as BSI, AFNOR, SIS, etc. Because
there is no n a t i o n a l - s t a n d a r d s writing body in the United
States, Congress has designated the American National
Standards Institute (ANSI), as the de-facto body and
therefore, the member from the United States.
At the s u b c o m m i t t e e meeting, a g r e e m e n t was reached that

numerous changes n e e d e d to be m a d e - - s o m e t o l e r a n c e s w e r e too
r e s t r i c t i v e , some were not r e s t r i c t i v e enough, but there
were problems in agreeing to the "correct" values.
Agreement was reached for some values because various
d e l e g a t e s i n f o r m a l l y p r e s e n t e d work that they had p e r s o n a l l y
performed, or reported on work that had been done in their
country. However, other values could not be agreed upon
because of d i v e r g e n t requirements in various national
s t a n d a r d s and the s u p p o r t i n g data for the various p r o p o s a l s
was not c u r r e n t l y available. It was s u g g e s t e d that an
i n t e r n a t i o n a l s y m p o s i u m be held to discuss the factors and
variables that effect the C h a r p y impact test so that
r e s e a r c h e r s around the world would have a forum at which to
present data that would answer some of the q u e s t i o n s that
had been raised. The USA r e p r e s e n t a t i v e , on behalf of ASTM
C o m m i t t e e E 2 8 - M e c h a n i c a l Testing, agreed to sponsor such a
s y m p o s i u m as part of the E28 m e e t i n g s in N o v e m b e r 1989.
This STP is the result of that symposium.
The original goal of having world-wide research

p r e s e n t e d on the factors and v a r i a b l e s of the Charpy test
was achieved. There were three sessions c o n t a i n i n g 16
papers presented by authors from five d i f f e r e n t countries.
Because a t t e n d a n c e exceeded e x p e c t a t i o n s , it appears as if
more than just those writing s p e c i f i c a t i o n s are i n t e r e s t e d
in the topic.
I
Copyright 9 of1990
University by ASTM
Washington International
(University www.astm.org
of Washington) pursuant to License Agreement. No further reproductions authorized.
2 CHARPY IMPACT TEST: FACTORS AND VARIABLES
SPECIFIC REMARKS
Twelve of the papers presented are being published in

this STP, and one will be published in the ASTM Journal of
Testing and Evaluation (Reference I). The twelve papers
fall into three categories, (1) those discussing the
p e n d u l u m - i m p a c t machine, (2) those discussing the specimen,
and (3) those d i s c u s s i n g the testing techniques; several
papers discuss more than one category. In summary, the
papers present i n f o r m a t i o n on:
the effect of many of the dimensional parameters of
an impact machine, including metrological
techniques to evaluate these parameters and a
compliance technique for verifying machine
acceptability;
the effect of the geometry of the striker, that is,
the 2-mm radius striker specified by ISO and much
of the rest of the world, and the 8-mm striker
specified by the ASTM;
the effect of the method of fabricating the notch of
a CVN test piece including fatigue precracking;
the effect of s p e c i m e n sizes in Charpy impact testing;
the effect of strain rate including slow-bend tests.
Because the dimensional parameters of the m a c h i n e s are

so very important to obtain "proper" impact values, the
papers by Porto, et.al., by Schmieder, by Revise, by Lowe,
and by Naniwa all discuss how the test machine can influence
the results obtained. These papers discuss the effects
ranging from the attachment of the machine to its foundation
to the m e t r o l o g i c a l methods used to determine angles and
linear dimensions. Several of the papers discuss several
potential sources for variation in test results due to
machine variations. Attention is drawn to the paper by
Porro, et.al, presenting the results of a study on the
compliance of a machine as a means of assessing its physical
condition. Naniwa presents the results of an in-depth
study of the differences in the behavior and the deformation
of the specimen when struck by an 8-mm striker (the "ASTM
striker") and when struck by a 2-mm striker (the "ISO
striker").
The specimen was investigated from two points of view:

(I) the method of preparing the notch, and (2) the size of
the specimen. The papers by Koester and by Fink studied the
effects of grinding versus s i n g l e - p o i n t machining; the
papers by Fields, et.al., by Mikalac, et.al., and by
Interrante, et.al, studied the effect of notch acuity and
the method(s) of obtaining a sharp notch. Alexander, et.al.
investigated specimen size.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authori
INTRODUCTION 3
The influence of the temperature c o n d i t i o n i n g media on

test results was reported by Nanstad, et.al. Their paper,
and R e f e r e n c e I, indicate that the t e m p e r a t u r e of the
specimen in the vicinity of the notch at the instant of
impact is not n e c e s s a r i l y the same as the temperature of the
c o n d i t i o n i n g media.
As a result of the various studies presented, ASTM

Committee E28 has initiated ballots changing some of the
r e q u i r e m e n t s of ASTM Method E23. ISO S u b c o m m i t t e e 4 has
begun to study the results to see how they apply to the
revision of their Method R442.
Prior to the Symposium, one attendee was overheard

saying, "I see that there is a symposium on the Charpy test;
what can be new there?" I believe that the s y m p o s i u m and
this STP are definite statements that much is happening in
the field of Charpy testing to further the u n d e r s t a n d i n g
of what is required to obtain a c c e p t a b l e Charpy test results
and the proper i n t e r p r e t a t i o n of those results.
ACKNOWLEDGEMENTS
I wish to thank the many people that helped to arrange

the s y m p o s i u m -- in particular, Dorothy Savini, and the
many other members of the ASTM staff, the session co-
chairmen, R.D. Koester, and R.J. Goode, and the many people
who reviewed manuscripts.
Thanks are also in order to the people who have been

instrumental in seeing that this STP was published. These
include Monica Armata, Rita Harhut, and the editors of the
ASTM Staff and Jim Perrin of the ASTM P u b l i c a t i o n Committee.
In addition, the original reviewers again reviewed the
revised papers to insure their quality.
John M. Holt
Alpha C o n s u l t a n t s & Engineering
1504 W i l l i a m s b u r g Place
Pittsburgh, PA 15235-4924
Reference I: Tobler, R.L., R.P.Reed, I.S. Hwang, M. Morra,

R.G. Ballinger, H. Nakajima, and S. Shimamoto; Journal of
Testing and Evaluation, Vol 19, January 1991, pp.34-40.
The Pendulum-Impact Machine
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions auth
Francesco Porre, Rodolfo Trippodo, Roberto Bertozzi and Gianluca
Garagnani
IMPACT TESTER COMPLIANCE: SIGNIFICANCE, SENSITIVITY AND EVALUATION
REFERENCE: Porro,F.,Trippodo,R.,Bertozzi,R.,Garagnani,G.,"Impact
Tester Compliance: Significance, Sensitivity and Evaluation",
Charpy Impact Test: Factors and Variables.
ASTM STP 1072 , John M. Holt, editer, American Society for
Testing and Materials, Philadelphia 1990.
ABSTRACT: The compliance is very sensitive to internal mechanical

factors concerning the load system, as the hammer, the tup, the
anvils and the base to foundation attachement.
In order to verify the sensitivity of compliance measurements, a
series of experimental tests has been performed, with artificial
and real defect located at the most critical parts.
In order to overcame the need of an instrumented impact tester an
instrumented specimen has been prepared, together with its
electronic system for impact tester compliance measurement.
The compliance measurement, after verification of the impact
tester with direct and indirect methods, as per ASTM E 23 or ISO
R 442, can be helpful for verification of the good working
condition of the pendulum and for the detection of onset of
anomalies.
KEYWORDS: compliance, impact testers, pendulum, Charpy specimens,
INTRODUCTION
As pointed out by Bluhm [i] the flexibilties and the softness of

the impact machine play a primary role in the determination of the
correct value of the energy spent to break the specimen.
Dr.Porro is Quality engineering supervisor at Ansaldo ABB

Componenti, via Lorenzi 8, 16152 Geneva, Italy;
Ing. Trippodo is the director of CERMET (Regional Research Center for
Materials), via More 26, 40068 San Lazzaro di Savena, Bologna, Italy;
Mr.Bertozzi is research scientist at CERMET; Dott. Garagnani is
research scientist at Department of Metallurgy, University of Bologna,
Italy.
7
Copyright9 1990 by ASTM International www.astm.org
The record of the strain of an instrumented tup actually made on

an instrumented impact machine, Fig.l, definitely supports the
hypothesis of the presence of vibrations during specimen rupture,
resulting in loss of energy by elastic deformations, in the case of
brittle fracture.
L
o
a
d
/
r,
TII'EIlRI~I .1$ m e o / d l u
Fig.N.l: load signal from an instrumented impact tester tup

showing typical vibrations during specimen rupture
In order to minimize the influence of this vibrational energy on

the adsorbed energy reading, it is necessary to have an impact tester
with low compliance.
This important conclusion motivated the authors to take into
consideration verification of the impact tester compliance to assure
homogeneity of behaviour from one tester to another.
It is well known that the reliability of the impact tester
measurements is a matter of discussion when two impact testers
(typically customer or inspection agency and manufacturer impact
testers) measure different energy values from specimens of the same
material.
This work is oriented to analyze the possibility to use the
compliance, together with other characteristic impact tester
parameters, for the detection of existing or impending anomalies.
BACKGROUND
The rule that governs the energy transformation during an impact

test is as follows:
Ep = Ea + Ek + Ee + Ef (i)
where Ep = potential energy of the hammer (weight * height) to be

converted into kinetic energy after the hammer release;
Ea = energy absorbed by the specimen during its rupture;
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authoriz
PORRO ET AL. ON IMPACT TESTER COMPLIANCE 9
Ek = kinetic energy remaining after impact;

Ee = energy stored by the system hammer~specimen~anvils by
elastic deformation;
Ef = energy lost by friction and windage during the blow;
The quantity Ee represents the energy stored and lost by the

loading system of the specimen and therefore unavailable for breaking
the specimen.
The energy dissipated as elastic deformation of the loading

system, for a given load P, introducing the definition of stiffness
that is the ratio load/deflection, is:
P
Ee = - - (2)
2 * Sm
where Ee = energy stored by the system hammer/specimen/anvils by

elastic deformation (J);
P = load (N)
Sm = system stiffness (N/m)
The compliance, or displacement under a given load, can be

expressed in terms of stiffness of the system as follow:
1
Co = (3)
2 * Sm
where
Sm = stiffness of the loading system (N/m)
Co = impact tester compliance (m/N)
After substitution, the formula (2) can be written:
Ee = Co * P (4)
After the original idealized model suggested by Bluhm [i] for the
determination of the stiffness, two methods are currently available.
The first, described by Venzi [2], has only experimental

difficulties; this approach has been followed by the authors and the
results obtained will be discussed in the following.
The second, used by Ireland [3], requires an instrumented impact

tester, presents sufficient mathematical difficulties to require a
computer for integration and shows lack of precision due to the
interpretation limits of the computer during the determination of the
characteristic points on load-time curve (yielding load and yielding
time).
Following the Venzi approach [2], the pendulum-specimen system

can be sketched as follows, during a blow in the elastic field :
loading center of
mass Sm Sp percussion
displacement
l<--loading system-->l< , specimen----> I
loading center of
mass Se percussion
vo--> Jk/h/k/h/k/k/~ * > x

displacement
I< equivalent system >I
where:
M = pendulum mass (specimen mass is neglected, as Bluhm [i])

Vo= impact velocity (just before impact)
X = displacement of the centre of mass M,coincident with the centre of
percussion (one degree of freedom assumed as Bluhm [i])
Sm = loading system stiffness, inverse of loading system compliance Cm
Sp = specimen stiffness, inverse to specimen compliance Cp
Se= equivalent stiffness (ratio load/deflection), inverse of
equivalent compliance Ce
The displacement "x" is the sum of the displacements of the

specimen and the loading system:
x = X + x (5)
loading system specimen
and the following law relates the three stiffnesses:
P P P
= - - + - - (6)
Se Sm Sp
where P is the load on the system (action and reaction at the

interface tup-specimen)
Then the relation between the stiffness and the compliance is:
1 1 1
= ~ + ~ (7)
Se Sm Sp
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorize
Ce = Cm + Cp (8)
In order to solve the equation, i.e to obtain the value of Sm, it

is necessary to have the values of Sp and Se.
The value of Sp can be c a l c u l a t e d theoretically. The f o r m u l a for
a u n n o t c h e d specimen with square cross s e c t i o n is :
4
I 4 * E * L )
Sp = 19)
3
W
where Sp = specimen s t i f f n e s s
(N/m)
E = elastic m o d u l u s of the m a t e r i a l
L = specimen width (typ. i0 * i0 mm)
W = span b e t w e e n anvils (typ. 40 mm)
The value of Sp for a s t a n d a r d unnotched s p e c i m e n made in AISI

4340 h a r d e n e d steel (55 HRC) is:
6
Sp = 133.4 * i0 N/m (i0)
The value of Se can be obtained e x p e r i m e n t a l l y with the following

considerations.
The law that d e s c r i b e s the e q u i l i b r i u m of the h a m m e r t r a n s l a t i o n
for a b l o w in an e l a s t i c field is (drop angle ~ < i0 ~ :
dX
M * - - + Se * X = 0 (ii)
2
d t
the s o l u t i o n of this e q u a t i o n is:
X = Xo * sin(~ * t) (12)
Se 89 2 * ~
where e = ( ) - (13)
M T
T = o s c i l l a t i o n p e r i o d of the s y s t e m
p e n d u l u m - s p e c i m e n for e l a s t i c b l o w
The value of Se is d e t e r m i n e d by the equation:
2 , 2
Se = ( ) * M (14)
T
By the k n o w l e d g e of Sp and Se it is now p o s s i b l e to solve the
e q u a t i o n (7) to obtain the value of Sm.
MEASUREMENT OF THE COMPLIANCE
The previous common approach for the measurement of the

compliance, following the Venzi approach, was to use an instrumented
tup in order to obtain the value of T by detecting the load signal, as
indicated in Fig. N.2.
J
Till .6 IlWeO/dlv
Fig.N.2: load signal from an instrumented impact tester tup

under low blow (deformation in elastic field) for
determination of typical oscillation period IT].
The requirement of having an instrumented impact tester, and the

scarcity of this type of machine, resulted in generally low interest
in using the compliance paramenter because of the difficulty in
determining it.
To overcome the need for an instrumented impact tester, and to
allow a low-cost determination of the compliance on impact testers, an
instrumented specimen was prepared together with an electronic system
for detection of the tup-specimen contact time, i.e., the half period
T/2.
The system consists of the following:
An unnotched specimen i0 mr, wide,10 mm heigh and 55 mm long, made
of AISI 4340 steel, hardened to 55 HRC, with a surface roughness of 32
rms. The specimen is provided with two threaded holes at its ends to
allow the insertion of two screw that are utilized as hooks for a thin
rubber band for fastening the specimen against the anvils during
repeated blows.
A longitudinal strain gage is cemented at mid length and mid
height on the specimen side opposite the hammer.
This strain gage is connected in bridge configuration to a strain
signal conditioner located near the specimen itself. This strain
signal conditioner is equipped with gain and balance (zero) adjustment
trimmers.
PORRO ET AL, ON IMPACT TESTER COMPLIANCE 13
The strain gage conditioner detects the strain signal due to the
displacement of the specimen during the impact of the hammer, i.e the
load. The output of the strain conditioner during absence of load
could be zero or a pre-set value.
The output of the strain signal conditioner is sent to an
electronic trigger that detects the strain signal as it changes from
the pre-set value.
The output of the trigger starts time counting (on a timer) when
the trigger detects strain signal and stops the counting when the
trigger detect the end of the strain imposed by the load
The output of the trigger is also sent to a counter that can
count the number of subsequent repeated blows while the timer measures
the total time of contact of each blow.
T/2, the oscillation half period , is the value that is
experimentally determined, as sketched in Fig.3.
load-time c u r v e
elastic low blow
angle of drop < 15 deg,
~/ T / 2 = half o s c i l l a t i o n period
~ H
first second third

blow blow blow
Fig.N.3: load signal as detected during a low blow and for two
subsequent rebound.
The accumulated time intervals (T/2 or its multiples), the number

of blows and the pre-set balance value are displayed on the
instrument.
The system is also provided with an output for an oscilloscope
for directly viewing the strain signal or for recording it.
The system arrangement for the measurement is presented in fig.4.
The interesting features of the system are the following:
i) portability: it is completly hand-portable;

2) simplicity : its electronic circuitry is very simple and very
common, made with standard industrial components;
3) flexibility: it is not fixed or made for a specific impact tester,
but can work on different machines, allowing
intercomparison between pendulums, labs, etc;
4) low cost : it is much less expensive than an instrumented impact
tester.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized
14 CHARPYIMPACT TEST: FACTORS AND VARIABLES
Fig.N.4: Typical instrumented Charpy specimen arrangement for time

of contact measurement under low blow.
The rubber bands prevent movement of the specimen under
repetitive low blows
SIGNIFICANCE OF THE COMPLIANCE
The value of the compliance in an impact tester is related to the

geometry and the ~aterial p~operties of the loading system.
At least the following components of the loading system should be
considered:
-the hammer and its fixtures to the supporting bar,
-the tup and its fixtures to the hammer,
-the anvils and their fixtures to the pendulum base,
-the pendulum base and its attachements to the floor.
The geometry of the loading system is defined by the manufacterer

of the impact machine and normally it is not possible to vary or
modify it.
Becouse the ~ompliance is affected by variation of the working
condition of the impact tester, i.e. change in the fastening condition
or wear of the mechanical components, it is therefore important to
periodically check the value of compliance in order to detect the
onset of anomalous conditions.
The following point out the significance of the compliance and
its power in the determination of change in the working condition of
an impact tester.
-First test:
After a m e a s u r e m e n t of the compliance of a low range i n s t r u m e n t e d

impact tester ( 20 J capacity) , several artificial defects were
introduced in order to v e r i f y the c a p a b i l i t y of the c o m p l i a n c e
m e a s u r e m e n t to detect such defects or anomalous w o r k i n g conditions
[5].
The choice of a low range impact tester was because, at the time
of this test, work in progress using the standard 360 J impact tester
did not permit the risk of loss of the W a t e r t o w n C e r t i f i c a t i o n due to
the i n t r o d u c t i o n of the artificial defects. M o r e o v e r it was
not possible to risk the p o s s i b i l i t y of irreversible damaging.
The i n s t r u m e n t e d impact tester was a Tinius Olsen model 74, with
a m a x i m u n impact energy of 25 J, "U" type hammer, with the tup
fastened to the hammer by four screws. The tup is presented in Fig.5.
The instrumentation, ETI 300 System, was m a n u f a c t u r e d by Effect
Fig.N.5: Tinius Olsen mod.74 (25 J) impact tester tup sketch

showing the four fastening screws.
The defects introduced in the system were as follows:

A =
referred as normal operating conditions
B =
loosened central bolt (bolt No.4)
C =
two loosened bolts (bolts No. 2 & 4)
D =
three loosened bolts (bolts No. 1,2 & 4)
E =
elastic deformation of fixing points the bolts I, 2 & 3 are each
loading two cup springs with a force of 50 Kgf.
F = elastic d e f o r m a t i o n of fixing points the bolts I, 2 & 3 are each
loading two cup springs with a force of 58 Kgf.
G = loosened attachement to the base
The results in terms of compliance measurements, taken by the pendulum

instrumentation, are presented below:
defect brief T/2 Sp Se Sm Co

s t a t u s description 6 6 6 -9
of defect msec N/m *i0 N/m *i0 N/m *i0 m/N *i0
A normal cond. 1.47 133.4 16.87 19.32 24.5
B 1 loose bolt 1.57 133.4 14.91 16.77 28.5
C !2 loose bolts 1.65 133.4 13.83 15.40 31.8
D !3 loose bolts 1.75 133.4 11.96 13.14 36.3
E 50 kgf springil.80 133.4 10.63 11.57 38.6
F 158 kgf spring 1.96 133.4 8.82 9.41 46.5
loose base 1.41 133.4 18.44 21.38 22.2
-Second test:
Compliance measurements were used to verify the working

condition of a standard (non instrumented) impact tester ( 360 J
capacity) manufactured by Metrocom Italy, during its initial
installation [6].
The impact tester was then moved and re-installed in another
laboratory, and new compliance measurements were taken.
All the measurements were taken utilizing the instrumented

specimen and the electronic equipment.
The results show the capability of the compliance measurements to

detect several anomalous situations during the installation, ranging
from the loosening of the foundation bolts, the presence of a thick
paint layer under the nuts (acting as an elastic medium), the
difference in anvil spacing, and the presence of an out of level
condition.
It is difficult to predict which is the correct compliance value
of an installed impact tester, because the value seems to be affected
by the system of fastening of the base to the floor.
Nevertheless after the istallation and the calibration of the
impact tester performed under the relevant specification (E 23, UNI
6882, ISO R442, etc.) the test of the compliance can detect the onset
of anomalous conditions.
The compliance values at the final fixing, for both the
first and second installations were determined after both direct
(metrological) and indirect (by standardized Charpy specimens)
verification had been completed.
The results of tests of the first installation, taken by the use
of the instrumented specimen and related electronics are the
following:
first installation: I
defect brief T/2 Se Sm Co

status description 6 6 -9
of the situation msec N/m *10 N/m *i0 m/N "10
anvil spacing 40.1 2.410 33.9 45.4 10.90
anvil spacing 40.6 2.340 35.9 49.2 i0.i0
loose base 1.775 62.5 117.0 4.25
looser base 1.846 57.8 101.0 4.90
1 loose anvil bolt 1.885 55.4 94.7 5.27
2 loose anvils bolts 2.25 38.9 54.9 9.10
7 all loose anvil bolts 2.350 35.6 48.6 10.20
8 first final fixing 1.911 53.9 90.4 5.52
The results of tests of the second installation, taken by the use

of the instrumented specimen are the following:
second installation: I
defect brief T/2 Se Sm Co
status description 6 6 -9
of the situation msec N/m*10 N/m*10 m/N*10
9 paint under base 1.884 55.5 94.9 5.26

fastening nuts 1.998 49.3 78.2 6.38
i0 cleaned base surfaces 1.852 57.4 i00.0 4.95

under fastening nuts 1.933 52.7 87.0 5.74
ii after pre-work 1.960 51.2 83.2 6.00
12 after base tigtening 2,000 49.2 78.0 6.40
13 after base levelling 1.940 52.3 86.0 5.80
14 after 2nd levelling 1.910 53.9 90.6 5.51
'15 final fixing 1.860 56,9 99.2 5.03
As previously stated, the parameters which affect the compliance

also affect the energy reading; the correlation between compliance
variation and energy variation is presented in the following table:
defect brief Co Delta Delta

status description -9 Compl. Energy
of the situation m/N *i0 % %
1 anvil spacing 40.1 10.90 +116.6 -4.7
2 anvil spacing 40.6 i0.i0 +100.7 -3.8

[ --
3 lloose base 4.25 -15.5 +l.1
8 first final fixing 5.52 +9.7 +0.3
14 after 2nd levelling 5.51 +9.5 -0.58
15 final fixing (ref.val) 5.03 0.0 0.0
It was important to measure the value of the compliance for

impact testers manufactered by different manufacturers, after the
completetion of both direct and indirect verification tests, in order
to have a table of the value of the compliance of each type of machine
in the "verified condition".
The measured values of the compliance, measured by the use of the
instrumented specimen and the electronic equipment, referred to the
impact tester are the following:
manufacterer capacity pendulum -9

J type Co*f0 m/N
Tinius Olsen 365 U 5.87
Metrocom 300 C 5.03
Galdabini 300 C 4.48
WPM (Ceast) 300 C 5.71
CONCLUSIONS
Tests performed in this study demonstrate that impact tester

compliance can be very helpful with other characteristic measurements,
in the verification of good working condition of impact testers and in
the detection of onset of anomalies.
A mandatory condition for the consistency of compliance
measurements is that the impact tester shall comply with standard
verification rules, both direct, as, for example ASTM E23 or ISO R442,
and indirect, with verification Charpy specimens.
Many of the parameters taken into account by these rules, as
shown with the tests performed during impact tester installation, will
greatly affect the time of contact, destroying the consistency of the
compliance measurements.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions autho
REFERENCES
[i] Bluhm, J.I, "The Influence of Pendulum Flexibilities on Impact

Energy Measurements", Symposium on Impact Testin@, Atlantic City,
june 27 1955, ASTM STP N.176
[2] Venzi, S, "La Resilienza Strumentata per la Misura del KIC
Dinamico",C.S.M (Centro Sperim. Metallurgico) Report I087R,
June 1978
[3] Ireland, D, "Effect Technology Inc. Technical Report TR 974-29R"
November 1974
[4] A S T M E-23 "Notched Bar Impact Testing of Metallic Materials."
Annual book of ASTM standards, Section 3, Vol 03.01
[5] Porro, F, Trippodo, R, Wagner, V, "Theoretical and Experimental
Evaluation of Compliance of Impact Testing Machines"
AMTT Report, Januari 1982,
BCR contract 861/I/4/143/80/12-BCR-I-10
[6] Trippodo, R, Bertozzi, R, "Programma di Messa a Punto e
Realizzazione di un Sistema per la Resilienza Strumentata"
CERMET Report, July 1987 N.50-el01
Albert K. Schmieder
COMPARISON OF MErROLOGICAL TECHNIQUES FOR CHARPY IMPACT MACHINE

VERIFI CATI ON
REFERENCE: Schmieder, A. K., "Comparison of Metrological Tech-

niques for Charpy Impact Machine Verification," CharDy Impact
Test - Factors and Variables: ASTM S_TP 1072 , J. M. Holt, Ed.,
American Society for Testing and Materials, Philadelphia, 1990.
ABSTRACT. Different measuring techniques were used to determine

some of the specified characteristics of nine Charpy impact
machines. In general, the techniques used were specified or
recommended by one or more national standards. For example,
the elevation of a raised pendulum was determined by direct
measurement with a ruler and also by calculation from the mea-
sured angle of the pendulum rod. Both methods gave equal values
with about the same reproducibility. On the other hand, signi-
ficant differences were found when the friction loss in the pen-
dulum was measured by a single swing and by multiple, successive
swings. Significant differences in the period of oscillation
were also found when the maximum angle of swing was 15 degrees
as compared with 5 degrees. Both values were specified as per-
mitred maximums in some national standards.
KEYWORDS: impact machines, Charpy machines, friction loss,

period of oscillation, clinometer
The increase in international trade has stimulated efforts to re-

duce the differences between national standards for materials speci-
fications and the methods of testing used to obtain the specified
values. This paper is part of that effort. The objective is to pre-
sent information which will be helpful in reducing the differences
between various standards which specify the characteristics of pen-
dulum impact machines.
In most cases, when the indicated value varies with the choice
of instrument or technique, the measuring technique is specified by
the national standards. In a few cases, different standards require
or at least recommend different techniques. These different tech-
niques were compared by using two or more to measure selected charac-
teristics of one or more testing machines. The characteristics cho-
sen for evaluation are:
Mr. Schmieder is a consultant on mechanical testing residing at

R.D.7, Box 330, Closson Road, Scotia, NY 12302.
2O
Copyright@ 1990 by ASTM International www.astm.org
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions a
SCHMIEDER ON METROLOG/CAL TECHNIQUES 21
(a) pendulum elevation

(b) friction and windage losses during a full swing
(c) period of oscillation of the pendulum
location of center of gravity

(f) affect of stem bending on elevation measurement
To simplify the presentation and reduce the need to refer to pre-
vious sections while reading, each of the six programs listed above
are reported and discussed under a separate major heading. An excep-
tion is that the conclusions drawn from each are gathered under one
heading. Elements common to several programs are reported in the
following section.
INFORMATION APPLICABLE TO ALL TESTS
Nomenclature
In most cases, the names of machine parts and quantities to be
measured will follow IS0-R442 [I] i. Most of these are defined pic-
torially on Figures 1 and 6 which are in that document. Uncommon
terms or specialized uses of common terms are defined below.
eg line - the straight line from the axis of rotation through
the center of gravity.
cg point - a point on the cg line at the same distance from the
axis of rotation as the center of strike. Note that the term center
of gravity has its usual definition.
specified aecurae[ - accuracy of a measurement required by a
standard method of verification.
~ermitted inaccuracy - one tenth of the specified tolerance.
Machines Whose Characteristics Were Measured
During the study of some of the variables listed above, nine
machines were measured; during others, only one. In each section, the
machines measured will be identified by the symbol shown in Table i.
The letter in the symbol indicates the form of pendulum hammer. The
letter C refers to the disk shape in which the striking edge can be
observed during a test. The letter U refers to the hammer form hav-
ing the striker projecting from an upper plate and hidden by side pie-
ces. It is not the intent of this report to identify and compare
individual machines, so the dimensions are nominal.
TABLE i -- Description of Machines
identifying Symbol Cl C2 C3 C4 UI U2 U3 U4 U5
Rating, J 3 20 350 2500 i00 350 350 350 400
(ft.lbf) (2) (15)(250)(1900) (75)(250)(250)(250)(300)
Angle of f a l l , degrees 150 150 110 130 135 135 135 120 135
Pendulum length, m 0.3 0.3 i 2 i I I I i
(ft) (1) (1) (3) (6) (3) (3) (3) (3) (3)
IFigures in square brackets identify references listed on the last

page.
Methods of Calculation
Unless stated otherwise, the methods of calculation were those
shown in the reference previously cited [i]
Measurement of An~/lar Position of the Pendulum

A clinometer was clamped to the pendulum shaft and read while
the pendulum was held in a stationary position. The instrument con-
sists of a frame in which is mounted a protractor carrying a sensi-
tive spirit level. The protractor is rotated by a micrometer screw
graduated each minute of arc. Angles were read to 0.5 minutes. The
angles at the end of the swing were retained by the position of the
friction pointer. This position was recorded by attaching a thin
strip of polished metal over the scale and marking this strip with
a fine scribe line at the tip of the pointer. A prop with a jack
screw was used to hold the pendulum at the marked position while the
clinometer was read. The prop was positioned so that the line of
action of the supporting force passed near the center of gravity of
the pendulum.
A 4X magnifier was used while reading or marking the pointer
position. The estimated accuracy of determing the pointer position
was 1/4 millimeter (0.01 inches). For a friction pointer of average
length, this corresponds to a maximum estimated error of 4 minutes
of arc.
If the pendulum is assumed to be rigid, the clinometer may be
mounted in any position without affecting the accuracy of the read-
ings relative to the reading at a known pendulum angle, in this case,
the vertical position of the pendulum. The only limitation on mount-
ing position is that the plane of the protractor be parallel to the
plane of swing of the pendulum. However, it is essential that the
clinometer does not move relative to the pendulum during all readings.
During these tests, the only situation in which the lack of ri-
gidity of the pendulum introduced a significant error was while the
pendulum was latched. The reported readings were corrected for this
error by a method explained in a later section.
COMPARISON OF METHODS OF DETERMINING P ~ D U L U M ELEVATION
Method of Test
Elevation of the pendulum of machine U4 was measured using two
methods: the first by direct measurement, the second by calculation
from measurement of the angular positions of the pendulum stem. For
the direct measurement, a beam with machined flange surfaces sup-
ported by jack screws was leveled using a precision level graduated
in intervals of 1.5 minutes of arc. The distance of a cg point above
the beam was measured using an engraved steel scale and a 4X magnifier.
The method of locating the eg point is described in a separate section.
Scale measurements were made at three positions of the pendulum:
latched, hanging, and supported on an adjustable prop at its static
position at the end of a free swing from the latched position.
SCHMIEDER ON METROLOGICAL TECHNIQUES 23
For the second method, the angular position of the stem was mea-
sured using a clinometer at three positions of the pendulum. These
were (1) while the striker was latched, (2) while the striker was
held in contact with a specimen in the testing position, and (3) while
the striker was propped at the position of the end of a free swing.
The first reading was corrected for stem deflection as discussed in
a later section. The second reading was corrected to the free-hanging
position.
Tests by each method were repeated five times to measure the re-
producibility under a variety of instrument orientations. Between
tests, both the reference beam and the clinometer were tuzned in the
sequence listed below.
Test Number 1 2 3 4 5
Orientation change Original Dnd-for- Upside End-for- Original
End Down ~hd
Result s
For the direct method, the elevation of any position is by defi-
nition the difference between the ruler reading at the position and
the reading at the free-hanging position. The non-dimensional fric-
tion loss per swing is the elevation at the latch position minus that
at the end of the upswing, that difference then divided by the latched
elevation.
The calculation of elevation using angular measurements was more
involved. The observed angle at the latched position was corrected
for stem deflection by the method described in the section on that
subject. The observed angle when the striker was in contact with a
specimen was corrected by the movement necessary to reach that posi-
tion from the freely hanging position.
The average friction loss for the five tests is 0.5_5 percent by
both methods. The standard deviations are 0.03 percent for the direct
measurement and 0.04 percent for the values calculated from angle
measurements, excluding the error in establishing the cg point.
Discussion
The values shown above indicate that direct measurement by a
scale resting on a level reference surface is equal in accuracy to
elevation values calculated from measurements of pendulum angle by a
clinometer.
The direct measurement has the advantages of requiring less ex-
pensive equipment which is available in many laboratories and of re-
quiring less knowledge of mathmatics to calculate the final result.
The major disadvantage of the direct method for an inspection ser-
vice is the difficulty of moving the reference surface and scale,
both being about two meters (six feet) in length. The clinometer and
associated equipment can be carried in a tool box that will fit under
an airplane seat.
The additional time required to set the level reference for the
direct method is about equal to that needed for the correction for
stem deflection when required. On average, the direct method requires
about i0 percent more time for a typical six-point scale calibration

and a single swing friction measurement.
FRICTION AND WINDAGE LOSSES DURING FULL SWINGS
Method of Test
Each machine was tested by the following series of free swings

from the latched position. The series was repeated at least once.
(1) single swing with the pointer set at full scale before release,
(2) with pointer set as in (1) swinging was allowed to continue
until the pendulum is near the latched position for the
fifth time, then the pointer is reset to ten percent of full
scale,
(3) with pointer set as in (1), repeatedly latched and released
without pointer reset until the pointer shows no further
motion,
(4) repeat (i!,
(5) repeat (2) but with the addition of a pointer reset to full
scale each time the pendulum is near the latched position.
The angle of the pendulum was marked at the following pesitions:
while latched, while hanging freely, and after each of the series
above. If the difference in marked position was greater than the
amount discernible by using a 4X magnifier, the series was repeated
twice more and the a v e r s e reported.
Results
The percentage values per swing are shown in Table 2 . The values
shown are calculated from the series of tests previously listed.
Test 3 of the series measures the los~ in the pendulum during one
swing. It is shown on the first line- of the table. Test 1 measures
the loss in the pendulum due to one swing plus the loss in the pointer
due to one upswing. The difference between the losses measured in
Tests 1 and 3 is the loss in the pointer. It is shown on the second
line.
Test No. 2 of the series measures the loss due to one upswing of
the pointer plus ten swings of the pendulum. This value minus the
pointer loss is divided by II and shown on the third line.
The fifth line shows the loss due to the pointer during one up-
swing. It is one fifth of the difference in loss during Tests No. 5
and 2.
The fourth line shows the average loss in the pendulum only. It
is equal to one tenth of the loss during Test 2 minus the single up-
swing loss in the pointer shown on the fifth line.
The last line is the ratio of the single swing loss in the pen-
dulum(determined by a single, isolated swing)to the corresponding
average loss from a series of ten successive swings. That ratio is
IThe line numbers in this section all refer to Table 2 .

',O
O'hO 0%'.O C ~ C q 0 0 o.~ O'x ~r~
u'h r O'-, (,~ O O~ ..~ ',.0 eq
9 9 9 9 ~D,--9o ~
00
O O
O ,--I OJ
O O O eqo O
oil
O
0'~ D..-
u~xO ~ - e,/",D cO O', Cm a x
9. ~ o,1 c,/ eq o'~ c~ p-.- O O O O O
9 9 ~ 9
D~- r'q O O O O O CqrH O
O~-'l
",O
~g'5o 0", (M o ~p,. 0.1 CO CO ",O eO Ex~
c'h o.]o cq 0'", O
9 9 . 9 eh
o. S O O O O O O'-,-~ D'-
OO O O O O O O O O O rHrHO
Or-I
O~ C'--
O ~".D
0"', C"~ 0 CO Oq - - ~ O Cq 0", r-t e..I O D. '.r~
P~ 9 ~ 9 9
"O OEX.-
..~ c~O o o o o CO r--I O O O O O 0'~ rH O
Oil
o
9
O'xO cq C~ ,--I c,~.~ O 9 Le'~ t.-I ~
P~ 1--1 r-I Orq 9
9 . ~ 9
g ~ OC:) o o o o D---
oil
O O O OJoO
o
r-t
9 D,-
9
-O
o , % ~ oTM Cx_ O'x \O O', O', O O",
O.--IO
~ oo D-- O O
C~
~ O O O ~ Ex- r-I O O O O O gq r--I O
OO,2
.r-t
c~
--~-o 0-, ..~ OO~Uh CO
r-t 0 CO o ~ ~ t-q t ~ , t h eq CO
r.D
o
o ~ o o d OJo~O
or-t
..0
o
~r-I C,/ 0", 0 0 , 1
(M C'--- ~"~ C'-- C'-.- r 0 ~, Cq 1/"10
9 ~ ~ 9 ED -r--I r - ~ t
! r.~ C~O 0 0 0 0
!
D-- 0 r-I OOr-I o./r-I 0
c,/
aOr-~ Ox ~ OO Oq Cch Cq
,-.I U ' ~ O (•-x•-h O ~ O t--I r-I u'hOO Cr~
r.D
6~ S6&d c
~ O~ O r'-I
o o o o o ~ o o
o ~
o O
.,H
~ o -~ ~ P~ ~ 1~.4 o
o ~ ~ r-I O
I ~ ~;~ ~ ~ ~
"" .c-I ---p o ~0o ~ u0~ ~ ~ ~0.H
I9~ o ~~, H ~"d
~ o~ , - q~ .p
I~
~
O 9
~
O
0
OH
~0 9
~a0~
~ a) ~q N g4 g~ 9 ~ b.oc6 horn ~
[rH r-'q ~D-H ~ t ~).H .,H

ib~O ~ - ~ . . ~ ~.-~ ~
I~"d '-' ,--t o'h " d nD 9,-I .H
I . H ~ .1-1 ~ 0~ ~ .H
1 ~ ~) 0 ~ 9 O O
l P~P~ ~ . ~
~ l ~ ~ -P O
~ o 9 orq o
obtained by dividing the v a l u e in the first line by that in the

fourth line. The average and standard deviations are 0.83 and 0.06.
Discussion
Typical standard values for maximum friction loss in the pendulum

and pointer combined are 0.5 percent (i~ and 0.75 percent (3) 9 The
sum of the first and second lines I of the tabulated results show all
except machines CI, C2, and UI meet the requirement of both standards.
These machines differ from the others in design. The first two are
small machines designed for testing nonmetallic materials. Machines
listed as UI and U2 are actually the same machine frame and bearings
supporting different pendulums. The bearings are adequate for the
rating of U2, which is four times that of UI. Presumably, the bear-
ings are larger than necessary for the rating of UI and, therefore,
have excessive friction losses.
A standard value (3) for maximum friction in the pointer alone

is 0.25 percent. This requirement is easily met by the machines used
for testing metals with the exception of U4, which slightly exceeds
the requirement.
The third line shows an arbitrary measure of the condition of

the bearings. The standard value (3) not to be exceeded is 0.40.
This criterion of the friction losses is in agreement with the one
above in the evaluation of the condition of the machines.
The bottom line shows that the friction losses per swing by the
multiple swing tests are somewhat greater than those for a single
swing. This is consistent with the concept of the following air flow
of one swing being an opposing air flow for the return swing. If
values from the multiple swing tests were compared to the maximum per-
mitted values shown in the standards, machines CI, C2, and UI would
again be found to have excessive friction. As would be expected due
to the measuring of a larger quantity with the sane instrument, the
precision of the value per swing by the multiple swing method is
greater than that for the single swing method. Other advantages of
the latter test are that it is less time consuming and that it can be
made without additional instruments if the accuracy of the energy
scale is assured by a previous calibration.
PERIOD OF OSCILLATION OF THE PENDULUM
Method of Test
The pendulum was displaced from the free-hanging position and

held manually against an adjustable, non-magnetized stop. At the
instant of release, a stopwatch reading in 0.01 second intervals was
started. The number of times the pendulum approached the stop was
counted. When a preselected number was reached, the watch was stopped.
The timed interval was i00 cycles unless prevented by the rate of
decay of the oscillation. Then, the count chosen was the maximum that
would be completed while the oscillation was still large enough to
be easily counted.
ILine number in this section refers to Table 2 .

Tests were made with the adjustable stop set to allow the pendulum
to be deflected from the vertical by approximately 7.5, 5.0, or 2.5
degrees. The adjustable stop was left in each position while the test
was repeated a minimum of three times. If the range of the observed
times was less than 0.1 seconds, the average was divided by the count
and reported as the period of the pendulum. If the range exceeded
0.1 seconds, the tests were repeated until the last test changed the
average by less than 0.02 seconds. Then, the last average was divided
by the count and reported as the period.
Results
Table 3 shows the change in the average period of oscillation due

to changes in the initial amplitude. In order to compare directly
machines of widely different sizes, the values of period of oscilla-
tion are shown as percentage decreases from the period with the largest
initial oscillation.
Discussion
Test methods for impact machine verification commonly require that

the center of strike be located within one percent of the distance
from the axis of rotation to the center of percussion. Since this dis-
tance varies as the square of the pendulum period, the permitted inac-
curacy of the period measurement is 0.05 percent. The sixth line I
shows that only about half of the machines tested achieved this degree
of agreement between the periods measured with the maximum specified
angle of swing, 15 degrees, and the minimum, 5 degrees. This indicates
that it would be desirable to have closer agreement between the vari-
ous standards on the magnitude of this angle. Factors pertinent to
the choice of this angle are considered next.
The derivation of the formula used to calculate the distance from

the axis of rotation to the center of percussion uses the fact that
for sufficiently small angles of swing, the sine of the angle and its
radian measure are equal. In this region, the period of the pendulum
is independent of the angle. The fifth and sixth lines show that the
period of the pendulum decreases progressively as the angle of swing
is decreased. This indicates that the range in which the assumption
above holds has been exceeded by the permitted angles of swing. Re-
ducing the maximum specified angle of swing to less than 5 degrees is
undesirable for two reasons. First, even at 5 degrees, some machines
with friction losses less than those specified elsewhere in the stan-
dard will not continue swinging for the specified I00 cycles. Second,
the reproducibility of the period during successive counts decreases
noticeably as the angle of swing decreases and also as the number of
cycles during the timed interval decreases.
Elliptic integrals ~] provide solutions for the period of the

pendulum which are not limited to small angles. If this calculation
would result in the corrected period being the same for all the angles
tested, use of the correction could be specified instead of further re-
stricting the angle of swing to be used during verification. The
seventh, eighth, and ninth lines show the results comparable to those
in the preceding two lines but corrected by elliptic integral
iLine numbers in this section all refer to Table 3.

solutions. For four of the machines, this correction reduced the varia-
tion due to angle of swing to less than the permitted inaccuracy. The
other machines showed a variation greater than twice the permitted in-
accuracy. Comparing the seventh, eighth, and ninth lines to the third
line shows that when the change in period is 0.05 percent or less,
the rate of decay of the oscillation amplitude from 7.5 degrees is
0.17 degrees per cycle or less. With m o r e rapid decay, the range of
the corrected values f o r the period increases progressively. Appar-
ently, the effect of friction on the observed period is not negligible.
Similarly, comparing the change in period to the twelfth line shows
that the corrected periods for three angles of swing vary by less
than 0.05 percent only if the angle after i00 cycles is greater than
60 percent of the intia/ value. An exception is machine C1 which is
not normally used to test metals.
VARIATION OF FRICTION LOSS WITH ANGLE OF SWING
Method of Test
Machine U4 is equipped with a device for changing the latch posi-

tion by five degree increments. Using this device, single swing tests
were made using the same test method described in the preceding sec-
tion on full swing tests; that is, by measuring the elevation at the
latched position and then at the end of the upswing. Since the dif-
ference is less than one half of one percent of the measured quanti-
ties, the results showed scatter large enough to leave the trend line
poorly defined. To reduce this scatter, tests were made by the mul-
tiple-swing method described in the preceding section on measurement
of the pendulum period by low angle swings. By this method, the change
in elevation is determined from the difference in position of the fric-
tion pointer at the top of the first upswing compared to the last
counted upswing of an uninterrupted series.
Two tests with successive swings were made for each latch posi-
tion. During the first test, the friction pointer was reset only
enough to contact the driving arm during the last lO percent of the
first and the last upswings. During the second test, the pointer was
reset to sweep from the maximum energy graduation to the end of the
upswing during each upswing.
Each type of test was repeated at each latch position at least

twice. If the results differed by more than twice the estimated read-
ing error of the scale at that level, the tests were repeated until
the change in the average due to additional tests was equal to or less
than the reading error.
Results
The scale of machine U4 reads absorbed energy. For a given num-

ber of cycles without full pointer reset, the change in reading is
the friction of the pendulum for a number of swings equal to two less
than double the number of cycles. The loss per swing was calculated
for each latch position. Table 4 shows the ratio of other vaAues to
the loss from the highest latch position. The amplitude of swing was
determined by two different measures: (1) the angle of swing and (2)
the residual energy. By definition, the residual energy is the machine
rating minus the scale reading. For each measure of amplitude, the
average of the value at first and last swings was taken as the point
at which the average loss per cycle occurred. The test conditions and
the ratios of these average values are also shown in Table 4. Table 5
shows the results of a linear regression analysis of the friction loss
and amplitude as measured by each method.
A similar series of tests were made with the friction pointer re-
set to the maximum energy graduation as each cycle was completed. The
energy loss with the reset minus that without the reset was divided by
the number of pointer resets to obtain the energy loss due to the
pointer. These values were converted in the same way as the values of
pendulum loss and reported in the same tables.
TABLE 4 -- Friction loss, amplitude of swing and residual energy

for various latch positions.
Latch Cycle Ratios for 2endulum Ratios for 2ointer
position, count
de~rees Loss Angle Energy ..Loss
.. Angle Energy
120 20 1.O00 1.000 1.000 1.000 1.000 1.000
ll0 20 0.828 0.965 0.908 0.91 0.965 0.908
I00 20 0.684 0.882 0.792 0.83 0.882 0.792
90 40 0.530 0.774 0.647 0.74 0.774 0.647
80 40 0.407 0.691 0.521 0.59 0.691 0.521
70 40 0.297 0.608 0.429 0.51 0.608 0.429
60 60 0.190 0.516 0.319 0.41 0.516 0.319
50 80 0.120 0.426 0.224 0.32 0.426 0.224
TABLE 5 -- Linear correlation of (a) angle of swing with friction loss

and (b) residual energy with friction loss.
For pendulum only For pointer only
Coefficient of correlation (a) 0.985 (b) 0.995 (a) 0,996 (b) 0.997
Slope of best fit line 0.662 0.893 0.85 1.14
Loss intercept -0.600 -0.170 -0.19 +0.13
Discussion
Most test methods that require or suggest a correction of the

absorbed energy for friction loss assume that loss to be proportions/
to the angle of swing. This is equivalent to assuming Coulomb fric-
tion in which the friction force is independent of velocity. A dif-
ferent reasonable assumption is that the loss is mostly due to wind-
age. Then, for blunt shapes such as the pendulums, the loss varies
as the square of the velocity, which in turn varies as the elevation
of the pendulum at the top of its down swing or upswing. This ele-
vation is proportional to the residual energy; that is, the energy
at the latched position minus the absorbed energy. The purpose of
these tests is to compare the results from these two assumptions with
the measured values of friction work during swings from various ele-
vations. To quantify this comparison, a linear regression analysis
was made of the friction work with each of these measures of the amp-
litude of swing. The pendulum loss and the pointer loss were con-
sidered separately.
If the two variables were perfectly proportional, the coefficient

of correlation would be 1.000. Due to the ratio form in which the
data were analyzed, a perfectly proportional relationship would re-
sult in a slope of 1.000 and an intercept at 0.000 loss. The values
in Table 5 show that for the pendulum loss, the assumption of loss
proportional to residual energy is significantly more accurate than
the angle~f-swing assumption. For the pointer loss, the two assump-
tions seem to be equally applicable.
LOCATION OF A LINE FROM THE AXIS OF ROTATION TO THE CENTER OF GRAVITY
Method of Test
Several methods were used to mark or measure the position of the

cg line and cg point on machine U4. Only one of these methods was
used when testing the other machines. Before any of the tests, the
pendulum was started in a small oscillation in a room without percep-
tible air currents. Measurements were made after the pendulum came
to rest. To redistribute the lubricant in the bearings, the pendulum
was swung from the latched position between each small-swing measure-
ment.
The methods used consist of two steps. The first step is to lo-
cate the striking edge relative to the specimen supports. The second
step is to determine the distance at which a vertical line through
the axis of rotation passes a specimen or pin resting on the supports.
The first step was accomplished by either of the two following

devices and procedures. The first device was a proximity detector
mounted on a micrometer calibrating stand. The oscillation decay to
rest was recorded on a chart. Then the pendulum was moved to contact
a pin resting on the specimen support. The micrometer was advanced
until the record again showed the rest position. The second procedure
was similar except that the proximity detector was replaced by a dial
indicator supported on a magnetic stand. The stand was advanced toward
the latch until the spindle tip was separated from the hammer by the
smallest visable gap. The bezel was set to zero, then the pendulum
moved to contact the pin and the indicator read.
The second step used one of two different devices, either a plumb
bob and scale or a elinometer. The plumb bob string was held above
the shaft so as to barely touch a machined portion while the bob tip
was just above a scale held horizontally against the anvil portion of
the specimen support. The clinometer was clamped to the pendulum stem
and read while the striking edge was pressed against a specimen or pin
on the supports. The reading was adjusted by an angle equal to the
motion measured in the first step divided by the pendulum length.
For U-type pendulums, a depth micrometer was used to transfer to

the outside surface the distance from the leading face to the striking
edge and also the distance from the plane of the bottom to the cen-
ter of strike. From the point so established, the distance determined
from the measurements in the two steps above was laid off horizontally
to establish the cg point.
Results
It was noted that when the striker was brought into contact with
the pin on the anvils and released, the pin rolled or slid to main-
tain the contact. This caused an obvious increase in the rate of de-
cay of the oscillation. This effect was eliminated by avoiding con-
tact between the pin and the striker during oscillation. In testing
machine U4, a standard Charpy specimen was used in place of the pin.
The specimen was dragged when the pendulum was released from contact
to start the oscillation. When the oscillation was started with a
gap, the presence of the specimen still had an effect readily measur-
able on the proximity detector record. The position at rest was
0.05 mm (0.002 in ) closer to the anvils when the specimen was located
there.
Proximity detector records of repeated tests on machine U4 showed
no discernable shift of the rest position after oscillation even though
the record was readable to 0.01 mm (0.0004 in).
From the reproducibility, it was estimated that the error in

measuring the distance between the striking edge and the specimen sup-
ports by the two detection devices is 0.i and 0. 3 minutes of arc for
the proximity detector and the dial indicator, respectively.
The error in establishing the cg line was similarly estimated at

0. 5 minutes of arc by the clinometer and 1. 5 minutes by the plumb bob
and scale.
Discussion
Standard values of accuracy for determination of the elevation

of the pendulum are 4 minutes of arc (I) or 0.I percent of the ele-
vation (3). These limits are equivalent for a typical machine having
an angle of swing of 240 degrees. Comparing these values to the es-
timated accuracies above shows that the plumb bob method of determin-
ing the cg point contributes to the error about one third of the speci-
fied maximum, which seems acceptable.
If an error of the maximum amount specified occurred in locating

the cg point, this amount would be added to the down swing and sub-
tracted from the upswing such that the loss in determining pendulum
friction would be 0.2 percent of the elevation. Since the pendulum
friction loss is specified as 0.5 percent (i), the effect of the error
is 40 percent of the quantity. This is four times the permitted in-
accuracy of i0 percent of the quantity being measured.
It is known that repeated blows to hardened steel, properly or-

iented to the earth's magnetic field, will cause the steel to become
magnetized. Such magnetization of the striking edge and anvils is
thought to be the cause of the specimen movement noted above. It
might cause a significant error if the free-hanging pendulum is very
close to a specimen of magnetic material.
The principal objective of the early section on comparisons of

methods of measuring elevation was to compare the clinometer method
to the scale method. Therefore, the cg line and cg point were es-
tablished once and used for all five tests. The results above indi-
cate that if these references had been re-established each time, the
standard deviation for the percent friction would have been 0.06 and
0.05 for the scale and the clinometer methods, respectively.
CORRECTION OF CLINOMETER READINGS FOR P~TDULUM ROD B ~ D I N G
Method of Test
The clinometer was attached at six equally-spaced positions along

the pendulum rod of machine U2 and read both while the pendulum was
supported by the latch and while the pendulum was supported on an ad-
justable vertical prop whose axis, extended, passed close to the cen-
ter of gravity of the pendulum. Machine U2 was selected for these
tests because the latch is located at the shaft hub where it did not
limit positioning of the extensometer.
The prop was adjusted to return the center of gravity of the un-
latched pendulum to the same position it had while latched. The mo-
tion of the center of gravity was measured by means of a dial indica-
tor supported on a rod resting on the machine foundation and having
the spindle touching the hammer at a point under the center of gravity.
To calculate the location of the center of gravity, the dimensions

of the hammer and pendulum rod were recorded, except for the wall
thickness of the cylindrical rod, which was not accessible.
Results
The correction values tabulated below are equal to the angle of

the cg line minus the clinometer reading. The tabulated position of
the clinometer is the distance from the axis of rotation to its mid
length as a percentage of the distance from the axis of rotation to
the center of gravity.
TABLE 6 -- Variation of clinometer correction with position.
Distance, percent 20 30 40 50 60 70 80
Correction, minutes of arc -4. 5 -2.5 -i +i +2 +3 +3
The position of the center of gravity was calculated by the mo-

ments of the calculated weights of the individual portions of the pen-
dulum about the axis of rotation and dividing by the total weight.
The value obtained by assuming the rod to be standard weight pipe dif-
fered by 1.3 percent from that obtained by assuming extra heavy pipe.
Both positions were within 1.5 percent of the mid point between the
top plane of the hammer and the center of strike. For these calcula-
tions this mid point was assumed to be the center of gravity.
Discussion
The central portion of the pendulum rod is elastically deformed

upward by the bending moment due to the force from the latch and the
component of the weight of the hammer perpendicular to the rod axis.
Thus, when a clinometer is attached near the axis of rotation, it will
read an angle larger than the angle of rise of the center of gravity.
Conversely, if the clinometer is attached on or near the hammer, the
observed angle will be smaller than the angle of rise. The theory
for deflection of simple beams shows that at the maximum deflection,

the tangent to the beam has the same slope as a line between the
points of support. Therefore, when the clinometer is located there,
the correction is zero. Furthermore, the theory shows that for an
approximately straight beam of uniform cross section, the point of
maximum deflection is
x =
~a (a
3
+ 2b) , when a is greater than b, (1)
where:
a = the distance from an outer loaded point to the intermediate
point;
b = the distance from the other outer loaded point to the
intermediate point; and
x = the distance to the point of maximum deflection measured
from the same loaded point as distance a.
For machine U2, distance x converted to be comparable to the positions
in Table 6 is 42 percent. Thus, the theoretical value of the point of
zero correction agrees with the experimental value interpolated from
Table 6 within the estimated experimental error.
For machines with a pendulum rod of variable cross section, the
formula above should not be used. It is usually simpler and faster
to measure the correction than to derive a comparable formula for
that specific shape. An example of this case is machine C3 which has
a tapered pendulum rod of I-beam cross section. Using position mea-
surements comparable to those in Table 6, the latch is at 40 percent.
With the clinometer at 63 percent, the measured correction was +1.8
minutes of arc.
CONCLUSIONS AND RECOMM~NDATIONS

The direct method of measuring elevation and the calculation of
elevation from measurements of pendulum angle are about equal in ac-
curacy and time required. It is recommended that both be permitted
by standard test methods.
The relationship between the loss per swing by the multiple swing
method compared to that of the single swing method is consistent
enough to allow the use of either in evaluating the machine condition.
However, the accuracy of the single swing method is not adequate for
measuring the specified friction losses. It is recommended that a
multiple swing method be specified. The multiple swing method takes
less time and requires no auxiliary equipment, which further recommends
its use.
The center of percussion can be determined with useful accuracy
if the period of the pendulum is measured while the friction losses
are limited to an amount which will permit i00 cycles of oscillation
after release from a 2.5 degree displacement from the vertical posi-
tion, with the added limitation that the amplitude at the lOOth swing
be at least one half of that at the first swing. For some machines
it may be necessary to suspend the pendulum and shaft from well lu-
bricated centers to meet this requirement.
For the eight machines tested, the friction loss in the pendulum
during an upswing was found to be more nearly proportional to the
change in the residual energy than to the change in angle. It is
recommended that standards requiring or allowing a friction correction
use that assumption.
For machines with pendulum rods of uniform cross section, the
error in the clinometer reading while the pendulum is latched can be
eliminated by attaching the clinometer at the point of maximum bending
deflection of the rod. The location of that point can be easily cal-
culated. If the pendulum rod has a non-uniform cross section or the
clinometer is attached at other locations, significant errors in the
angle of fall may result unless the observed angle is corrected for
the deflection of the pendulum rod.
ACKNOLEDGMENTS
The scope of this investigation would have been more limited if

the author did not have permission to visit several laboratories and
make measurements on machines located there. The assistance of the
following people in arranging such visits is gratefully acknowledged.
N.V. Cjaja, Schenectady Materials and Processes Laboratory
G.J. Leclerc, General Electric Co., Corporate Research and Development
R.E. Pasternak, Army Materials and Technology Laboratory
REFER~CES
[1] Recommendation R424 Verification of Pendulum Impact Testin G

Machines for Testin 5 Steels, ist ed., July 1965, International
Organization for Standardization.
[2] Designation E 23-86 "Standard Method for Notched Impact Testing
of Metallic Materials," 1989 Annual Book of ASTM Standards, Volume
03.01, American Society for Testing and Materials, Philadelphia,
Section 5.2.6.2, p. 196.
[3] von Karman, T., and Blot, M.A., "Elementary Problems in Dynamics -
Motion of a pendulum," in Mathematical Methods in Engineering,
Mcgraw Hill Book Co., New York and London, 1940, pp. 115-119.
G. Revise
INFLUENCE OF DIMENSIONAL PARAMETER OF AN IMPACT TEST MACHINE

ON THE RESULTS OF A TEST
REFERENCE: Revise, G., "Influence of Dimensional Parameter of an

Impact Test Machine on the Results of a Test," Charpy Impact Test:
Factors and Variables, ASTM STP 1072, John M. Holt, Ed., American
Society for Testing and Materials, Philadelphia, 1990.
ABSTRACT: The calibration of impact test machines is done by two

methods. A direct method which consists to verify and control the
dimensions of the machines; an indirect method which consists to
compare the results of a test done with reference test pieces,
between a reference machine and a machine which is to verify. The
values of the geometrical parameters of the machines, have an
influence on the results of a test. The present work quantifies
the parameters variations and compares the results obtained with
bending specimens and with charpy specimens. As a matter of fact,
the influence of the dimensional parameters of machines can be
finally expressed in different values of energy obtained with the
test pieces, and allow the comparison of different types of
specimens. The choice of reference specimen is now done by ISO and
by ECISS. This present work h a s brought some constructive element
before the choice.
KEYWORDS: impact test, charpy machine, resilience,

bending, anvils
Gilbert Revise
Chief of Department
Laboratoire National D'Essais
5, r u e E n r i c o F e r m i
ZA T r a p p e s Elancourt
Trappes FRANCE
35
Copyright 9 of
University 1990 by ASTM
I - Introduction
The impact test machine often known as Charpy machine, is

used for the characterisation of resilience which is a
relevant characteristic of a material and specially for
steel 9 The test is very simple in its principle and in
its procedure, but it gives highly dispersed results even
when the homogenity is very good9
This high dispersion is not acceptable today and, if it

is due to the machine, it can be improved 9
Some years ago, the "Bureau Communautaire de R~f~rence"

from European Community undertook a research to make a
resilience standard sample to calibrate the impact test
machines in the same conditions of an ordinary test. This
study had two parts :
9 study of the manufacture of reference test piece

which should reduce the dispersion of the results.
9 study of the influence of the mechanical and

dimensional parameters of the impact test machine
on the results.
This part should have been done after the first one,
using the reference sample 9 Because some difficulties
have appeared in the manufacture of the sample, we
performed the work with bending specimens which were used
in the ISO recommendation, and in the French standard,
for the calibration of the test machines 9
At the same time, we found in literature other works

which were done with Charpy specimens showing the
influence of an impact test on the results of a test.
A part of these works is presented here.
REVISE ON DIMENSIONAL PARAMETER 37
2 - The impact test machine is constituted of a pendulum

oscillating around a horizontal axis which is supported
by a rigid fixed frame. The different active and reactive
parts of an impact test machine are showed on figure I.
The French and ISO standards allow a certain tolerance

for the geometrical parameters of the machine.
We have made some variations of those parameters, and we

made bending as indicated by the French Standard tests in
the different configurations obtained.
The tests have been performed on metallic samples of 3, 5

and 7 mm thickness, as per the French standard, on a
300 J Wolpert impact test machine.
The different parameters for the active part, were :
- the position of the center of percussion

- the radius of the knife
-the speed of impact
And for the reactive part of the machine :
- the distance between the anvils

- the radius of the anvils
- the position of the plan of the anvils
- the position of the plan of the support
- the position of the sample between the anvils
Figure 1
axe o f
rotation "~
>
--V-
I /
/ anvils
A - Influence of the p o s l t l o n of the c e n t e r . p e r c u s s i o n
If the center of p e r c u s s i o n and the impact point are not

similar, occurs a torque which gives e l a s t i c deformations
to the arm of the pendulum, during the test.
This p h e n o m e n a was shown with two arms of the pendulum

which were of the same lenght but with different cross
sections. The characteristics of the two arms, are shown
in the table I.
Table I
Arm Length Type and Moment of Equivalent of length pendulum

mark radius ine=ti~
(=) (=) (==4) Without With
additional additional
Rod
A R = 14 36,2 " 106 788,8 800,i
635
I Tube I
I B r = 27 280 " 106 773,3 I 800,4
I I
[ I
In theory it is necessary to place an additional mass on

the two pendulums to bring the center of percussion on
the impact point. The distance between the rotation axis
and the impact point is fixed at 800 mm on the impact
test machine.
The deformations of the arm of the pendulum measured

during a bending test of 5 mm thick samples, shows that,
with the arms A and B, the vibrations have a larger
amplitude if the pendulum has an additional mass. This
phenomena is reproduced on the records shown on figure 2.
Figure 2
Deformation Deformation
Arm A Arm B
~#m "m
50
tl
~m/m 9
5O
~m/m
I00,
9
2 4 6 T~ms 0 ~2 ~4 ~]6 T~.~ms
If we simulate the arm of the pendulum as a beam on two

anvils which is submitted at a concentrated charge, after
calibration of the measured deformations, a very simple
calculation permits the evaluation of the energy which is
stored in the arm. In this case, the energy is
approximatlvely equal to 0.I Joule, which is negllgeable
in the evaluation of the energy absorbed by the bending
of sample.
As shown in table II, the influence of the rigidity of

the arm appears as negllgeable in the case of a bending
test. However, the dispersion is greater when an
additional mass is mounted on the hammer of the pendulum
with the aim of changing the center of percussion.
Table I I
Arm mark Average Maxi Standard

Energy Distance deviation
(J) (J)
I
I Without
[ additional 69,7 0,5 0,2
I m~SS
[
With
additional 68,8 1,2 0,5
m~ss
I
,,,
Without
mass
With
mass
B - Influence of the radlus of the knife
At the ISO meeting held i n Washington in September 1988,

the Japanese delegation presented a report on the
comparison between the ISO and ASTM knives. What we show
here, is only the effect on energy value with both
knives, during a bending test.
Table III
t I
Bends ISO Knife ~%~Kns J Absolute J Relative
sample Idiffere=ce Jdis
thickness "Average MsxCmu~ Average Maxlmum (J) to ISO
(mm) Value difference Value difference
Ca) Ca) Ca) Ca)
21,7 0,5 22,0 0,5 +0,3 +1,4

65,8 0,9 66,7 0,3 +0,9 +I,4
156,8 0,9 161,8 4,6 +5,0 +3,2
i
University of Washington (University of Washington) pursuant to License Agreement. No further reproductio
The values given in table III are the averages of 5

tests. The absolute difference between the measured
values exist whatever the energy level, and its value
increases with the energy level.
The examination of the tested sample showed that the mark

of the knife is clearly visible ; with a profile
projector, it is possible to verify the radius of the
knife.
For exemple, on figure 3, are given the marks obtained

with three types of knives (ASTM, ISO, and a mark with a
damage knife).
Figure 3
u " d. "
Pr
University of Washington (University of Washington) pursuant to License Agreement. No further reproduction
C - Influence of the impact speed
The potential energy and the impact speed are both

related to the fall height of the pendulum. Some tests
have been done with 3 mm thick sample and with impact
speeds equal to 5.3 and 3.2 m/s. The results are in
table IV.
Table IV
I
Impact Speed I Measured Energy
(m/s) l (J)
.... i
5,3 I 21,7
3,2
I-
I 21,4
f
This difference which appears is not significant. The

tolerance of the standard appears as very good (for
bending specimen).
D - Influence of the distance b e r g e n anvils
The French standard states 0.5 ram tolerance for

industrial machines and 0.2 mm tolerance for reference
machines. The distances which has been selected for the
test are 40 mm, 40.2, 40.5 with a tolerance of .02 mm.
The distances and the position of the knife between the
anvils has been verified with a casted print.
The measured energy decreases with increasing distance

between anvils. The relative difference in percent is
more important when the energy is low.
Table V
1 I
]rblck~essofl DISTANCE B E T ~ E N ~-~ILS
Itbe be~ding 1
s&urple Id = 40,02 d = 40,2 d = 40,5
Energy
(d) Energy [D~fs Difference Energy I Di~erence~Difference
(J) (J) (~) (a) (J) (%)
24,4 23,0 -1,4 -5,7 22,9 -1,5 -6,1
69,0 67,6 -1,4 -2,0 66,9 -2,1 -3,0
156,8 154,1 -2,7 -1,7 150,3 -6,5 -4,1
Figure 4 Influence of the distance between anvils
U.
~s61
154
t52 .- e =7 rc,~
150 I I
4'0 40,?. 40,5 mm
d.
70 l, t
66 . e=5mm.
4b 46,a 46,g,~,.
24 I~'-~ 3[ -
22 .... l 1 e:3mm
40 4~,2 40,5 mm
4o,0 .40,2 ,~o,:.s "d~
Figure 5
-2 \
-3 % "" D ~ : 5rr.m
\
-4
\ "'<
-5
-6.
I
distar~e
E - Influence of the radius o f the a n v i l s
The French standard states a radius of anvils between 1

and 1.25 mm. We manufactured anvils with the same
dimensions, rigidity, fastening system, but with
following radius 0.9, 1,0 and i.I mm. These anvils have
been set up on the impact test machine.
Table VI
I I I
[ Radius It = 0.9 ml [ Radius R = I.I ml I
I 1 1 I
Thlck~ess of!Kadi~s=lO m~ l~verage Absolute IDis~ance' ~ A~erage 1D~s~ance ~ 1
Icbe ben,lng Mean Value , Value dlff ...... [RI - RO 9 . Value Absolute I .I - Rl., I
sample 1 ~ 9 ~'~ J Idlfference
(=) (J) I (J) (J) I zx I (J) (J) l ~ l
3 21,7
1-I 21,4 -0,3 -I,4 20,6 -i,I -5,1
I
5 65,5 ] 64,9 -0,6 -0,9 64,1 -1,1 -1,7
I
7 148,6 I 148,9 +0,3 +0,2 146,1 -2,5 -1,7
[
46 C H A R P Y IMPACT TEST: F A C T O R S A N D VARIABLES
Figure 6 Influence of anvils radius
~r'
1~ radius(ram.}
X
/ '/
m/5// %1
-1 /
&/
e=7mm
r
-2
--3
-4"
-5
e=3mm
Figure 7 Values obtained

d
149
\
148
147 \\
146
I
d
66
65
r
64,
I
q9 %1
20 I ......
0,9 1 %1
CopyrightbyASTMInt'l(allrightsreserved);TueDec1512:59:52EST2015
Downloaded/printedby
UniversityofWashington(UniversityofWashington)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
In the aim of having more information on the influence of

this parameter, measurements of strain and displacement
were made for each test. The maximal effort and
deformation has been measured. The averages are shown
table VII.
Table VII
I I
l I R= 0,9 m m R = 1,0 m m R = I,I = m
Ithickness ofl
]Che sample F maxl d F mmy~ I d F ~.4 d
(,=) (s~) (~=) (k~) I (,~) (~) (=)
I
I
0,90 21,0 0,95 I 20,8 0,92 22,4
I
3,0 23,1 2,95 I 23,2 2,85 23,9
I
6,47 22,6 6,50 I 22,9 6,29 23,0
9 I.
The observed differences are more important between

radius 1.0 and 1.1 ram than between radius 0.9 and 1.0 mm.
However those differences are not negligeable and, when
there are compared to the reference values, shows the
important role of the radius of anvils in the measured
energy.
Since the anvils are opposing the pendulum action, the

mark on the sample increases with increasing the anvil
reaction and decreasing anvil radius. As the print of the
anvil is kept on the test sample, it is possible to look
for damage of the anvil. The shape of the sample is
measured with a profilometer. The exemple on figure 8
confirms that these phenomena are more important for the
sample with the highest bending energy.
figure 8 Print of the anvil on the sample
R = q9 , ~ . R .%0 ram. R=11 ram.
2C,,;m
F - Influence of the position of the plan of the anvils
Due to the real difficulty of the evaluation of the

position of the vertical plane of the anvils and the very
severe specification in the French standard on this
point, we have carried out investigations to evaluate the
influence of this parameter.
We used anvils with dimensions generaly equivalent to

those normally used, but with the planes P.' and P."
identical and making on angle 8 with the p~ane which
contains the rotation axis. The tangent is successively
equal to 5/1 000, and I0/I 000
(@I = 0~ 17' Ii" , @I = 0~ 34' 22'').
figure 9
i/if /
The results of the tests are shown on table VIII.
Due to the difficulty in evaluating this parameter and of

its influence on the result of the test, we recommand
taking 2.5/I 000 for (tan @).
In the case when PI' and Fl" are parallele, the distance
between these two planes must be such as the plane on
which the sample rests makes an angle @I with the
vertical plane so that its tangent is less than
2.5/I 000. A simple calculation gives a distance less or
equal to .118 mm. The tolerance of + .I mm seems to be
acceptable. 0 mm
G - Influence of the position of the plane of t h e s u p p o r t
The same questions which arise about the anvils, have

been set up about the position of the plane of the
support of the sample.
Figure i0.
Different supports has been successively mounted with the

same anvils. Their planes P_' and P2" are identical, and
make with the horizontal ~lane P9 (which contain the
rotation axis), an angle 89 whZch tangent was set
successively 5/1 000, then 10/F 000 (@2 = 0~
82 = 0"34'22").
Table VIII
I i
l'hicle~ess ltg(P i P"I) =0 I ~g (Pi P"I ) = 5/1 000 Cg (Pi P"l ) ~ 10/1000
of the
bendfug Average I" Average Absolute [ D ~ t a n c e Average I Absolute I Distance
s~mple Value 1 Value difference Value Idifference I
(~a~) (J) I (J) (J) (~) (J) (J) I (%)
I
I ' 'I
21,7 I 21,2 -0,5 -2,3 20,9 -0,8 1 -3,7
65,5 1 6~,2 -1,3 -2,0 63,8 -1,7 I -2,6
l 1
IL8~6 I 147,2 -1,2 -0,8 i&6,6 -2,0 1 -1,4
If the axis of the knife doesn't hit the sample in

direction perpendicular to its longitudinal axis the
sample can be twisted.
If one of the branchs of the sample is in the reference

plane, the other one makes a c e r t a i n angle with this
plane, but it is difficult to evaluate this angle.
Table IX.
Table IX
I
~Itk~ess Itg(~ P~ )=o tg (P~ P%l = 5/1 ooo tg (P, p~) = lO/~Ooo
of ~he
5ending Average Average IAbsolute IDfstancel Average I Absolute I Distance
sample Value Value [difference Value Idifference I
(==) (J) (J) (J) (~) (J) I (J) I (~)
i I
3 L
21,7 21,1 -0,6 -2,8 21,0 l -0,7 I -3,3
] I
65,5 64,8 -0,7 -I,1 64,2 l -1,3 I -2,0
148,6 147,9 -0,7 -0,5 166,4 I -2,2 I -i,5
H - Influence of the posltlon of a bend s a ~ l e
The position of the sample depends on the technique of

specimen placement and is Independance of the machine.
The tests were made with an offset equal to 0.5, and 1.0
Table X
7
I Offsett~.ng 9 = 0,54 O~f~ettiU~ e = 1,00~
1 Offse~tln g
Ir~ick=ess of~ e = O ~
I the sample E~ergy Energy Absolute ]Distance Energy Ab6olute Distance
Idifference Idiffere~cel
I (=z) (J) (J) (J) (~) (J) (J) (%)
I
I 3 24,4 23,7 -0.7 -2,9 22,7 -1,7 -7,0
I
I s 69,0 68,1 -0,9 -1,3 65,9 -3,1 -4,5
I 7 156,8 L55,8 -I,0 -0,6 152,8 -4,0 -2,6
The results show the very important influence of the

position of the sample on the energy value. A defect of
centering of 1 mm can be easily avoided by visual
inspection but thls value is too large. A compromise was
set up to 0.5 mm for industrial machines and 0.I mm for
reference machines.
This work on the effect of the geometrical parameters of

impact tests machines on the results of the test, is not
complete ; however, it gave some results which were
included in the subsequent revision of the French
standard.
It is still possible, even when all geometrical parameters

are within the tolerances, that the calibration of the
test machine by the indirect method gives results out of
the standard tolerances. This situation can be easily
explained : all geometrical differences, even though they
are within the tolerance, correspond to energy
differences ; the addition of all these energy differences
may exced the limits.
Therefore, the calibration by comparison with calibrated

specimens is not sufficient. At the same time, it is
necessary to give dimensional measurements of the impact
area of the machine and to verify that the dimensional
differences are all going in the same direction. This
procedure is the only one which will permit an answer to
the question :
"What can we do when the two methods, direct and indirect,

don't lead to the same conclusions ?".
However, because all the work reported here was done with
bending specimens, it will have to be carried out with
Charpy reference specimens as required by the future ISO
and ECISS standards.
Arthur L. Lowe, Jr.
FACTORS INFLUENCINGACCURACYOF CHARPY IMPACTTEST DATA
REFERENCE: Lowe, A. L. Jr., "Factors Influencing Accuracy

of Charpy Impact Test Data," Charpy Impact Test: Factors
and Variables, ASTM STP 1072, John M. Holt, Editor,
American Society for Testing and Materials, Philadelphia,
1990.
ABSTRACT: The Charpy impact test was o r i g i n a l l y developed
as an acceptance test for steel products. However, using
the Charpy test to evaluate the i n t e g r i t y of nuclear
reactor vessels has caused the test data to be subject to
s c i e n t i f i c evaluation. The many small errors in parameters
affecting the results produced r e l a t i v e l y large errors in
any analysis of the data. The sources affecting these
errors are identified and an effort is made to quantify
them. The Charpy impact test is a simple and convenient
test but the end result is engineering data. To provide
data for s c i e n t i f i c evaluations greater control must be
placed on all phases of the testing process.
KEYWORDS: Charpy impact test, Charpy data, neutron radia-
tion, error analysis, controlling parameters, transition
temperature s h i f t , upper-shelf energy
INTRODUCTION
The Charpy impact test was developed as an acceptance test for
steel products. Over time, the test has come to be used for a number
of standard acceptance and regulatory requirements that require
evaluating the test results based on specific s c i e n t i f i c assumptions.
However, l i t t l e has been done to understand the basic test procedure,
or to define which test parameters influence the test results. In
performing and evaluating the Charpy impact testing of hundreds of
unirradiated and irradiated Charpy specimens as part of the Babcock &
Wilcox Owners Group (B&WOG) Reactor Vessel I nt egri t y Program [ I ] , a
Mr. Lowe is an Advisory Engineer (Materials) at B&W Nuclear

Service Company, Engineering and Plant Services Division, Post Office
Box 10935, Lynchburg, Virginia 24506-0935.
54
Downloaded/printed
Copyright 9 1990 by by
ASTM International www.astm.org
LOWE ON TEST DATA ACCURACY 55
great deal has been learned about the sensitivity of results to test
parameters and material conditions.
Variations in the test data can be attributed to parameters
divided into two categories: test-procedure related, including the
test machine i t s e l f ; and material related. Only after the influence
of these parameters, or variables, are understood, or at least their
contribution to data v a r i a b i l i t y recognized, can the Charpy impact
data be reliably analyzed. This paper describes the variables that
influence the Charpy impact d a t a obtained from reactor vessel
surveillance programs and demonstrates how certain of these variables
influence the accuracy of the analyzed data.
BACKGROUND
Charpy impact data is widely used in the nuclear industry to

evaluate the integrity of nuclear reactor pressure vessels. This
usage developed from the fact that the early studies of neutron
radiation on the properties of reactor vessel materials were per-
formed, in part, using the Charpy impact specimen. The size of the
Charpy specimen permitted the inclusion of a greater number of
specimens than would be practical using other types of test specimens
such as the drop-weight test. Besides, the study of fracture
behavior associated with merchant ship failures had established a
correlation between Charpy data and fracture toughness as defined by
dynamic tearing or drop-weight testing [2,3]. The characteristic
response of steels to irradiation as described by the Charpy impact
data curve is shown in Figure I. The Charpy transition temperature
(RT~nT) increases with a corresponding decrease in upper-shelf energy
(USE~ as the result of exposure to neutron radiation. The Charpy
specimen was also adopted as the primary fracture toughness specimen
for inclusion in reactor vessel surveillance programs [4].
The Charpy test was further established as the primary method
for evaluating irradiated material properties by the specification of
the tests in nuclear licensing regulatory requirements IOCFR50,
Appendix G, "Fracture Toughness Requirements [5]" and Appendix H,
"Reactor Vessel Materials Surveillance Program Requirements [6]."
Both of these documents reference the use of Charpy test data to
establish operating limits for the reactor vessel. In addition, the
Nuclear Regulatory Commission (NRC) published Regulatory Guide 1.99,
"Radiation Embrittlement of Reactor Vessel Materials [7]," that
describes a procedure for estimating neutron radiation induced
changes in Charpy properties as applied to various licensing require-
ments. In the original version, bounding conditions were defined,
but the latest version is based on statistical mean values and
margins using standard deviations of the data used to develop the
mean values.
More recently, the NRC published the regulations concerning
pressurized thermal shock containing a screening criterion based on
Charpy data [8]. The latter requirement also contains a criterion of
Charpy mean values, plus a margin for the uncertainty of the Charpy
data, based on the applicability of the supporting data.
UNIRRADIATED , ~ ~ f
,,=,
J_.
E /'!~ &RINDS,, / "I~ "~" ----'--"
I..^o,^TED
Temperature
FIGURE I -- Characteristic response of a Charpy

impact curve to neutron radiation.
Of special interest to the B&WOG Reactor Vessel Integrity

Program are the Charpy data related to the Mn-Mo-Ni weld wire/Linde
80 flux submerged-arc weld metals used to fabricate a large number of
reactor vessels during the late 1960's and early 1970's. This weld
metal exhibits a high sensitivity to neutron radiation, and thus, is
the controlling material for licensing purposes for many reactor
vessels fabricated during this time period. Also, the margin
required to be added to the analysis of these weld metals (per
Regulatory Guide 1.99/2 = 28F or 15.6C) represents a significant
conservatism compared with the actual d a t a values used in any
analytical evaluation. These developments have necessitated a review
of the available Charpy test data from reactor vessel surveillance
programs in an attempt to identify those parameters that influence
the scatter, or uncertainty, in the data. All the Mn-Mo-Ni/Linde 80
submerged-arc data was obtained from appropriate surveillance capsule
reports and each data point and related parameter was verified [9].
These data constitute a Charpy data base that is as free from errors
and inaccuracies, as is possible. However, this does not mean i t is
error free, as w i l l be demonstrated. This paper is based on some of
the results from this ongoing effort.
VaEiBbles Influencing Charpy Data
The materials used in fabricating reactor vessels were reviewed
and the role of the many processing and fabrication parameters, or
variables, were identified. These parameters are listed with general
comments as to the relative importance to influencing the Charpy data
from a given piece of steel or weld metal.
Material History
Base metal (plate or forging)
9 Product form - large variations are attributable to the

differences in product form ( i . e . forging or plate).
9 Manufacturing process large variations because process

controls cleanliness of the steel which in turn influences
Charpy data.
m Fabrication process - large variations because of changes

in properties related to working directions. Such as the
directional properties resulting from r o l l i n g of plate,
9 Heat treatment variations within product form - most

important to control sampling requirements to insure
uniformity of samples.
9 Chemical composition - minor variations within product form

but is important to neutron radiation s e n s i t i v i t y of the
material.
Weld metal
e Welding process large variations and must be controlled

to provide for good intercomparisons.
9 Weld wire chemical composition m i n o r variations within

classification but variations are important to neutron
radiation sensitivity.
9 Weld flux type - can cause large variations between classi-

fication because of its effect on weld cleanliness.
9 Weld flux chemical composition - small variations in

chemical composition w i l l affect metallurgical charac-
t e r i s t i c s of the deposited weld metal, which can affect
Charpy data.
m Welding procedure - variations are primarily a function of

heat input, which in turn, is a function of amperage,
voltage, and welding travel speed.
9 Heat treatment - temperature and time at temperature, plus

cooling rate; pre-heat temperature, intermediate stress
r e l i e f cycles; temperature and time of temperature at the
final stress r e l i e f - all can have a significant affect on
Charpy data,
Specimen Preparation
e Location in material properties within base metals vary

as to location in the final product form; this is not as
important a factor for weld metals as for plates or
forgings.
University of Washington (University of Washington) pursuant to License Agreement. No further repro
9 Specimen fabrication the accuracy of the machined

specimen and notch geometry must be maintained within
specified dimensions.
9 Orientation of notch - orientation of the notch relative to

the major axis of the product form must be kept consistent
with type data to be obtained.
Service Environment
Neutron spectrum - importance not defined but i t is assumed

that the higher the spectrum energy the greater the effect
on both transition shift and upper-shelf decrease.
9 Neutron flux importance is not defined but effect

believed to be similar to that of spectrum.
Neutron fluence the greater the fluence the larger the

transition temperature shift and greater the decrease in
upper-shelf energy.
9 Temperature - the amount of shift, or upper-shelf decrease,

is inversely proportional to irradiation temperature.
Testing
Machine installation must be installed according to

applicable instructions and standards. An improperly
installed machine will produce erroneous data.
Machine calibration must meet applicable calibration

requirements, An uncalibrated machine produces ques-
tionable data!
Operator experience - most important to have an experience

operator who understands the test procedure and has
developed a smooth and consistent testing technique.
9 Operator technique must demonstrate consistent perfor-

mance in conducting the actual testing.
Evaluation
Evaluator experience - experience with the type of material

to be evaluated provides for a better interpretation of the
data.
e Evaluator technique - an understanding of the principles of

statistics is beneficial to interpreting data.
Manual vs, computer plotting computer techniques are

desirable but must address measured data and not function
on idealistic Charpy c u r v e interpretations. Manual
plotting has consistently demonstrated better interpreta-
tion of data.
Statistical Evaluation Method
The s t a t i s t i c a l indicators used to evaluate the role of the

various parameters that influence the Charpy data are the coefficient
of determination and the mean square error, or standard error. The
coefficient of determination is a measure of the goodness of f i t of
the mathematical model to the data with a value of 1.0 being indica-
tive of a perfect f i t . The mean square error is a measure of the
variation of the data about the regression model. The smaller the
standard error of the estimate the smaller the scatter of the
observed data to the predicted values produced by the model. A small
standard error is desirable.
The coefficient of determination or f i t , expresses the propor-

tion of the variance of one variate as given by the other, when the
f i r s t is expressed as a linear regression of the second. I t is a
measure of the usefulness of the terms, other than the constant, in
the model. For simple linear regression the coefficient of deter-
mination measures the proportion of total variation about the mean of
the dependent variable explained by the regression. I t is the
correlation between the observed dependent variable and the predicted
values of the dependent variable. A perfect f i t of the model to the
data would produce a value of 1.0, given no repeated measurements.
The mean square error provides an estimate, based on the degrees

of freedom, for the variance about the regression, based on the data
set and parameters in the model. I f the regression equation were to
be estimated from an i n f i n i t e l y large number of observations, the
variance about the regression would represent a measure of the error
with which any observed value of the dependent variable could be
predicted from a given value of the independent variable using the
f i t t e d equation.
The formula for the determination of the standard error, or

estimate error, from a number of independent contributing factors
whose estimates of error are o I , o 2 ".. ~n is
= ~ i 2 + o22 + ... On2
where
= Total error from all contributing factors
Review of Variables Influencinq Charpy Shift Data
The role of various parameters on Charpy 30 f t - l b transition

temperature s h i f t was f i r s t observed during the development of new
correlations for the Mn-Mo-Ni/Linde 80 weld metal to predict the
change resulting from neutron radiation [7]. At the time a number of
correlations were being tried that produced different coefficient of
f i t values. However, for select sets of data the coefficient of f i t
values remained relatively unchanged but the standard deviation for
the f i t t e d curves would vary significantly. Two typical sets of data
are presented in Table I.
University of Washington (University of Washington) pursuant to License Agreement. No further
These sets of data demonstrate that by carefully defining the

selection of data not only was the coefficient of f i t changed, but
the standard error can be s i g n i f i c a n t l y changed. This observation
i n i t i a t e d an evaluation as to the role of the various parameters
identified in the previous section and t h e i r effect on the standard
error. Recognizing that a larger number of factors influence the
standard error than could be identified, or quantified, a l i s t i n g was
prepared that identified the most probable parameters. A probable
error value was assigned to each based on either recognized error
values for the data, or an estimate of the probable error was
calculated from variations in the parameter. The parameters and
t h e i r corresponding estimated error are shown in Table 2.
Using the data in Table 2, i t is shown that a l l the estimated

errors combined produce a standard error of approximately 20.9F
(11.6C) which approximates the larger values shown in Table I. A
lower standard error value of I0.7F (5.9C) is calculated using items
i , 3 and 4 in Table 2 and not including items 2 and 5. This lower
value closely approximates the lower values given in Table I. The
exclusion of the two items (Items 2 and 5) is possible because all
the fluence analyses were performed using the same procedure and the
data interpretation was performed by the same person. Therefore, i t
is assumed that these errors are constants and did not contribute to
the standard errors. This is not to imply that there is no error
from these parameters but in the analysis of this selected data base
they probably contribute a similar error to each data analysis. The
reason that the values for Trials A-I and B-I cannot be reduced is
that the early fluence analysis was performed using an out-dated
procedure which probably had a standard error s i g n i f i c a n t l y larger
than the one l i s t e d in Table I .
Another group of Charpy data analysis results is presented in

Table 3. Two important observations can be made from these data.
First, in weld wires A and B, the correlations based on the i r r a -
diated 30 f t - l b d a t a produce lower standard errors t h a n those
correlations based on 30 f t - l b s h i f t data. This difference implies
that errors in the i n i t i a l values influence the correlation error.
This trend appears true whether two or four laboratories (test sites)
provided the irradiated data.
Weld wire C poses a different relationship between the test

sites when weld metal test results are separated into sub-groupings.
When the same test site produced both the i n i t i a l data and test data,
good coefficients of f i t and small standard errors were obtained.
However, when the same weld metal tested by two sites was evaluated
the coefficient of f i t was reduced and the standard error increased.
This implies the enclusion of a larger error which in this case may
be related to the different testing techniques of the two test sites.
The inclusion of three test sites expanded the standard error and
produced mixed results as to coefficient of f i t . This evaluation
further demonstrates the interactions of both test procedure para-
meters and test laboratories.
University of Washington (University of Washington) pursuant to License Agreement. No furthe
TABLE I -- Selected Charpy transition shift correlation data

showing variations in coefficient of f i t and standard
error of the estimate (Mn-Mo-Ni/Linde 80 weld metal).
Trial Welds Fluence

Number Selected Range R2a Syxb, F(C)
A-I B&W All 0.86 23.9(13.3)

2 B&W >3E18 n/cm2 0.90 12.8(7.1)
3 Non-B&W All 0.83 19.3(10.7)
4 Non-B&W >3E18 n/cm2 0.83 21.6(12.0)
B-I B&W All 0.86 22.0(12.2)

2 B&W >3E18 n/cm2 0.96 10.9(2.8)
3 Non-B&W All 0.70 22.1(12.3)
4 Non-B&W >3E18 n/cm2 0.81 19.8(11.0)
aCoefficient of f i t .
bstandard error of the estimate.
TABLE 2 -- Estimated error values for parameters affecting

the accuracy of Mn-Mo-Ni/Linde 80 weld metal
Charpy 30 f t - l b transition temperature data.
Estimated
Item Parameter Error, F(C) Comments
Charpy test temperature* • 5(3) Assumes oil baths and

minimum transfer time
Neutron fluence • Normal accepted error

for calculations
Irradiation temperature • 5(3) Assumes steady-state

operations, no transi-
tions in power
Chemical composition • 8(4) Calculated value based

on normal variations in
compositions
Charpy data interpretation • Assumes no experience

with materials data
*Includes the machine error.
University of Washington (University of Washington) pursuant to License Agreement. No further repr
Review of Variables Affectinq Charpy Upper-Shelf Enerqy
An e f f o r t was made to evaluate Charpy upper-shelf energy data in

the same way that the Charpy transition temperature s h i f t data was
evaluated. This study did not produce any clearly defined effects of
the various parameters with the exception of chemical composition.
This effect has been recognized from the earliest studies. However,
an interesting observation can be made from a plot of data from a
group of weld metals fabricated from the same heat of weld wire. The
unirradiated and irradiated data is shown in Figure 2. These data
exhibit a variation in responses to testing at various laboratories.
Certain of the data sets follow the expected trend while others
exhibit sharp changes. While i t is d i f f i c u l t to assess the cause of
these variations, they are probably the result of either inadequate
testing procedures, or related to the installation of the impact test
machine. The variations in the data vary between 5 to 10 f t - l b s , not
only between different testing laboratories, but between different
data sets tested by the same laboratory. Unfortunately, there are no
standard test specimens for calibrating impact testers on the Charpy
upper-shelf energy region such as are available for the Charpy
transition region. Standards would provide a means for cross
referencing upper-shelf energy test results.
The interaction of the i n i t i a l upper-shelf energy on the

interpretation of irradiated data is demonstrated by comparing the
decrease in Charpy upper-shelf energy to the irradiated Charpy upper-
shelf energy for a group of weld metal. Figure 3 shows the effects
of irradiation on the upper-shelf as a percent of the i n i t i a l , or
unirradiated, value plotted as a function of the fluence. Based on
the power reactor data, the coefficient of f i t is 0.47 and the
standard error is 3.4%, or approximately 10 percent of the mean
decrease value. The same data is shown in Figure 4 as the irradiated
Charpy upper-shelf energy as a function of fluence. Again, based on
the power reactor data, the coefficient of f i t is 0.90 and the
standard error is 1.3 f t - l b s , or approximately 2.5 percent of the
mean irradiated value. This comparison implies that the error of the
i n i t i a l upper-shelf energy values may be significantly larger than
the irradiated values. This evaluation of upper-shelf energy change
i l l u s t r a t e s a case of basing the evaluation of well characterized
data on data that is not of equal quality and thereby producing less
than desirable results.
Recommendations to Minimize Errors
The review of Charpy data presented demonstrates that relative

small standard errors can be achieved in Charpy data by closely
adhering to an established testing procedure. However, to achieve
the ultimate in Charpy evaluation error, all the parameters identi-
Fied as affecting the data must be controlled as close as practical.
In the case of surveillance program testing, controlling the final
test parameters will not reduce the errors introduced in the i n i t i a l
data set. Since, the final evaluations are often based on the
eO
Id..
.,... ~
.J~ COO',
X
(1)
~n
R--4-~
0 m
S.- ~,"~ r
r 9 ~ 1 7 6
~o
V-=
e.- o
~ t=- "1::1
(1)
rO r--- i ! I I I I
r",
:>
l.J.d
~ e--
e - I I.-- ~
"~ o
e-.
9~-'1 -I~ -I~ N ~

,.-.~ ~1"~ Z Z Z ~ Z Z
t~
E
~
~~
.,.; -~
O
C~I r r ,"-I ,--I
e,. r E R-- S-
O 0
Zr---
.~ e--
Q;
e'- ~)
I
I
0 ~
~
I I
.,-I
I"--
go , (fj, , 120
8o 110
Lab 1 100
3 70 - ,,~,,j~-Lab 2
c 90
m-
~o 60 - . " \ ~ , ~ ' ~ " ~ Lab 5 Lab 3 80
e- ~L~c"---Z:~
---'~'~-.==~-- ---m Lab 3
7O J
~, 5o
\ "Lab 2 LaD / =. _=
--~,.~.~, . . . . "ZLaO ..l .,:"K~ Lab 4
60
.Io 40
Lab2 50
:~ 30 40
20 Ir"l I I I I 30
0J'J7EI7 1E18 3E19 5E18 JEt9 5E19
Fluence, n / c r n 2 (E > 1MeV)
FIGURE 2 -- Irradiated Charpy upper-shelf energy

as a function of fluence.
8O I I l I
~E
70
r
(1. 9 Power Reactor Data
~ 6o oTest Reactor Data
Coefficients for Power Reactor Data
~ 50 Rz = 0.47 Std, Oev. = 3.4
a
~ 4o
~ 3o 2Std, Dev. = 6.8
"0 20 0,, 9 ,,,, ,=m
-~ 1o
o
- o I I n I I
5E17 1El8 3E18 5E18 1El9 5E19
Fluence, n / c r n z (E > l M e V )

decrease as a function of fluence.
analysis of two sets of data obtained at different times and by

different laboratories, the best results (minimum errors) can only be
obtained by both testing laboratories closely controlling a l l testing
parameters that can affect data.
80 I t I I
ioo
70
90
"T 60 80
uJ" 70
u) 50
60
2Std. Dev, = 2.6ft-lb
'-
(1: 40 (3,5J)
r 50 J
O
-o 30 40
30
20 9 Power R e a c t o r D a t a
P.
o Test Reactor Data
20
tO Coefficients for Pow er R e a c t o r Data
I0
R 2 = 0.90 Std. Dev, = ].3ft-lbs(1.SJ)
0 I I,, J I -- o
5Et7 1E18 3E18 5E18 1E19 5E19
Fluence, n/cm 2 (E >IMeV)

as a function of fluence.
ACKNOWLEDGEMENTS
This paper reports work done in support of the Babcock & Wilcox
Owners Group Reactor Vessel I nt egr it y Program. The author acknow-
ledges and expresses appreciation for t h e i r support and t hei r
permission to report program results. The support of J. W. Pegram in
providing the s t a t i s t i c a l analysis is acknowledged with appreciation.
REFERENCES
[I] Lowe, A. L., Jr., Moore, K. E., and Aadland, J. D., "Integrated
Reactor Vessel Materials Surveillance Program for Babcock &
Wilcox 177-FA Plants," Effects of Radiation on Materials:
Twelfth International SymDosium,_ASTMSTP 870, F. A. Garner and
J. S. Perrin, Editors, American Society for Testing and Materi-
als, Philadelphia, 1985, pp. 931-950.
[2] Stout, Robert D., and W. D'Orville Dory, We l d a b i l i t ~ of Steels,
Welding Research Council, New York, 1953.
[3] C o n t r o l o f S t e e l C o n s t r u c t i o n t o Avoid B r i t t l e Failure, M. E.
Shank, E d i t o r , Welding Research C o u n c i l , New York, 1957.
[4] American Society for Testing and Material, Standard Practice for
Conducting Surveillance Tests for Light-Water Cooled Nuclear
Power Reactor Vessels, E185-82, July 1982.
[5] Code of Federal Regulation, T i t l e 10, Part 50, Fracture Tough-
ness Requirements for Light-Water Nuclear Power Reactors,
Appendix G, Fracture Toughness Requirements.
[6] Code of Federal Regulation, T i t l e 10, Part 50, Fracture Tough-

ness Requirements for Light-Water Nuclear Power Reactors,
Appendix H, Reactor Vessel Material Surveillance Program
Requirements.
[7] U.S. Nuclear Regulatory Commission, Radiation Damage to Reactor

Vessel Material, Requlatory Guide 1.99, Revision 2, May 1988.
[8] Code of Federal Regulations, T i t l e 10, Part 50.61, Fracture

Toughness Requirements for Protection Against Pressurized
Thermal Shock Events.
[9] Lowe, A. L., Jr., and Pegram, J. W., "Improved Correlations for
Predicting the Effects of Neutron Radiation on Linde 80 Sub-
merged-Arc Weld Metals," Effects of Radiation on Materials:
I4th International Symposium, ASTM STP I046~ Volume I I , N. H.
Packen, R. E. Stoller and A. S. Kumar, Editors, American Society
for Testing and Materials, Philadelphia, 1989.
Toshio Naniwa, Masayuki Shibaike, Megumu Tanaka, Hiroshi Tani,
Kenichiro Shiota, Namio Hanawa, and Toshiaki Shiraishi
EFFECTS OF THE STRIKING EDGE RADIUS ON THE CHARPY IMPACT TEST
REFERENCE: Naniwa, T., Shibaike, M., Tanaka, M., Tani, H., Shiota,
K., Hanawa, N., and Shiraishi, T., "Effects of the Striking Edge
Radius on the Charpy Impact Test," Charpy ImDact Test: Factors and
~ari~bles, ASTM STP i072~ John M. Holt, Ed., American Society for
Testing and Materials, Philadelphia,. 1990.
ABSTRACT: Effects of the striking edge radius of the Charpy impact

test on the absorbed energy and the abrasion of the striker were
investigated. The absorbed energy of the 8mmR was higher than that
of the 2 m m R w h e n the absorbed energy was above 200N-m. The cause
of this difference is that the bending deformation of the test
specimen and the friction between the anvil and the test specimen
are large in case of the 8mmR. The abrasion of the 8mmR striker
was extremely large and it is too difficult to maintain the radius
dimension according to A S T M E 23.
KEYWORDS: Charpy impact test, Charpy apparatus, pendulum striking

edge, absorbed energy, abrasion
INTRODUCTION
The Mechanical Testing Subcommittee found the difference in the

absorbed energy of the Charpy impact test between the 8mmR and the 2mmR
striker (1). The absorbed energy of the 8mmR striker was higher than
that of the 2mmR in case of 400 N/mm 2 class steel. This relation was
reversed in case of 800N/mm 2 class steel.
The purpose of this paper is to investigate effects of the strikln@

edge radius on the absorbed energy and the abrasion of the striking
edge in addition to our previous investigation.
T. Naniwa and M. Shibaike are with Nippon Steel Corp., M. Tanaka is

with NKK Corp., H. Tani is with Sumitomo Metal Industries, Ltd., K.
Shiota is with Kobe Steel, Ltd., N. Hanawa and T. Shiraishi are with
Kawasaki Steel Corp., All the members belong to The Mechanical Testing
Subcommittee, The Joint Research Society of The Iron and Steel
Institute of Japan, 9-4, Otemachi-l-Chome, Chiyoda-ku, Tokyo, Japan.
67
Copyright9 1990
University by ASTM(University
of Washington International www.astm.org
EXPERIMENTAL
Material
Grade and class: Carbon and low-alloyed steel. Values of the

tensi--~estr-~ngt-h~re from 400 to 600 N/mm 2.
Absorbed energy: From i00 to 500 N-m
Test specimen: ASTM E23 Type A
Instrumentation
ASTM type Charpy impact testing machines (The striking edge radius
is 8mm) and JIS type Charpy impact testing machines (The striking edge
radius is 2mm) were used.
Dynamic load-time measuring method during the Charpy impact test:

Two strain gages were glued on both sides of the striking edge. It was
possible to record load during the Charpy impact test. The schematic
diagram is shown in Fig. i.
Static bending device: We made the same devices as strikers and

anvils of the Charpy impact test for the static bending test. Devices
are shown in Fig. 2. The static bending tests were performed by these
devices and the compression testing machine. The load and the
displacement of the striker were measured by the compression testing
machine and recorded on the X-Y recorder.
X-Y recorder
Hammer Strain gage
, ~ mm Load=k.~
Strain gage ~: Strain
k:Experimental
coefficient
Fig. i Dynamic load measuring method for the Charpy impact test.
NANIWA ET AL. ON STRIKING EDGE RADIUS 69
Guide
Fig. 2 Devices for the static bending test.
RESULTS
Comparison of Charpy Impact Test Results between the 8mmR and the 2mmR
Striker
Charpy impact test results are shown in Fig. 3 (a)-(d). No differ-

ence was recognized in the absorbed energy between the 8mmR and the
2mmR less than 200N-m energy. However, the absorbed energy of the 8mmR
was higher than that of the 2mmR above 200N-m energy. The higher the
absorbed energy was, the larger the difference of absorbed energies
between the 8mmR and the 2mmRwas. No difference was recognized in
shear fracture, lateral expansion, and transition temperature.
500 l I I
400
E
E
Oo o ~
300 /
>
o
n-
Lu
E 200
a
u,l
en
r
0 i00 ~
m
ii1
/
O I I I I
0 I00 200 300 iOO
ABSORBED ENERGY BY 2mmR, N-m
Fig. 3-(a) Comparison of the absorbed energy between the 8mmR
and the 2mmR
IOO
E
E
00
>
n~
uJ
n~
D )f?o ~
<
n-
U.
n~
<
uJ
-r
o9
20 I I I I , I I i
2O 5O I00
SHEAR FRACTURE BY 2mmR, %

Fig. 3-(5) Comparison o f th e shear f r a c t u r e between the 8n~LR
and the 2 ~ .
>
E
E
o0
z
o
4
o
o oo
y
~
o
z
<
a_
X 2 ~%~o ~oo
< O ~ O0 O O
oo
I-
<
,-I
i I
2 5 4
LATERAL EXPANSION BY 2mmR, mm
Fig. 3-(c) Comparlson of the lateral expansion between the 8mmR
and the 2mmR.
i 40p
FAT]-, degree C"
500 I I
/
i 8 mmR
I
40s I
E I
/
o
z ~ 8
30s 2 mmR
,,,
u.l
r~
u.I
n-200
m o
<
100
o-14o-12o-loo .o & 14o _~ l

TEST TEMPERATURE, degree C
Fig. 3-(d) Comparison of the transition curve between the 8mmR
and the 2n~..
Measurement of Fracture Process
An example of dynamic load-time curves during the Charpy impact

test is shown in Fig. 4, and an example of load-displacement curves
during the static bending test is shown in Fig. 5. A comparison of
Figures 4 and 5 indicates the following.
(i) Load-time curves and load-displacement curves in same test

conditions resemble each other.
(2) Shapes of test specimens after test resemble each other, too.
If the static bending energy is defined as the area surrounded by

load curve and displacement-axis, it is expressed by the following
equation.
a
Ks = I f(x).dx (i)
o
where
Es= static bending energy
x = displacement
f(x) = load at displacement(x), function of x
o = displacement at start point
a = displacement at end point
72 C H A R P YIMPACT TEST: FACTORS AND VARIABLES
D
<
O
0~
2mmR
I 1 ~ .... 400 N/mm2steel
TIME
Fig. 4 Examples of the dynamic load vs. time curve
obtained by the Charpy impact test.
Z
,o
15
5
8mmR
400 N/mm2steel
. O
Q
,,~ 15 2mmR
0 400 N/mm2steel
d
I0
0
DISPLACEMENT
Fig. 5 Examples of the load vs. displacement curve obtained
hy the static bending test.
Comparison between the Charpy absorbed energy and the static

bending energy is shown in Fig. 6. Both energies were strongly
correlated. From facts described above, we may conclude that cesults
of the static bending test are applied to explain fracture process of
the Charpy impact test.
500 I' ' 1 I I /

E I -'Striker ~ /
Z
400 -I ~ I 2ram Rt I / ~ ~ -
>_"
(,.9
LIJ 300
Z
W
~ 0f 0 --
(.9
~--
Z
200
Z
W
m
_o 100
<
I-
0 9 I 1 ,,1
0 100 200 300 400 ;00
CHARPY ABSORBED ENERGY, N-m

Fig. 6 The relation between the Charpy absorbed energy
and the static bending energy.
Results of the static bending test are shown in Fig. 7 and

Photo. I. Results may be summarized as follows.
20: I I
High Energy Level
8mm R
Striker Energy, N-m
15 a b
8mm R 459 395
10
", / 2mm R 340 248
0
20
Medium Energy Level
Z 15 Striker Energy, N.m

,,I a b
8 mm R 281 234
10 ~~8mm R
.< 2 mm R 232 204
0
._1 5 2mm R/~", "~
0
20
Low Energy Level
Energy, N. m
15 Striker
a b
8 mm R 78 97
10
2 mm R 92 87
00 10 20 3O
/a Charpy absorbed energy t
DISPLACEMENT, mm / b Static bending energy /
Fig. 7 Load vs. displacement curves by the static bending test.
Photo i. Observation of bending stages in static bending test

(High Absorbed Energy Steel).
Low absorbed energy steel:

(i) Load-dlsplacement curves of the 8mmR and the 2 m m R w e r e similar.
(2) The maximum load of the 8mmR was higher than that of the 2mmR.
The difference in loads was about i kN.
High absorbed energy steel:

(i) Load-displacement curves changed into complex ones. The second
load peak appeared at the point of about 15mmdisplacement in medium
energy steel in case of the 8mmR. The second and the third peak
appeared in high energy steel. The third peak appeared at the point of
about 25mm displacement.
(2) The second and the third peak load of the 8 m m R w e r e higher than
those of the 2mmR. The second peak load of the 8 m m R w a s about 1.5
times that of the 2mmR. The third peak load of the 8 m m R w a s about 4
times that of the 2mmR.
The Abrasion of the Striker
Changes of the striking edge dimensions with tests are shown in

Fig. 8. Sections of the plaster molds out of the strikers are shown in
Photo. 2. There was a contrast between the 8 m m R a n d the 2mmR. The
radius of the 8mmR markedly diminished. If we observe AS~M E 23
strictly that the dimensional tolerance is within ~0.05mm, we must
exchange the 8mmR striker within 1,000 tests. On the contrary the 2mmR
striker was little worn after 30,000 tests.
~8
'~ 8R 2R 8 mmR
R 9 o
==" W 9 =
~ r 9
k- 6
-
U.
0 4 . . . . . . . Wl
2 mmR
i
~UM OF ABSORBED ENERGY
2 i 4 5 (Xt0' N.m) R~
LO
8 1'6 (Xt03 pieces)

THE NUMBER OF TEST PIECES (8mmR)
I i I
0 16 31 (xl0 3 pieces)
THE NUMBER OF TEST PIECES (2mmR)
Fig. 8 Change of striking edge dimensions with tests.
NANIWA ET AL ON STRIKING EDGE RADIUS 77
8 mmR 2 mmR
t-
o
rn
I
I-
<
Photo. 2 Cross sections of striking edges.
DISCUSSION
The Difference in the Absorbed Energy
We would like to discuss the cause why the absorbed energy of the
8mmR was higher than that of the 2mmR in high absorbed energy steel.
Load-displacement curves had three peaks in case of high energy steel.
A typical load-displacement curve is shown in Fig. 9.
The first peak: It is assumed that this peak corresponds to the

initiation of the crack at the bottom of the notch. It showed the
maximum load. That of the 8 m m R w a s slightly higher than that of the
2mmR. Owing to the fact that the contact length of the 8mmR between
thestriking edge and the specimen was longer than the 2mmR, we proved
it by using elastic dynamics (1) .
The second peak: From results of the static bending test, it is

assumed "the second peak corresponds to the bending deformation occurred
near the shoulders of the specimen near the notch. Please refer to
Photo i. The deformation of the 8mmR specimen was larger than that
of the 2mmR. For your information, the relation between the striking
edge radius and the plastic deformation of the specimen is shown in
! : Maximum load point

II : T u p s h o u l d e r b e n d o c c u r r e n c e
I l l : A n v i l sli ) o c c u r r e n c e
Low Energy High Energy
9 ] L
s
o<
d
mm R 9
8 mmR
2 mm~'~_ ..... .
DISPLACEMENT, mm DISPLACEMENT, mm
Fig. 9 Three energy regions of the load vs. displacement
curve in energy.
r
E
E [00 t Charpytest t
Static bend
z
O Area=(~B_) x t C+D)
I-
<
IT V NOTCHED SURFACE ~,
O
Ii
Iii A ~B
1:3
O
}--
O9
< I I I 1
,_1
n
IJ_
0
<
UJ
IT
< I I I I I
0 2 3.5 5 6.5 8
RADIUS OF STRIKER, mm
Fig. i0 The relation between the striking edge radius and
the deformation of the specimen.
Fig. i0(i). The second peak enlarged in proportion to the absorbed

energy of the specimen, too.
The third peak: It is assumed that the third peak is due to the
friction between anvils and the test specimen. The friction did not
occur in case of low absorbed energy steel. The friction occurred in
case of high absorbed energy steel. And the friction of the 8 m m R w a s
larger than that of the 2mmR.
The Abrasion of the Striker
As shown in Fig. 8 and Photo. 2, the abrasion of the 8mmR striker

was larger than that of the 2mmR. It is considered that the 8mmR
striker was worn more than the 2mmR because of more bending deformation
and more friction of the specimen.
The Radius Size of the Striking Edge
As mentioned above, it is possible to say that the load-

displacement curve having only the first peak corresponds to energy
about the initiation and the propagation of the crack. The first peak
is the most important peak for the Charpy impact test. On the other
hand the second and the third peak are caused by the deformation and
the friction respectively. These energies should not be included in
the Charpy impact test. The absorbed energy of the 8mmR includes more
extra energy than that of the 2mmR.
The abrasion of the 8mmR striker is very large. Since the

dimensional tolerances change faster, the striker must be changed more
frequently.
From the above experimental results, it is concluded that the 8mmR

striker is not as desirable for the Charpy impact test.
CONCLUSIONS
We investigated effects of the striking edge radius on the absorbed

energy and the abrasion of the striker. The summary is as follows.
(i) The absorbed energy of the 8mmR striker was higher than that of
the 2mmR when the absorbed energy was above 20(~q-m. No difference was
recognized in the shear fracture, the lateral expansion, and the
transition temperature.
(2) The difference in the absorbed energy between the 8mmR and the
2mmR is caused by the bending deformation of the test specimen and the
friction between the anvil and the test specimen.
(3) The Abrasion of the 8mmR striker was larger than that of the
2mmR. We must exchange the 8mmR striker within 1,000 tests in order to
satisfy the dimensional tolerance of striker radius. On the contrary
the 2mmR striker was little worn after 30,000 tests.
ACKNOWLEDGMENTS
The authors would like to acknowledge the cooperation and the

assistance of the others in The Mechanical Testing Subcommittee, The
Joint Research Society of The Iron and Steel Institute of Japan.
REFERENCE
[I] M. Etoh et al., "Investigation of the Effects on Charpy Impact

Characteristics by Shape of Pendulum Striking Edge," ASTM A01,13
Subcommittee on testing (seminar), Louisvill K.Y. May i0, 1983.
The Specimen: Notches
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
Robert D. Koester and Steven E. Barcus
EVALUATION OF FABRICATION METHOD FOR MAKING NOTCHES FOR

CHARPY V-NOTCH IMPACT SPECIMENS
REFERENCE: Koester, R. D., and Barcus, S. E. "Evaluation of

Fabrication Methods for Making Notches for Charpy V-Notch
Impact Specimens," Charpy Impact Test: Factors and Variables:
ASTM STP 1072, John M. Holt, American Society for Testing and
Materials, Philadelphia, 1990.
ABSTRACT: The effect of preparing the notches for Charpy V-

Notch impact specimens is evaluated. Two methods of
preparation are used for this evaluation. They are grinding and
broaching. Impact specimens with two energy levels were used:
low energy, 15 J (11 ft.lbf) and high energy, 94 J (69 ft.lbf).
Twenty-four specimens of each level were made with ground
notches and with broached notches. Each was tested with a
calibrated impact tester. The material used for the specimens
was AlSl-4340 alloy steel. The notch characteristics were
documented by the following methods: notch angles, notch radii,
surface conditions, and near-surface microstructure. Other
parameters documented for this study were the ligament
thickness, specimen width, and specimen length. In addition, test
conditions and results are provided as follows: test temperature,
absorbed energy, and lateral expansion. Besides tabular data,
photographs give views of the notch surface. Evaluation of the
results indicated both types of notches gave equally consistent
values, but the ground and broached values differed in their
averages. Possible causes for this variation were differences in
ligament sizes and bottom of notch radii, surface tears and
shallow microstructural deformation in the case of the broached
notches.
Mr. Koester is an engineering manager for Southwestem Laboratories,

Post Office Box 8768, Houston, -rx 77249; and Mr. Barcus is president of
Sure Tool & Engineering, 302 East Pleasant Street, Churnbusco, IN 46723.
83
Copyright 9 1990byby ASTM International
Downloaded/printed www.astm.org
84 CHARPY IMPACTTEST: FACTORSAND VARIABLES
KEYWORDS: Charpy V-notch, impact test, notch machining, AISI-

4340 alloy steel, lateral expansion and toughness.
BACKGROUND
The key part of fabricating a Charpy V-notch impact specimen is the

method of producing its notch. The other parts of the fabrication cycle
must be capable of producing consistent dimensions, squareness, and
reasonable surface finish; but the aspect of the machining that can have
the most telling effect on the results is the notch itself. For this reason, it
was determined that a study aimed at identifying any differences from
specimens produced by different notching methods would be a worthwhile
endeavor. Of the three most common ways to produce the notch,
broaching, grinding, and milling, the former two were chosen for this
comparison.
Since the results of this nature are governed by statistics, the use of
24 specimens for each condition being evaluated was felt necessary. This
number of samples permits a statistical analysis to be performed that gives
validity to any differences that are observed. The results were measured
in the English system and converted to the International System of Units.
Previous studies [1] have shown that impact results are reproducible.
This study clearly indicates that consistent test specimens and calibrated
test machines give this reproducibility. Other studies [2] have addressed
the aspect of notch depth and notch radii.
EXPERIMENTAL METHOD
Material
The material was modified AlSl-4340 that was processed at the same
time that a lot of standard specimens being made for impact machine
calibration purposes was processed. The material is, therefore, the same
except for the notching that is purchased from the Army Materials
Technology Laboratory. This source of material ensured that no variables
from the material or heat treating aspects of processing would enter this
evaluation.
Machining
The specimens were fabricated, i.e. sawed, milled, and ground, with
a lot of standard specimens being made for impact machining calibration
purposes. The finished Charpy specimens are, therefore, the same except
KOESTER AND BARCUS ON FABRICATION METHOD 85
for the notching that is purchased from the Army Materials Technology
Laboratory. This fabrication approach also ensured that no variables from
the fabrication methods used would enter this evaluation.
Notching
The specimens were notched with the following equipment. The

ground notches were made using a Mitsui, Model 6-12 Grinder equipped
with an Engis Diaform Wheel Dresser. The grinding wheels were Norton
No. 38A60K8VBE, and they were 17.8 cm (7") x 0.64 cm (1/4") x 3.18 cm
(1-1/4") in size. Grinding is done in a normal pJunging style to the
appropriate depth. The number of wheel dressings will vary; however, the
wheel is dressed just prior to grinding the last 0.076 mm (0.003") of notch
depth. The grinding operation is done with the specimen flooded in
coolant.
The broaching was accomplished on a Blacks Equipment, Type

CNB14 Broacher. It is a motorized broach. The broach cutter is No. NBT-
VT .010 that has 56 teeth. The notch is cut with one pass of the broach
cutter while adding a liberal amount of cutting oil to the broach cutter prior
to the cut. The broach used for the notching had been used previously to
notch about 1000 specimens.
Testing
The Charpy specimens were tested on a Tinius Olsen Impact Machine,

Model No. 64. This machine has as its capacity 358 J (264 ft.lbf). The
machine had been calibrated using standard Charpy specimens in
accordance with ASTM E 23 [3] procedures on January 27, 1989. The
results of that calibration showed that all aspects of the testing procedure
and equipment met the requirements of ASTM E 23. The results compared
to the standard values given by Army Materials Technology Laboratory
appear in Table 1. It can be seen that this machine is within the calibration
requirements of ASTIVl E 23. It may be noted that it is on the low side of
the nominal value established by the Army Materials Technology
Laboratory.
The Charpy tests conducted for this evaluation were performed on

March 4, 1989, which was a short time after the calibration described
above. The tests were performed per ASTM E 23 procedures at -40~
(-40~ The specimens met all of the above dimensional requirements for
an ASTM E 23 Type A specimen. The cross sections were 10.0 mm
(0.394") x 10.0 mm (0.394"). The ligament sizes were 7.98 mm (0.314") to
8.03 mm (0.316"). The ground notches typically had ligament sizes of 7.98
mm (0.314") while the broached notches, 8.03 mm (0.316"). The notches
had angles of 46 ~ with the ASTM E 23 required bottom radii of 0.25 mm
(0.0107") being 0.229 mm (0.009") for the broached notch and 0.259 mm
(0.010") for the ground notches. The surface finish on the notched face
and the face opposite from the notch was observed to be within the
requirement of two microns (63 micro-in) per ASTM E 23. Comparison on
an optical compartor of the ground and broached notches to a certified
outline of the Charpy V-notch requirements of ASTM E 23 indicated that
both notches met all requirements.
TABLE 1-CalibralJon of "nnius Olsen impact testing machine
Source I_ow-Energy High-Energy

of Charpy Specimens Charpy Specimens
Value Absorbed Energy Absorbed Energy
J (tt.ll~) J (ft-lb 0
Measured on standard
AMTL specimens by SwL 15.7 (11.6) 99.0 (73.0)
Published as nominal
energy value by AMTL 16.7 (12.3) 101.1 (74.6)
Allowable range of 15.3 (11.3) 96.1 (70.9)

values to meet ASTM to to
E 23 requirements 18.0 (13.3) 106.2 (78.3)
Test Results
Tables 2 and 3 show the absorbed-energy values and lateral-

expansion results of the two energy levels of Charpy specimens that were
tested. The low-energy specimen results appear on Table 2 and the high-
energy level on Table 3.
TABLE 24mpact results on low-energy Charpy impact specimens

with broached or ground notcher
Absorbed Enemy Lateral Expansion

Value No. of Value No. of
J ~t.U~r) Specknens mm (has) Specimens
Ground Notches
13.6 (10.0) 9 0 (0) 20
14.2 (10.5) 10 0.02 (1) 1
14.9 (11.0) 3 0.05 (2) 3
15.6 (11.5) 2
Broached Notches
13.6 (10.0) 5 0 (0) 21
14.2 (10.5) 1 0.02 (1) 3
14.9 (11.0) 10
15.6 (11.5) 5
16.3 (12.0) 3
TABLE 3---Impact results on high-energy Clmr~ impact specimens

wi~ broached or gn:xJrr notches.
Absorbed Energy Lateral Expansion

Value No. of Value No. of
J (lt-lb0 Specimens mm (mils) Specimens
Ground Notches
88.1 (65) 1 0.89 (35) 2
89.5 (66) 3 0.91 (36) 1
90.8 (67) 1 0.94 (37) 3
92.2 (68) 4 0.97 (38) 8
93,6 (69) 5 0.99 (39) 5
94.9 (70) 6 1.02 (40) 1
96.3 (71) 1 1.04 (41) 4
97,6 (72) 0
99.0 (73) 3
Broached Notches
85.4 (63) 1 0.84 (33) 1
86.8 (64) 1 0.86 (34) 3
88.1 (65) 3 0.89 (35) 3
89.5 (66) 5 0.91 (36) 5
90.8 (67) 2 0.94 (37) 4
92.2 (68) 4 0.97 (38) 5
93.6 (69) 3 0.99 (39) 3
94.9 (70) 4
96.3 (71) 1
A statistical analysis of these results was performed in order to define

any difference. The results of this analysis are shown as Tables 4, 5, and
6 for the low-energy absorbed-energy values, the high-energy absorbed-
energy values and the high-energy lateral-expansion values, respectively.
The low-energy lateral-expansion values were not included in the evaluation
since they were, by and large, zero values.
TABLE 4--Statistical analysis of a b s o d o e d ~ n e ~ values from low-energy Charpy

impact specimens with broached or ground notches.
Parameter Ground Notches Broached Notches
Number of specimens, n 24 24
Average of n specimens,
s J (ftolbf) 14.2 (10.5) 14.9 (11.0)
Standard deviation for

n specimens, O'n
J (ft.lbf) 0.7 (0.5) 0.8 (0.6)
Percent variation from

average, an/s ~ 100 4.8 5.5
The average for the low-energy absorbed-energy value is 0.7 J (0.5

ftolbf) higher for the broach notched specimens than for the ground
notched specimens. The standard deviation for the low-energy absorbed-
energy values is 0.1 J (0.1 ftolbf) higher for the broach-notched specimens
than for the ground-notched specimens (see Table 4). This difference
reversed itself for the high-energy absorbed energy values. The average
for the high-energy absorbed-energy value is 2.4 J (1.7 ft.lbf) lower for the
broach-notched specimens than for the ground-notched specimens. The
standard deviation is the same for both the ground- and broach-notched
specimens in this case (see Table 5). Comparison of the lateral-expansion
results for the high-energy level specimens indicates the same situation as
for the average absorbed-energy values for this range. The average for
the broach-notched specimens is 0.04 mm (1.8 mils) lower than the
average for the ground-notched specimens. The standard deviations are
again the same (see Table 6).
TABLE ~ ana~ts~ of absod0ed energy vaZuesfrom high-

energy Charpy impact specimens with broached or ground notche~
'~ J (ftolbf) 93.6 (69.0) 91.2 (67.3)
n specimens, on
J (ft.lbf) 2.8 (2.2) 2.8 (2.2)

average, on/~ x 100 3.2 3.2
TABLE 6-Statistical analysis of lateral-expansion results from high-

energy Cherpy impact specimens wi~ broached or ground notcher
,~ mm (mils) 0.97 (38.3) 0.93 (36.5)
n specimens, Gn
mm (mils) 0.04 (1.7) 0.04 (I/7)

average, an/s 9 100 4.4 4.7
In order to characterize the condition of the notches, the samples were

examined by stereomicroscopy. Figures 1 and 2 show the bottoms of the
notches. These views show continuous score marks in the bottom of the
ground notch and the presence of surface checks in the bottom of the
broached notch. In an effort to show any microstructural effect of the two
methods of notching, a metallographic specimen was prepared through
each notch on specimens that were not impact tested. This view was
prepared transverse to the notch direction. The specimens were prepared
by the standard methods of ASTM E 3 [4] and E 407 [5].
FIG. 1-Visual appearance of notch bottom for ground-notch Charpy specimen.
FIG. 2-V'~uJal appearance of notch bottom for broached-notch Charpy specimen.

Prior to metallographic preparation, the specimens were nickel plated

in order to improve edge retention during these operations. Views of these
cross sections are shown in Figures 3 and 4. These views give additional
information on the shapes of the notches made by grinding and broaching,
respectively.
FIG. 3-Cross sectional view of ground notch ~ nickel plazJng added for edge retention.
FIG. 4-Cross sectional view of broached notch with nickel pl~Jng added for edge retention.
While the ground notch has a uniformly rounded bottom, the broached
notch has a slight flattening apparent. The results are shown as the
photomicrographs in Figures 5 and 6 for the ground and broached
specimens. The microstructure for the ground specimen showed no
evidence of disturbed metal at the bottom of the notch. In the case of the
broached notch, a shallow effect was observed. The depth of this effect
was measured as 6.4 microns (250 micro-in) at the notch bottom.
FIG. 5-Etched (2% nital) microslnx~tureat notched edge of ground-notched

Charpy specimen. (The white nickel plating at top field of view is for edge retention.)
FIG. 6-Etched (2% nital) mimostmcture at notched edge of broach-notched

Charpy specimen. (The white nickel plating at top field of view is for edge retention.)
A higher magnification view of this same effect on the side wall of the
broached notch is shown in Figure 7. Its depth on the side was 15.9
microns (626 micro-in).
FIG. 7-Etched (2% ~ microstnectureat notch surface of Ioroach-

notched Charpy specimen. (Thewhite nickel pla~ng at the right
side of me r ~ d of view is for edge retention).
Discussion of Results
This evaluation has indicated that both grinding and broaching yield
acceptable Charpy V-notches. The grinding method of notch fabrication
has an advantage in that the grinding wheel is dressed prior to the final
notch pass of 0.076 mm (0.003"). The broached notch produced for this
study had surface checks (tears) that ran transverse to the notch direction.
A slight indication of this effect was also apparent in the microstructure
immediately adjacent to the notched surface. This effect is assumed to
be deformation in the microstructure caused by the broach. It was also
observed on the cross section for the microstructural examination that the
ground notch had a uniformly-rounded bottom while the broached notch
had a slightly flattened bottom. Dimensionally, the ground notches had
smaller ligament sizes and sharper bottom of notch radii than the broached
notches. The statistical comparison of the 24 values for each method of
notching indicated nearly identical consistency, i.e. in standard deviations.
KOESTER AND BARCUSON FABRICATION METHOD 93
Condusion
Notches acceptable for ASTM E 23 Charpy V-notch impact testing can

be made either by broaching or grinding. Possible causes for the minor
variations observed were the following factors: ligament sizes were
different; radii at bottom of notches were different; the broached notch had
a slight flattening at its bottom; the broached notch had surface checks
(tears) and a shallow amount of microstructural damage at the bottom of
the notch.
ACKNOWLEDGMENTS
The authors wish to thank Messrs. W.M. Buffaloe, S.H, Nguyen, R.C.
French and W. Bodenhamer of Southwestern Laboratories for their
assistance in performing the test work, and Mr. D. Burns of Spraymetal,
Inc., Houston, TX, for the nickel plating. We also thank Ms. Pat Koester
for help in the preparation of this paper.
REFERENCES
[1] D.E. Driscoll, "Reproducibility of Charpy Impact Test," Symposium

on Impact Tes'dna. ASTM STP 176, Amedcan Society of Testing
and Materials, 1955, p. 170.
[2] N.H. Fahey, "Effects of Variables in Charpy Impact Testing,"
Materials Research & StandardS, Vol. 1, No. 11, November 1961,
page 872.
[3] ASTM Standard E 23, "Standard Methods for Notched Bar Impact
Testing of Metallic Materials," 1988 Annual Book of ASTM
Standards, Vol. 03.01, Metals-Mechanical Testing; Elevated and
Low-Temperature Tests; Metallography, by Amedcan Society for
Tes~ng and Materials, Philadelphia, 1988, p. 191.
[4] ASTM Standard E 3-80 (Reapproved 1986), "Standard Methods
of Preparation of Metallographic Specimens," ibid. p. 74.
[5] ASTM Standard E 407-70 (Reapproved 1982), "Standard Methods
for Microetching Metals and Alloys," ibid. p. 505.
David A. Fink
0UANTITATIVE COMPARISON AND EVALUATION OF VARIOUS NOTCH MACHINING

METHODS AND HOW THEY AFFECT ASTM E23 AND ISO R442 TESTING
EQUIPMENT RESULTS
REFERENCE: Fink, D. A., "0uantitative Comparison and

Evaluation of Various Notch Machining Methods and How
They Affect ASTM E23 and ISO R442 Testing Equipment
Results", Charpy Impact Test: Factors and Variables,
ASTM STP 1072, John M. Holt, American Society for
Testing and Materials, Philadelphia, 1990.
ABSTRACT: The differences obtained when conducting

Charpy impact tests using an ASTM method E23
specification, versus the ISO type impact tester have
been reported with little agreement upon the
comparisons. There has also been historically much
controversy as far as the proper method of machining
the notch into the Charpy specimen. This paper
compares the effects of the two different types of
impact testing equipment and evaluates machining of the
notch by the methods of grinding, broaching and milling
with a single toothed fly cutter.
It was determined that results obtained with the ISO

type impact machine were elevated by approximately 4%.
It was also determined that a slightly lower impact
value was obtained with broached specimens.
Photomicrographs of typical notch profiles were also
obtained.
KEY~ORDS: Charpy impact testing, notching, machining

methods, impact testing
Mr. David Fink is a Project Engineer for The Lincoln Electric

Company, 22801 St. Clair Avenue, Cleveland, Ohio 44117-1199. Mr.
Fink is a member of the AWS A5 Filler Metal Committee, and is
Chairman of the AWS A50 and ASP Subcommittees on Consumables for
gas metal arc, electroslag and electrogas welding.
94
Copyright 9 1990
of Washington International
of Washington)www.astm.org
pursuant to License Agreement. No further reproductions authorized.
FINK ON NOTCH MACHINING METHODS 95
INTRODUCTION
Charpy impact testing is used for many purposes in industry today.

It is used as a design tool to allow a relatively quick and
inexpensive measure of a steel or a weldment's fracture toughness.
These results are then correlated with successful applications and
service life histories which have been used previously. Charpy
impact testing is also used as a procurement and quality control test
~<~ verify minimum acceptance levels in specifications. It is this
application which is most often controversial.
The two sources for the basic specifications covering Charpy

impact testing today are ASTM and ISO. ASTM (American Society for
Testing and Materials) is an organization made up largely of
volunteers which prepares consensus specifications on a multitude of
topics. ISO (International Standards Organization) is made up of
member countries and prepares specifications for international
application.
In view of the international nature of today's economy, it is not

unusual to evaluate Charpy testing done in several locations
worldwide. The problem this presents is the fact that the results
reported generally are not comparable due to the different nature of
the impact machines being used, and the lack of a recognized proof
testing program for ISO machines.
This state of affairs often leads to much controversy, especially

when the results of the test may show an out-of-specification or
out-of-conformance condition on one type of machine and not
necessarily on the other. For this reason, it is becoming
increasingly important to understand the differences between the ASTM
and ISO type impact testers [1,2].
The method of the machining of the notch has received some

attention previously [3], but it is often overlooked in the study of
the effects in Charpy testing [4,5]. The use of grinding to machine
the notch is often regarded as the referee method; however, it is slow
and expensive. The use of a broach to produce the notch is much
faster and less expensive, but experience has shown that it can often
lead to poor notch profiles if the broach is not properly maintained
or replaced at sufficient intervals. An alternate method which is
thought to be a compromise in speed and cost is the use of milling by
means of a single toothed fly cutter. All three of these methods will
be examined.
The U.S. Army Materials Technology Laboratory has for many years
maintained a program of verification testing and qualification of
impact machines [2]. The author is unaware of any equivalent program
for ISO machines. Verification of the ISO type machine is based on
physical measurements of dimensions and of velocities of the hammer
and weights of the various components.
DESCRIPTION OF EQUIPMENT
The two machines used for this test program were both Tinius 01sen
Universal Impact machines. The machine used for the ASTM testing has
been used for this purpose for a number of years. It has been
subjected to annual verification testing based on U.S. Army Materials
Laboratory standards and has been recertified annually based on those
specimens. It is also subjected to monthly calibration checks with
standardized specimens to ensure its continuous conformance.
The impact machine used for the ISO testing was originally
configured for ASTM testing. Replacement parts were purchased from
Tinius Olsen and retrofitted to the machine to bring it into
conformance with IS0 Standard R442. This machine was then calibrated
and inspected by a representative of TUV America, Inc. (a subsidiary
of TUV Bayern). Their report concluded that, "The evaluation confirms
full compliance with ISO standard R442-1963(E) and the production of
acceptable energy values at high and low energY levels. The impact
machine is in good condition." This machine, was used for all of the
ISO testing. For simplicity's sake, the resulting values were
recorded in terms of ft-lbs, from the dial, rather than Joules. (It
should be noted that the high and low values referred to the MTL
specimens as shown in Table I.)
The difference between the ASTM E23 specification and the ISO R442
standard is primarily the difference of the striker tip on the hammer.
The ASTM specification has a much blunter striker profile with
nominally an 8 mm radius, while the IS0 striker profile specifies a 2
to 2.5 mm radius as shown in Figure I. The ASTM specification also
requires annual verification testing using standard specimens obtained
from the Army Materials Technology Laboratory.
As part of the initial calibration check of the IS0 machine by TUV

America, two sets of Army MTL standard specimens were broken after
verification of the machine had been completed. These low and high
energy standards gave values within the allowable variation for proof
testing. This confirmed the fact that this IS0 designed machine could
meet the proof testing requirements of ASTM. The results of the proof
testing are shown in Table i.
EXPERIMENTAL PROCEDURE
A large number of un-notched impact blanks were purchased from the

manufacturing facility which provides them to the U.S. Army Materials
Testing Laboratory. These were of the low energy level and high
energy level type, finish machined, with the exception of the notch
being absent. These specimen blanks were prepared from SAE 4340
steel, specially heat treated to give specific notch toughness levels
at -40~ in a reproducible fashion. These impact specimen blanks were
then notched by the following methods:
I. Grinding conducted by the contractor for U.S. Army

Materials Testing Lab (data identified as "Standard
Ground" or "Ground #i"), and also a duplicate set by another

machining concern (data identified as "Ground" or "Ground
#2" ).
2. Broaching.
3. Milling with a single toothed fly cutter.
One half of each group of specimens were broken on each of the

ASTM and ISO machines after examination of the representative notch
profiles and inspection of all specimens for proper depth under the
notch. These specimens were all broken at -40~
In addition, samples of both ASTM A537 steel and HY-80 type steel
were used to prepare specimens which were notched by the method of
milling with the single toothed fly cutter. These specimens were
used to compare the ASTM and ISO machines in addition to the previous
mentioned specimens. These specimens were broken at various
temperatures as reported.
RESULTS
Results obtained by testing the standardized impact specimens are

shown in Table 2. This shows the values obtained when testing the
various notching methods on both ASTM and ISO machines at two
different impact levels. These results are shown graphically in
Figures 2 and 3, which show the comparisons for each method or
source of notching. The results obtained by breaking impact
specimens machined from steel plate are shown in Table 3, for both
ASTM A537 and HY-80 base material. The results for ASTM A537 are
shown graphically in Figure 4.
Notched profiles for the various methods were studied using a

scanning electron microscope. Representative photographs of the
profiles obtained with each method are shown in Figures 5 thru 8.
Figure 9 shows the contour obtained with a broach which was

not replaced soon enough.
DISCUSSION
The overall correlation of results from all tests including

several other smaller tests not related to this study are shown in
Figures I0 and 11. These graphs show that in almost every case,
the results obtained with the ISO machine fall above the 1:1
correlation line. The best fit regression line was determined to
be:
(ISO) = 1.0420 (ASTM) + .5160
There is an excellent agreement with the regression line, with a

Coefficient of Determination ( r ) of .9987, and a standard error of
estimate of 1.36.
This would indicate the ISO results are approximately 4% higher

than ASTM values. This is supported also in reviewing specific
notching results shown in Figures 2 and 3, where in each case the ISO
results showed a noticeably higher value.
It is also apparent that the method of grinding the notch gives

the smoothest and most consistent profile. The broached notches give
a lower impact value, which may be related to the roughness of the
surface and the microscopic tears observed in the surface as a result
of the machining. The milling using the single toothed fly cutter
produces results very similar to those of the ground method.
Although in several cases the milling results are elevated slightly,
they still fall well within the normal acceptable tolerances, as
generally required by the MTL program (Table I).
ACKNOWLEDGEMENTS
The author acknowledges with appreciation the help of numerous

members of the staff of both The Lincoln Electric Company, Cleveland,
Ohio, U.S.A., and Lincoln Electric (Europe) S.A. (Rouen, France); in
particular, for their assistance in the preparation and testing of
specimens; Mr. Charles Hayes and Mr. John Baker of the U.S.A.; and
Mr. Leon Keromnes and G. Bordenave of France; and for the preparation
of the presentation, Susan Powell and Beverly Young of the U.S.A.
REFERENCES
I. "ISO Recommendation R442," Verification of Pe~dul~..!mpact

TestinK Machines for Testin K Steels, ISO/R442-1965(E),
First Edition, July 1965.
2. ASTM E23-88 "Standard Test Methods for Notched Bar Impact

Testing of Metallic Materials," 1989 Annual Book of ASTM
Standards, Vol. 03.01, ASTM, Philadelphia, PA pp. 198-213.
3. Towers, O.L. "Charpy V-Notch Tests: Influences of Striker

Geometry and of Specimen Thickness," Report
7709.01/83/354.2, The Welding Institute, Abington Hall,
Cambridge, England, March 1983.
4. Fahey, Robert N. "Effects of Variables in Charpy Impact

Testing," Technical Report WAL TRI12.5/2, Watertown Arsenal
Laboratories, Watertown, Mass. March 1962.
5. Cuthbert, Alan "Impact Properties," Australian Welding Journal,

Vol. 28, No. 4, Summer 1983, pp 26-27.
CO
0
CIO~
"-- N"
0 ~.rr
0
L.
f" 9
0
~W
04 E
o~
co
W
n-
i,
TABLE 1 - - Testing of ISO Impact Tester using U.S.

Army Materials Testing Lab Specimen
LOW HIGH
SET SET
12.5 73.0
11.5 73.0
12.5 75.0
12.0 75.0
14.0 75.0
Average (ft-lbs) 12.5 74.2
MTL Reported Average (ft-lbs) 11.8 72.7
Allowable Variation per MTL + 1.0 if-lb. + 5.0O/o
Actual Variation + 0.7 ft-lbs. + 2.1%
University of Washington (University of Washington) pursuant to License Agreement. No further reproduct
~ ~ ~ ..... o o
>.n-
ed ~.
U.
r- J
~ o4
E
0 ~
O
=oz o ed ~.
~n
O9
~D
~D
~D
O9 nO
~=o
on-_
c-
s
(9
~9
CC
I
I i~ o
nO
04
on-_
(9
W
m
<
o
(9
o I
o t-
. ~
oo
~o
O3 ~- ~oo ,<
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions aut
r
. i
o
o (D
v
c-
I- O -J o o -
S
E rn
o
"o
Q. r-
I 0 ~ 0 ~ 0 0 0 ~ 0 ~ ~ ~ ~
9 <
N 113
"o
e-
. ~
0)
n-
~_~-o
i "(D
0--I09
~ -
I
I
W
..J
m
< <
: o .Q
azI
r y p o C
o l n w o D
~00
(D
s
il
e v i n U
2~
E
E
.$
"E,
I,-
c~
I.IJ
I.,-.
~C
ffl
o ~
o~ ~
E
' ~ ~
o
l ~-
c/J I
I
o Jr ~"
, ~
o
co
o o o o o o ,
lID ~ co ~
Figure 5a -- G r o u n d Notch (#i) - Entry Side
Figure 5b -- G r o u n d Notch (#I) - Exit Side
Figure 5c -- G r o u n d Notch (#1) - Root of Notch
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions author
Figure 6a -- G r o u n d Notch (#2) - Entry Side
Figure 6b -- G r o u n d Notch (#2) - Exit Side
Figure 6c -- G r o u n d Notch (#2) - Root of N o t c h
Figure 7a -- B r o a c h e d Notch - Entry Side
Figure 7b -- B r o a c h e d Notch - Exit Side
Figure 7e -- B r o a c h e d Notch - Root of Notch
Figure 8 a -- M i l l e d (Flycut) Notch - Entry Side
Figure 8b -- M i l l e d (Flycut) Notch - Exit Side
Figure 8c -- Milled (Flycut) Notch - Root of N o t c h
Figure 9a -- Notch Produced with Worn Broach - Entry Side
Figure 9b -- Notch Produced with Worn Broach - Exit Side
Figure 9c -- Notch Produced vith Worn Broach - Root of N o t c h
F I IX I /
I I 1\ i ! "~
+ + + \ t I
I I I \I ] /
I I I ~J I /
l ! f k 1
1 I I J
I! II II I\
II,\
,I ,I II .m, o~
i ! I i'\ I + i
f I 1 l\ i / .A
i i I I l\ /
t I I i
1 1I 1I 1i 1\
J f
/
,~
II fL 1
I
1I i] ~"N! ,I /
/
! I i + l l\ I /
i I I l J t\ + t
'0" I + + L + + .\ i /
I I I I I I \I I
+I""'*"
I I I I i i ~, -I
I I I I i i i\ I
+-
I I I I I i I\ I
t ! I i I I +lL /
I I i i i I I "~ I
! I I ~ i I Z \ I
! I I ~ I I I \I
0 1:3 0 0 0 0 0 0
i ! "-
",i i ~ 0 o
'~ i ~ i OD ,t-
1 ~", ' ' i . 0 oD I
~ i i ! ~ ,!\ i i ~ ~.~
i i i \ 1 i ,,~
i i i \ i i
-~" 1 I I ~ ! +X I 0
i 1 } i ! i ",.. l
0 0 0 0 0 0 0 0 0
T-- T" 9"
0
(J
University of Washington (University of Washington) pursuant to License Agreement. No further repro
B. Anne Fields, Samuel R. Low, III, and James G. Early
THE EFFECT OF FATIGUE PRE-CRACKING VERSUS
V-NOTCHING ON IMPACT TESTING OF CHARPY SPECIMENS
REFERENCE: Fields, B. A., Low, S. R., and Early, J.G., "The

Effect of Fatigue Pre-Cracking Versus V-Notching on Impact
Testing of Charpy Specimens," S v m g o s i u m o n Char~v Impact Test:
Factors and Variables, ASTM STP 1072, John M. Holt, Editor,
American Society for Testing and Materials, Philadelphia, 1990.
ABSTRACT: Charpy impact testing was carried out on both V-

notch and fatigue pre-cracked specimens of two steels: AAR
M128, and ASTM A212-65. Values of total energy absorbed,
lateral expansion and shear fracture appearance were found for
both types of specimen. The total energies absorbed by the
fatigue pre-cracked specimens were markedly less than those by
V-notch samples.
Transition temperature ranges were found for both the V-notch
and pre-cracked specimens. The energy absorption results show
a small increase in transition temperatures for the pre-cracked
specimens. The lateral expansion results are the same in both
types of specimen for the AAR M128 steel and show only a small
change in the ASTM A212-65 steel. Also for the A212 steel the
shear fracture appearance results can be represented by a
single curve for V-notch and pre-cracked specimens. However,
for the M128 steel these curves occur at lower temperatures for
the pre-cracked specimens.
KEY WORDS: AAR M128 steel, ASTM A212 steel, Charpy test,
fatigue pre-crack, instrumented impact test, transition
temperature.
When a standard V-notch Charpy specimen is tested under impact

loading the energy absorbed includes both the energy needed for the
initiation of a moving crack and that for its propagation. The former
may be considerably greater than the latter due to the large amount of
energy required to generate a crack at a notch. Using only the total
energy absorbed by a standard V-notch test can mask the fact that in a
Dr. Fields is a guest scientist in the Metallurgy Division, National

Institute of Standards and Technology (NIST), Gaithersburg, MD 20899; Mr.
Low is a mechanical engineer in the Metallurgy Division at NIST; Dr.
Early is a metallurgist and a scientific advisor to the director of the
Institute for Materials Science and Engineering at NIST.
120
Copyright 9 of
FIELDS ET AL. ON PRE-CRACKING VERSUS V-NOTCHING 121
given material the energy required to reinitiate a sharp crack already

present in the material may be quite low. Thus, when using a fatigue
pre-cracked specimen, the initiation energy may be a much smaller part
of the total energy. This means that a pre-cracked Charpy test can more
closely simulate the conditions under which an existing sharp crack
extends.
However, if pre-cracked specimens are to be more commonly used, it
is necessary to consider whether the standard transition temperatures for
energy absorbed, shear fracture appearance and lateral expansion, as
evaluated for V-notch tests, are still relevant or whether new
specifications need to be introduced. This report contains results
pertinent to such considerations.
EXPERIMENTAL DETAILS
Materials
All the samples were extracted from the head plates of three
railroad tank cars, two of which were fabricated from Association of
American Railroads (AAR) M128-69 steel [I] and are given the notations
of G and U. The third plate used was ASTM A212-65 Grade B steel and will
be referred to as plate S. The compositions of the two steels are given
in Tables i and 2. The head plates from which these samples were taken
are presumed to be in the hot-rolled, hot-formed and stress-relieved
condition. Details concerning the exact thermomechanical condition of
the three plates were not reported to NIST. The tensile properties of
the 'as received' steels from the three plates are given in Table 3.
Specimens
Both standard V-notch specimens and specimens with an additional

fatigue crack were tested [2,3,4]. The nominal dimensions were those
given in ASTM E 23-72 (Notched Bar Impact Testing of Metallic Materials -
Fig. 4, Charpy type A). These are I0 mm thick by i0 mm deep by 55 mm
long. For pre-cracked specimens the machined notch was extended about
2.5 m m b y fatigue cycling such that the total depth of the notch plus the
crack was between 4.5 mm and 5.5 mm. The fatigue cracking was carried
out following the procedure given in ASTM E 399-74.
Two orientations of specimens were used. In one, the longitudinal
specimen axis was aligned parallel to the principal rolling direction and
the plane of notching was in the long transverse direction. In the other
the longitudinal axis was transverse to the rolling direction, while the
notch plane was in the rolling direction. These two orientations are
given the standard notations of LT and TL respectively.
Test Method
Standard Charpy V-notch tests were carried out in accordance with

ASTM E 23-69. The values of total energies absorbed were obtained from
the dial energies recorded. The pre-cracked specimens were tested using
a standard Charpy machine modified for the acquisition of load-time data.
Strain gauges placed on the striking tup were used to sense the load
TABLE i -- Chemical Compositions of Plates G and U (percent by weight)
Specification Tank Car Tank Car

AARMI28-69 GATX 93412 UTLX 38498
Element Ladle Analysis Check Analysis
Grade A Grade B Plate "G" Plate "U"
Carbon 0.25 max 0.23 0.24

Manganese 1.35 max 1.15 1.24
Phosphorus 0.04 max 0.01 0.01
Sulfur 0.05 max 0.017 0.014
Silicon 0,30 max 0.19 0.28
Copper (a) 0.35 max 0,02 0.06
Nickel (a) 0.25 max 0.20 0.15
Chromium (a) 0.25 max 0.09 0.06
Molybdenum (a) 0.07 max 0.05 0.01
Vanadium 0.02 min (a) 0,026 0.01
Aluminum (a) 0.02 0.025
(a) Element not specified, fine grain practice is required,
TABLE 2 -- Chemical Composition of Plate S (percent by weight)
Specification Plate "S"

ASTM A212-65-B
Element Ladle Analysis Check..Analysis
Carbon 0.31 max 0.24

Manganese 0.90 max 0.73
Phosphorus 0.04 max <0.005
Sulfur 0.05 max 0.026
Silicon 0.13/0.33 0.26
Copper (a) <0.05
Nickel (a) <0.05
Chromiun (a) 0.07
Molybdenum (a) <0.05
Vanadium (a) <0.01
Aluminum (a) <0.01
(a) Element not specified, either fine- or coarse-grain practice

allowed.
TABLE 3 -- Tensile Properties of AAR M128 and ASTM A212 Steels
Material Ultimate Yield

Sample Tensile Strength Elongation Reduction
Code and Strength 0.2% Offset in 25.4 mm in Area
Orientation MPa(1) MPa(1) % %
M128 G TL 621 389 26.9 57.4
M128 G LT 582 380 31.7 63.2
MI2S U TL 608 364 33.9 61.0
M128 U LT 610 403 31.4 61.0
A212 S TL 466 222 35.8 57.2
A212 S LT 468 227 37.4 61.4
(i) To convert stresses from MPa to ksi multiply by 0.145.
variation with time. An idealized load-time plot is shown in Figure I.

PGy is the load at general yield and PM is the maximum load. The energy
absorbed, E, can be calculated at time t as discussed by Server [5]:
E = v~ P dt (i)
where V o is the initial impact velocity and P is the load. This equation
assumes that the velocity is nearly constant throughout the impact time,
ie. that the kinetic energy of the hammer is much greater than the
energy absorbed by the specimen. This assumption is valid since the mass
of the hammer is large and E is small. The energy absorbed at maximum
load, E~, can be determined from equation i when t=tM, where tM is the
time at which PM is reached. E M is the energy at crack initiation.
Propagation energy is calculated by subtracting E M from total energy.
Additional requirements for acceptable frequency response, initial
oscillation damping, velocity reduction and electronic curve fitting as
discussed by Server [5] are summarized in Appendix A of reference [4].
Details of the calculation of the true energy absorbed at maximum load
are also described in reference [4].
When comparing energies absorbed in standard Charpy V-notch
specimens and pre-cracked specimens it should be noted that the ligament
area, B(W-a) (where B is the specimen thickness, W is the depth ,and a
is the length of the notch plus the crack), will be smaller in the latter
case because of the extension of the notch by fatigue cracking. Thus,
it is not accurate to directly compare energies for the two types of
specimens. A solution is to normalize the energy values by dividing by
the fracture ligament area. This was done for all subsequent results.
RESULTS AND DISCUSSION
The normalized energies absorbed during standard Charpy V-notch

tests for the plates G, U, and S are shown in Figures 2-7. Also included
in these figures are the normalized total and initiation energies
obtained for the pre-cracked specimens. These results, along with the
calculated propagation energies for the pre-cracked specimens are all
tabulated in previous reports [2,3,4]. Figures 8, I0, 12 show the
PQy
n
"0
0 M
,J
tGy tM
Time, t
FIG. 1 -- Idealized load-time record for impact

loading of a three point bend specimen.
lateral expansion measurements for the TL orientation of both the V-notch

and pre-cracked specimens of plates G, U, and S. Similarly Figures 9,
II, 13 give the results of percentage shear fracture appearance for the
TL specimens. Transition temperatures for the three plates and for three
criteria are given in Table 4: the 270 KJ/m 2 energy absorption
(equivalent to the 15 ft-lb energy in V-notch specimens), the 50% shear
fracture appearance and the 0 . 3 8 m m (15 mils) lateral expansion. For the
initiation energy results the critical temperature is given as that found
at the midpoint of the transition range. Also included are the nil
ductility temperatures previously reported [2,3].
There are some slight variations ($7C) between some of the
transition temperature values given here for the V-notch specimens as
compared to the results listed in reference 4. These are due to
differences in fitting the curves for the transition temperature ranges,
but are not felt to be significant considering the amount of scatter
among the individual data points.
As was stated in the introduction, the energy required to reinitiate
a sharp crack already present in a pre-cracked specimen may be quite low.
This can be seen to be true in Figures 2 to 7, where the lowest curve in
each case is the energy absorbed during initiation. For the lower shelf,
when cracking is cleavage, the energy required for initiation is close
to zero. On the upper shelf the initiation energy is of the order of 125
KJ/m2; or approximately one quarter of the total energy absorbed.
Absorbed Energy Transition Temperatures
Figures 2 to 7 also show that for any given temperature the total
energy absorbed by a pre-cracked specimen is markedly less than that
I ] ] I I
0
u~ O ~
~ v
o; ~ m
. . == E
N 0 ',~
1--
,--I ~ r
8
T
I I I I
G~
C,~u/f"H) ~eClUo~qv A O J e u 3 p e z l lOU~UON
t t t
0
8O v
N 0 ,,'~
~> ~
- ~ o .~
-T 4 0 0~-I
I I ] 18
(,j
C~W/p~) pequosqy ~Oueu 3 p e z l l O l U O N
I 1 - I I I
0
rn 0
~- - 8
v
~~
~ m
- o =o .,9
T
1 I 1 I l
60
< ,,.u/f,~) peclJosqv /,EmJeu 3 p e z ! IDUJON
[ I I I I
8
v
0
~ m
9 ~ t0
N 0,.1
T O0,-1
r E~ ,,q
1 I I I
J~
C,~/p~) peq~osqv AOJeu 3 p e z l l ~ U J o N I-I
I I I l I ..~ U 9
0
~ U ~J
Q ~ 0
0
0 ,,-I
,,.4 ~ U
,-4 ~ Q)
9i i.~ -.: ~.
(~/F'H) peqJo~qy A~Je, U 3 p e z l IO~ON I--I
I I I I I
O
~176 V ~ ~
0 ~ m
t~
N 0 .,4
i! ~
~m.q 0 0,-~
Z 4.-I E-I
i
I I I J8
C~/F'H) pgqJosqy .,',~J~u 3 p g z ! io~u-~ON
' I ' I ' 1 ~ 1 ~ I '
~ > ~
!o" 0 0
T
|
1 ~ I ~ I I I r t
~0 (0 ~1" ~J
eouo..Ioedd V e,.~n],ooJ.= I ~oe~l~
I ' l ' l ' l ' l ' l ' l ' ,tJ U 9
~ U Q)
0
v
0 .,.4
~ k
|
, i I , I , I , I , I , I , CO
0
' I T I ' I ' I ' I '
~.~
.~ ~ ~ 9
-- (~ u) 9
L I
0 rn
i
, I -~ I .., I , I ~, I_ I
I--4
souoJoedd V eJn~ooJ~ JOe4S
, i , [ , i , I , I , I , I = ~a U 9
o r~ ra v .,'4 'x~ u~
~ ~ 0
0 Q 13 m ~ m
9 0 ,,-I
T j~
i
ora
L_ , L I L~ , I , I , I , I
(0 (.0 ~. 04
(mW) uotsuoclx 3 l O J e ~ D - I
' I ' I ' 1 ' I ' I ' 0; O;

o E!
0
v
m
o o
0
o.
E
~ m
m ~ ~ 0
Om i
i
I ~ ] , ] , I , I , I~ eel
eouojD~'dd V ~,Jn~.::)DJ-I JOeM~ ./. H
' l ' l ' l ' l ' l ' l '
J 4-)
~ ,,,--I
v 0
0 ~ ~ 0
| |
0 X 0
L
@
T
| r~
i
i
I I , I , I , ] r I , J i I ,
(0 ~4
(~w) uotsu~dx 3 ]~Je~,o 7

r-~
1-.10 O
~
~[--~
ch ',0 o'~
[-i
~
O
~u
~
',D 9
0 9
03 9
r~ I t,J
0 I1~ 0 i
f-I
r-i~l
i , i ,
0 v v ~Jv
r .
.5
i
,IJ
c~ P~ 91~ 9 O
~ 4J
i I
0
~ ,-4
v v , O
0 ~' C .~ .~ .~
i ~,1 ~ 4.,}
~- 0
OF,~ ~
i i
M v v v v
<
O-4"
i i
[.~ ~~
o 0 , ~
~ $~ 0 0 ~ 0
~ . ~ ~
r~ "~22 q
absorbed by a V-notch specimen. This is due to the large amount of

energy required to generate a crack at the notch. This difference is
greatest at the upper shelf level where the deformation is entirely
plastic. At lower temperatures where fracture is partly cleavage the
difference is not so large. One point to be noted is that at
temperatures where the initiation energy for the pre-cracked specimens
is close to zero, i.e. for cleavage, there is still a difference between
the total energies absorbed by the V-notch and pre-cracked specimens.
This means that there is still a significant energy required to initiate
a crack at the V-notch even when this energy is close to zero for the
pre-cracked specimens.
It should be noted that because the pre-cracked specimens show a
lower total absorbed energy at all temperatures, the 270 KJ/m z energy
absorption temperatures (15 ft-lb energy in V-notch tests) are
necessarily higher than those for the V-notch tests. This can easily be
seen in Figures 2 to 7. However these differences are not large, less
than 20 degrees C for each of the tests.
Lateral Expansion Results
The results of the lateral expansion measurements for the TL

orientation specimens of the G, U, and S plates are shown in Figures 8,
i0 and 12 respectively. For the G and U plates the results can be
represented by a single curve up to the upper shelf for the pre-cracked
tests. This means that the temperature for which there is a lateral
expansion of 0.38 nun (15 mils) is the same for both plates. For plate
S the curves, while separate, are not greatly different and the 0.38 mm
temperatures are only 9 degrees apart. This is not large when the
scatter in the data is considered. Therefore the 0.38 mm lateral
expansion temperature appears, from the present results, to be a
temperature common to both V-notch and pre-cracked tests.
Shear Fracture Appearance Results
Shear fracture appearance results for the TL orientations of plates

G, U, and S are shown in Figures 9, II and 13. For plates G and U it can
be seen that this transition temperature range is actually lower for the
pre-eracked specimens. These tests are the only ones of those carried
out here where this was observed. The differences obtained for the 50%
shear fracture appearance are 15 and 17 degrees C respectively. In spite
of the scatter in the data this appears to be a real effect, although the
reason for the difference is not known. For plate S the results for both
types of specimen can be represented by a single curve.
CONCLUSIONS
I. The energy required to reinitiate cracking in a previously pre-cracked

specimen is close to zero for cleavage and only of the order of 125
KJ/m 2 for ductile fracture.
2. The total energy absorbed by the pre-cracked specimens is m a r k e d l y

less than that for the V-notch specimens. This is due to the large
amount of energy needed to generate a crack at a blunt notch. This

effect is greatest at the upper shelf where fracture is 100% shear.
3. The energy absorption transition temperature range is slightly higher

for the pre-cracked specimens. For the 270 KJ/m 2 (15 ft-lb) energy
absorption temperatures this difference is less than 20 degrees for
all tests.
4. The temperatures at which there is a lateral expansion of 0.38 mm

(15 mils) are the same for both types of specimen from M128 steel. For
A 212 steel the pre-cracked specimens show an increase of 9 degrees
over the V-notch tests.
5. The transition temperature results for shear fracture appearance in

both types of specimen from A212 steel can be represented by a single
curve. For M128 steel the results for the pre-cracked specimens give
lower transition temperatures than for the V-notch specimens. This
is not understood at this time.
6. Present results indicate that for pre-cracked specimens of M128

steel the best correlation with standard Charpy V-notch transition
temperatures is found using values of lateral expansion, while for
A212 steel the best correlation is found using shear fracture
appearance.
ACKNOWLEDGEMENTS
The authors wish to express thanks to Mr. D. E. Harne for his

contributions to the mechanical testing involved in this project and to
the Federal Railroad Administration for support under Interagency
Agreement A R-40008 and for valuable comments and advice.
REFERENCES
[i] AARMI28 Specification for High Strength Carbon Manganese Steel Plates
for Tank Cars, Manual of Standards and Recommended Practices, Section
C - Part III, Specifiaction M-I002, pp M-7, M-8, The Association of
American Railroads, 50 F St., N.W., Washington, D.C. 20001, 1985.
[2] Early, J.G., "AMetallurgical Analysis of an ASTM A212-B Tank Car Head
Plate", Report No. 9, NBSIR-78-1582, National Bureau of Standards,
Sept. 1978.
[3] Early, J.G. and Interrante, C.G., "A Metallurgical Evaluation of Two
AAR M128 Steel Tank Car Head Plates Used in Switchyard Impact Tests",
Report No. 10, NBSIR-80-2039, National Bureau of Standards, May 1980.
[4] Early, J.G., Interrante, C.G., Low, S.R., and Fields, B.A., "Dynamic
Mechanical Properties of Two AAR M128 and One ASTM A212-B Steel Tank
Car Head Plates", Report No. 14, NBSIR-88-3690, National Bureau of
Standards, June 1988.
[5] Server, W.L., "Impact Three-Point Bend Testing for Notched and
Precracked Specimens", Journal of Testing and Evaluation, JTEVA, Vol.
6, No. i, Jan. 1978, pp. 29-34.
Stephanie Mikalac, Michael G. Vassilaros, and Harold C. Rogers
PRECRACKING AND STRAIN RATE EFFECTS ON HSLA-IO0 STEEL CHARPY SPECIMENS
REFERENCE: Mikalac, S., Vassilaros, M.G., and Rogers, H.C.,

"Precracking and Strain Rate Effects on HSLA-100 Steel Charpy
Specimens," Cha__b_~Xp.~Impact Test: Factors and Variables, ASTM
ST__~P1072, J. M. Holt, Ed., American Society for Testing and
ABSTRACT: Improved predictions of the ductile-to-

brittle transition behavior of structures are possible using
Charpy specimens that have increased constraint. This report
contains the results of an investigation into the combined
effects of a change in notch acuity and a change in loading
rate on the transition curve of a high-strength, low alloy
(HSLA) steel. Specifically, the changes in absorbed energy
and ductile-to-brittle transition temperature were noted.
Standard Charpy V-notch specimens were fatigue precracked to
a total crack depth (notch and precrack) of approximately
0.280 cm. The specimens were subsequently tested using an
impact loading rate of 5.1 m/see over a wide range of
temperatures to fully develop the transition temperature
curve. The transition temperature was determined using the
50% Fracture Appearance Transition Temperature (FATT). As
expected, the level of absorbed energy was lower for
precracked specimens as compared to standard specimens. The
transition temperature of the preeracked specimens was
shifted upward by approximately 40~ Both standard V-notch
specimens and fatigue precracked specimens were tested at a
slower loading rate of 0.0025 cm/sec to determine the
combined effects of the sharper crack tip and the change in
loading rate on both the energy absorbed and the transition
temperature. At the slow loading rate the absorbed energy
was lower for precracked specimens while the transition
temperature was shifted upward by approximately 20~ The
results were also compared with those from 1.6-cm thick
dynamic tear (DT) tests. It was found that the transition
curve developed through fracture appearance for the DT test
was identical to that of the precracked CVN tested at the
impact loading rate.
KEYWORDS: ductile-to-brittle transition, Charpy, notch

acuity, loading rate, transition temperature, dynamic tear,
energy absorption, precracking, impact testing
S. Mikalac and M.G. Vassilaros are research engineers at David

Taylor Research Center, Code 2814, Annapolis, MD 21402; Professor
Rogers is head of the Materials Engineering Department, Drexel
University, Philadelphia, PA, 19104.
184
Copyright @ of
MIKALAC ET AL. ON PRECRACKING AND STRAIN RATE EFFECTS 135
BACKGROUND
Since its introduction in the early 1900's [i], the Charpy V-notch
(CVN) impact test has been the one most commonly used to characterize
the notched impact behavior of structural metals. It is especially
useful for characterizing the transition from ductile to brittle
fracture behavior in steel under varying conditions of temperature and
loading rate. Ductile fracture is the dominant mode in the plastic
regime, that is, the upper shelf region of the transition curve.
General yielding occurs in this region usually accompanied by the
development of shear lips oriented at 45 degrees to the direction of
the applied stress, In the lower shelf region of the transition curve,
brittle fracture occurs under elastic stresses. In this case, fracture
is normal to the direction of the principal stress with little or no
shear lip development. Within the transition portion of the curve,
sometimes known as the elastic-plastic region, CVN specimens show a
mixed mode of fracture.
The CVN test has many advantages over the more complex methods for
analyzing material toughness. The sample size is small, the test
itself is quick and easy to conduct, and it is relatively inexpensive.
However, there are some significant disadvantages. First, because of
the small sample size, the constraint developed in the specimen may be
very limited. This may lead to non-conservative predictions of the
toughness and transition behavior of the material. Second, the test
cannot differentiate between energy consumed in crack initiation and
that absorbed during crack propagation. Finally, the Charpy impact
test does not provide results that can be utilized directly in
structural design as can Klc, the fracture toughness level of a
material.
There have been countless reports deriving empirical relationships

between CVN energy values and fracture toughness. Any correlations
that have been developed between fracture toughness measured by
standard procedure, i.e., ASTM E399, and CVN test results may not be
applicable when the materials or operating conditions change. Although
it is doubtful that a single correlation between Charpy test results
and fracture toughness that is valid for all materials and conditions
may be obtained, there is a possiblity that a modified Charpy test
could replace other toughness testing that requires larger specimens
such as the dynamic tear test [2,3].
Fatigue precracking is one method of increasing the constraint

under which the Charpy specimen deforms. The sharper crack tip
increases the level of tensile stress below the notch. It also reduces
the energy required to initiate crack propagation. Precracking
generally results in a lowering of the upper shelf energy of the
material and shifts its transition temperature upward. [2-7].
Variations in the strain rate will also change the constraint by
changing the yield stress of the steel. For a given geometry, an

increase in flow strength (which occurs when the strain rate is
increased) will drive the specimen behavior toward linear-elastic plane
strain behavior. This would be manifested by a drop in energy
absorption. In addition, at standard impact loading rates, the
transition temperature will usually be shifted upward as the strain
rate is increased. [5,7,8].
INTRODUCTION
The objective of this study was to understand the separate and

combined effects of notch acuity and loading rate on the ductile-to-
brittle fracture behavior of HSLA-100 steel. The effects of
precracking standard Charpy V-notch specimens on both the energy
absorption and the transition temperature were investigated. The
effects of a change in loading rate on CVN behavior were also examined
with respect to the absorbed energy and transition temperature shift.
Results were used to establish the change in constraint in the specimen
measured through energy absorption and fracture appearance.
The chemical composition, heat treatment and mechanical properties

of the steel are shown in Table I:
TABLE I -- HSLA-100 3-cm thick plate chemistry (weight percent)

Lukens Steel Company
Austenitized at 900~ for 75 minutes, water quenched
Aged at 640~ for 75 minutes, water quenched
C Mn S__~i Cu N_ii C_Kr Mo AI Cb

0.04 0.86 0.27 1.58 3.55 0.57 0.60 0.032 0.030
Transverse tensile properties:

0.29 Yield strength: 758 MPa
Tensile strength: 807 MPa
Elongation: 259
Reduction in area: 769
PROCEDURE
Charpy V-notch (CVN) specimens (type A, ASTM E23) were cut from a
3-cm plate and tested in the transverse-longitudinal (T-L) direction.
Testing was performed over a wide range of temperatures in order to
fully develop the temperature transition curve of the steel. Specimens
were tested at temperatures of -120~ -90~ -30~ 0~ and at room
temperature. The total energy absorption and fracture appearance were
recorded.
Precracked specimens were prepared by loading the specimen in three

point bend and subjecting them to an oscillating load guided by ASTM
E812.
Standard V-notch and fatigue precracked CVN specimens were tested

according to ASTM E23. The impact velocity generated by the testing
machine was 5.1 m/sec. Standard and precracked CVN specimens were also
tested at a loading rate of 0.0025 cm/second using anvils and tup from
a standard Charpy testing machine. The absorbed energy was measured as
the area under the recorded load versus cross-head displacement curve
corrected for machine compliance. Measured energy absorption values
were normalized with respect to the remaining ligament below the V-
notch or fatigue precrack.
Transition temperatures were developed using the 50% Fracture

Appearance Transition Temperature (FATT) method; this is defined as the
temperature at which the surface of the Charpy specimen exhibits equal
amounts of cleavage and ductile (shear) fracture.
Dynamic Tear (DT) specimens (ASTM E604) were also cut from the same
plate and tested in the transverse-longitudinal direction over a wide
range of temperatures in order to fully develop the transition curve.
Fracture appearance as a function of testing temperature was recorded.
Figure i shows the transition curves of the two different notch

configurations for CVN specimens tested at the impac~ loading rate of
5.1 m/sec. The normalized values of energy absorbed by the steel was
lower in precracked specimens at all temperatures. The results agree
with those of [3,4,6]. The transition from mostly ductile fracture to
mostly cleavage (50% FATT) occurred at -90~ (+/- 3~ as shown in
figure l(b) in the standard V-notch CVN specimens. The precracked
Charpy specimens had a much higher 50% FATT of -50~ Clearly, the
sharp crack tip of these specimens caused the transition temperature to
shift upward significantly, consistent with [3-7]. By reducing the
notch tip radius in a CVN specimen, the stress concentration at the tip
of the crack is increased. This also causes an increase in the
effective strain rate and also increases the concentration of strain at
the crack tip. The end result is an increase in the tensile stress
level below the notch [9,10]. Brittle fracture is thus promoted with
the sharper crack tip; this is seen in an increase of the 50% FATT and
a decrease in normalized energy absorption.
The transition curves for the two different notch configurations

for CVN specimens tested at a loading rate of 0.0025 cm/sec are shown
in figure 2. As in the tests performed at the impact loading rate, the
normalized absorbed energy was lower with precracked specimens at all
temperatures. The decrease in absorbed energy at the slow loading
rate, however, was not as large as when the tests are performed at the
impact loading rate. Also, with the reduction in strain rate, the
notch effect on the transition temperature was smaller. Although there
was substantial scatter in the transition region at this loading rate,
the transition temperature appeared to increase for the precracked
specimens by about 20~ from -75~ to -55~ The smaller change in
energy absorption and in transition temperature with precracking at the
slower rate may be the result of opposing constraint changes. At the
slower loading rate, the effective yield strength of the material is
lower [7], thus reducing the constraint developed. Therefore, the
changes in the transition temperature curves, when testing is carried
out at the slower loading rates, would be expected to be smaller than
those at the impact loading rate because the increased constraint
caused by fatigue precracking is geometrical, independent of the rate
of testing.
The transition curves of the standard V-notched specimens tested at

the two different loading rates are shown in figure 3. At the slower
strain rate there was a decrease in normalized energy absorption at all
temperatures. In terms of fracture appearance, the upper and lower
shelves were quite similar. There appeared to be an increase in the
50% FATT on the order of 15~ with the slow loading rate. This
behavior cannot be explained using the strain rate effect model
discussed earlier in the background section.
The transition curves of the fatigue precracked specimens tested at

the two loading rates are seen in figure 4. The upper shelf energy
(USE) at the two loading rates was similar. In the transition region,
however, the energy absorption at the slow loading rate was higher than
at the impact loading rate. These res61ts corroborate with those
observed by [8]. The transition temperature was approximately -50~
(+/- 3~ for the impact loading rate. The transition temperature at
the slower loading rate was approximately -55~ As seen in figure
4(b) the values of percent shear as a function of temperature measured
for fatigue precracked CVN specimens tested at the impact loading rate
appeared to provide a lower bound for the results obtained at the
slower rate. The increased flow strength caused by the increase in
strain rate appears to change the transition characteristics of the
steel, particularly in terms of normalized energy absorption.
It is expected that in front of a sharp crack (such as the fatigue

precracked CVN specimens), the plastic zone size is smaller than that
found in a blunt notch (such as the V-notch CVN specimen). Also the
maximum stress developed is higher and located closer to the crack tip
[11,12]. Thus the microstructural features that control fracture (thus
controlling energy absorption and transition temperature) in the
fatigue precracked specimens are not necessarily identical to those
controlling fracture in the standard V-notch specimens [5]. The end
result is that the models [2,3,8,9], which appear to work well for the
sharp notch configuration when describing loading rate effects cannot
always be used successfully when considering blunt notches [12].
Loading rate changes did not affect the standard V-notch specimens in
the same manner as they did the fatigue precracked specimens. For
example, the upper shelf energies of the V-notched specimens were
dramatically changed with changes in loading rate while the precracked
specimens had similar upper shelf energies despite the difference in
loading rate. This may be the result of a difference in the
microstructural factors controlling fracture in the two specimen types
as discussed by others [5,12].
The following table summarizes the major findings of this study of

the separate and combined effects of a change in notch acuity and
loading rate on the energy absorption and 50% FATT of HSLA-IO0 steel
Charpy specimens.
TABLE 2 -- Summary of Charpy Test Results
Standard V-notch Precracked V-notch
Impact Loading Rate:
USE = 3.06 J/mm^2 USE = 2,19 J/mm^2

50% FATT ~ -90~ 50% FATT = -50~
Slow Loading Rate:
USE ~ 2.43 J/mm^2 USE ~ 1.97 J/mm^2

50% FATT = -75~ 50% FATT = -55~
Figure 5 shows the transition temperature curve developed from 1.6-

cm dynamic tear specimens made of the same HSLA-100 steel. The 50%
FATT is approximately -50~ Comparing the precracked CVN curve tested
at the impact loading rate with the DT transition curve, one can see
the results: the transition temperature curves were almost identical.
Through-thickness stresses increase as the thickness of the specimen is
increased. These increasing stresses cause greater plastic constraint
at the notch root [13]. As discussed earlier in the background
section, precracking increases the constraint under which the Charpy
specimen deforms by increasing the level of tensile stress below the
notch. The constraint developed in the 1 0 - m m p r e c r a c k e d Charpy V-notch
specimen, as measured through fracture appearance, was increased enough
to mimic the constraint developed in the tricker 1.6-cm DT tests.
Similar results were reported by [2,3].
CONCLUSIONS
Fatigue precracking of Charpy V-notch specimens of HSLA-100 steel

caused the energy absorption to drop and the 50% Fracture Appearance
Transition Temperature to shift upward by approximately 40~ when
tested at an impact loading rate of 5.1 m/sec.
At a slower loading rate of 0.0025 cm/sec, fatigue precracking

still resulted in a decrease in energy absorption at all temperatures
although the change was not as large as observed at the impact loading
rate. There was a smaller increase in the 50% FATT of approximately
20~ when the V-notch was fatigue precracked.
With a blunt notch, the slower loading rate decreased the energy
absorption at all temperatures. At the slower loading rate there was a
shift upward of approximately 15~ in the 50% FATT.
With the sharper crack tip, there was a slight increase in energy
absorption at the slower loading rate in the lower transition region.
The 50% Fracture Appearance Transition Temperature changed only
slightly with loading rate; however, at temperatures below the
transition temperature, the standard rate appeared to provide a lower-
bound estimate of the transition curve.
Precracked Charpy V-notch specimens were found to provide a more

conservative measure of a material's notch toughness and were less rate
sensitive than the blunt V-notch specimen. The constraint developed in
a fatigue precracked Charpy V-notch specimen tested at an impact
loading rate appeared to be similar to that of a dynamic tear specimen
made of the same HSLA-IO0 steel. Both specimen types had a 50% FATT of
approximately -50~
ACKNOWLEDGMENTS
The authors acknowledge Dr. A.K. Vasudevan from the Office of Naval
Research for his support. Thanks also go to E. Tees, J. Sanders,
C. Fraser, E. Czyryca, E. Hackett, R. Link, and P. Joyce.
REFERENCES
[I] McAdam, D.J. and R.W. Clyne, "The Theory of Impact Testing:
Influence of Temperature, Velocity of Deformation, and Form and Size of
Specimen on Work of Deformation," Symposium o__nnImpact Testing, American
Society for Testing and Materials, 1938, pp. i12-134.
[2] Sharkey, R.L. and D.H. Stone, "A comparison of Charpy V-notch,
dynamic tear, and precracked Charpy impact transition-temperature
curves for AAR grades of cast steel," Transactions of the ASME, May
1976, pp. 446-452.
[3] Diesburg, D.E., "Transition temperature characterized by

precracked Charpy specimens," Toughness Characterization an___dd
Specifications for HSLA and Structural Steels ed. P.L. Mangonon, Jr.,
AIME: 1979, pp. 19-36.
[4] Louden, et al., "The influence of specimen size on Charpy impact

testing of unirradiated HT-9," ~ournal of Nuclear Materials, 155-157,
pp. 662-667, (1988).
[5] Tvergaard, V. and A. Needleman, "An analysis of the temperature

and rate dependence of Charpy V-notch energies for a high nitrogen
steel," International Journal of Fracture, Vol. 37, pp. 197-215,
(1988).
[6] McKinney, K.E. et al., "An evaluation of the toughness of ductile

iron vs cast steel using modified Charpy test specimens," AFS
Transactions, 84-122, pp. 239-250.
[7] Wullaert, R.A., "Applications of the Instrumented Charpy Impact
Test," Impact Testing of Metals, ASTM STP 466, American Society for
Testing and Materials, 1970, pp. 148-164.
[8] Barsom, J.M. and S.T. Rolfe, "Correlations between Klc and Charpy
V-notch test results in the transition-temperature range," Impact
Testing of Metals, ASTM STP 466, American Society for Testing and
Materials, 1970, pp. 281-302.
[9] Dieter, Jr., G.E., Mechanical Metallurgy, McGraw-Hill Book

Company, Inc., New York, NY, 1961.
[I0] Wullaert, R.A., Ireland, D.R., and Tetelman, A.S., "Radiation

Effects on the Metallurgical Fracture Parameters and Fracture Toughness
of Pressure Vessel Steels," Irradiation Effects on Structural Alloys
for Nuclear Reactor Applications, ASTM STP 484, American Society for
Testing and Materials, 1970, pp. 20-41.
[Ii] Veistinen, M.K. and K. Wallin, "Cleavage Fracture in 26Cr-IMo

Ferritic Stainless Steel," Materials Science Technology, July 1986, pp.
666-670.
[12] Lin, T. and R.O. Ritchie, "On the Effect of Sampling Volume on the
Microscopic Cleavage Fracture Stress," Engineerins Fracture Mechanics,
Vol. 29, No. 6, pp. 697-703, 1988.
[13] Hertzberg, R.W., Deformation a_u Fracture Mechanics of Engineerinz

Materials, 2nd ed., John Wiley & Sons, New York, NY, 1983.
ERRATUM FOR ABTM BTP 1072
The publisher regrets that five pages of figures to the pap

"Precracking and Strain Rate Effects on HSLA-100 Steel Char
Specimens," by Stephanie Mikalac, Michael G. Vassilaros, an
Harold C. Rogers (pp. 134-141), were inadvertently omitted.
Normalized Energy (J/ram^2) Standard Loading Rate
4
Specimen Type ~ l
Standard notch A
[~Precrackednotch ~ <) "~- ~-~1
3
8 []
2
N ~ ~ _ ~ ~I , __ , __ ~
0
-140 -120 -100 -80 -60 -40 -20 0 20
Temperature (degrees C)
Figure l(a): Normalized Charpy V-notch Energy versus Temperature;

Standard Loading Rate; for HSLA-IO0 Steel
Percent Shear Standard Loading Rate
100
Standard notch
80
60
[] Precracked notch
j~
40
20
0 1 ~J. I I
-140 -120 -100 -80 -60 -40 -20 0 20

Figure l(b): Percent Shear in Charpy V-notch Specimens versus

Temperature; Standard Loading Rate; for HSLA-IO0
Steel
Normalized Energy (J/ram'2) Loading Rate 9 0 . 0 0 2 5 cm/sec
4
SpocimenType
Standard notch
D Precrscked notch
[]
( l I I I I I
0 - -
-140 -120 -100 -80 -60 -40 -20 0 20

TemDerature (degrees C)
Figure 2(a): Normalized Charpy V-notch Energy versus Temperature;

Loading Rate - 0.001"/sec; for HSLA-IO0 Steel
Percent Shear
100 Specimen Type ~

Standard notch
[] Precrackednotch
80
60 []
40
20
0 ] I L I I I I
-140 -120 -100 -80 -60 -40 -20 0 20

Figure 2(b): P e r c e n t S h e a r i n Charpy V - n o t c h S p e c i m e n s v e r s u s

T e m p e r a t u r e ; L o a d i n g R a t e - O . O 0 1 " / s e c ; f o r HSLA-IO0
C o p y r i g h t b y A S TSMt eIenlt ' l ( a l l r i g h t s r e s e r v e d ) ; T u e D e c 1 5 1 2 : 5 9 : 5 2 E S T 2 0 1 5
University of Washington (University of Washington) pursuant to License Agreement. No furth
Normalized Energy (J/am'2) Standard V-notch
4
Loading Rate
Standard rate
r-I SIowrat. ~ ~ ....r~ " ..~
[]
<>
G I .... I I I J i !
-140 -120 -100 -80 -60 -40 -20 0 20

Standard V-notch; for HSLA-100 Steel
Percent Shear Standard V-notch
100 Loading Rate [] ~

Standard rate A//
[] Slow
80
60-
4o!
t
t
oL
-140 -120 -100 -80 -60 -40 -20 0 20
Figure 3(b): Percent Shear in Charpy V-notch Specimens versus

Temperature; Standard V-notch; for RSLA-IO0 Steel
Normalized Energy (J/mm'2) Fatigue Precracked Notch
4
Loading Rate
0 Standard rata
[] Slow rate
0
[]
0
-140 -120 -100 -80 -60 -40 -20 0 20

Fatigue Precracked Notch; for HSLA-100 Steel
Percent Shear Fatigue P r e c r a c k e d N o t c h
100 - Loading Rate

Standard rate
[] Slow rate
80
60
I
J
I
0 L~
~ J ...............................
-140 -120 -100 -80 -60 -40 -20 0 20
T e m p e r a t u r e ( d e g r e e s C)
Figure 4(b): Percent Shear in Charpy V-notch Specimens versus

Temperature; Fatigue Precracked Notch; for HSLA-100
Downloaded/printed by Steel
Percent Shear Standard Loading Rate
100 - Specimen Type

0 PrecrackedCVN
[] DynamicTear
80 []
60
40
20
S
r n j ~ J L , i
0 r
40 -120 -100 -80 -60 -40 -20 0 20 40
Figure 5: Percent Shear in Precracked Charpy V-notch and Dynamic
Tear Specimens versus Temperature; Standard Loading
Rate; for HSLA-100 Steel
Charles G. Interrante and James J. Filliben
SIGNIFICANCE OF PRECRACKING VARIABLES FOR SLOW-BEND CHARPY TESTS
REFERENCE: Interrante, C. G. and Filliben, J. J.,

"Significance of Precracking Variables for Slow-Bend Charpy
Tests," Charpu Impact Test: Factors and Variables, ASTM STP
1072, John M. Holt, editor, American Society for Testing
ABSTRACT: The significance of four variables in the

technique used to precrack Charpy specimens of metallic
materials is determined by analyses of seven responses
computed from results of slow-bend tests. The variables
include crack size, stress-intensity factor at the start of
precracking, notch preparation prior to precraeking, and
material. All four variables are shown here to be
significant for more than one of the computed responses.
Seven response parameters, each representing alternative
methods for evaluations of fracture toughness, were evaluated
for each test. Responses are based either on a single value
of load or energy absorbed in the test. The results indicate
that (I) all seven computed responses are linearly related to
crack length and the sensitivity to crack length is a
function of both response parameter and material, and (2)
precraeking at either very high or very low levels of stress-
intensity factor, Kf, are to be avoided. This work is the
result of a study conducted hy ASTM Task Group E-24.03.03 and
members of eight participating laboratories.
KEYWORDS: aluminum; Charpy; crack size; fatigue precracking;

fracture toughness; Kf maximum; Kruskal-Wallis test; linear
regression; notch preparation; precracking; statistical tests;
steel; stress-intensity factor; titanium
List of Symbols
A Area of uncracked ligament at start of test:

A=B(W-~)
A, Aluminum Aluminum alloy 2419-T851 in the aged condition.
Dr. Interrante is currently a Senior Metallurgist with the U.S.

Nuclear Regulatory Commission, Washington, DC 20555 and Dr. Filliben
is a Statistician in the Mathematical Analysis Division at the
National Institute of Standards and Technology, Gaithersburg, MD
20899, where this paper was produced.
142
Copyright 9 of
INTERRANTE AND FILLIBEN ON PRECRACKING VARIABLES 143
a2 5 Crack length at one quarter-thickness location

a5 o Crack length measured at mid-thickness location
a7 5 Crack length measured at the other quarter-thickness
location
(a25 + as0 + a75)/3 = crack-size factor
Normalized crack size
B - Specimen thickness = I0 mm (0.394 in)
A coefficient that is a result of a regression
analysis
CVN - Charpy V-notch (test specimen)
CPS Cumulative probability statistic
C(T) Compact specimen loaded in tension
C Experimental determination of total elastic
compliance of specimen (=DI/PI)
Cm Machine compliance
Cs Theoretical, elastic specimen compliance
D1 Displacement to an arbitrary point "i" in the
elastic region
5 Displacement
E Young's modulus
E= Energy correction based on specimen compliance and
crack length
EDM Electric discharge machining
EM Energy to maximum load, under the load-displacement
trace
ET Total energy under the load-displacement trace
ET/A Total energy divided by area of uncracked ligament
E~ Corrected energy to maximum load EM = E~ - E c
K Stress-intensity factor
Kd A response, computed from P* and Y*, termed lower-
bound or equivalent energy (EE)
Ks Stress-intensity factor in fatigue precracking
Kf maximum Kf at the start of precracking in a constant-
deflection apparatus
Kf ratio Kf maximum/K smallest
K smallest The smallest value of K computed from equations i,
2, and 3 of this report
KI c - Reference values of fracture toughness obtained from
CTS tests, whether valid (E 399) Klc or invalid ~
K3 A response computed from energy to maximum load
K~ Response Kj corrected for machine compliance of
specimen
A response computed using PQ as in ASTM Method
E 399
~Q-PM A response computed as for KQ except PM is used in
place of PQ
A response computed from total energy absorbed in
fracture
Poisson's ratio
N The number of observations
v Residual degrees of freedom
MLR Multiple linear regression
NP Notch preparation
N/A Not applicable
%RD of K Percent relative deviation = (K - KI= ) x 100/Kic
%RD of Rsb Percent relative deviation = (Rsb - Rsb) x 100/Rs5
P1 Load to an arbitrary point "i" in the elastic region

PM Maximum load
PQ Load at the 5-percent secant intercept
p* A value of load used to compute equivalent energy
RSD Residual Standard Deviation
Rsb Specimen strength ratio in slow-bend testing
Rsb Mean Rsb for each material
S, Steel Steel alloy A L M A R (200) in the maraged condition
Sy Significant differences based on magnitude of test
responses
Ss Significant differences based on reproducibility of
test responses.
S Standard deviation for replicate responses
Sp Standard deviation pooled for all responses of a
material
aU Ultimate tensile strength
Gy Yield strength
Oy 1 ay at the precracking temperature
Gy 2 ay at the Charpy test temperature
T, Titanium Titanium alloy Ti-6AI-4V in the annealed condition
W Specimen width = IO.O mm (0.394 in)
Mean response
Y* A function of ~/W used to compute K d
Yl An individual test response
I. Background
A proposal for standardization of a precracked Charpy impact test

was made by the Executive Committee of ASTM Committee E24, in January
of 1971. Task Group E24.03.03 was formed to deal with this problem.
The task group drafted a preliminary document titled "Proposed Method
for Precrack Charpy Impact and Slow-Bend Testing of Metallic
Materials," which required experimental work to determine the
significance of variables in the fatigue precracking procedures
prescribed in the proposed method. The "best procedures" for fatigue
preeracking had to be established. Further, the expected variability
of test results had to be determined for a multiplicity of
laboratories using a prescribed best method.
At the request of the Chairman of Task Group E24.03.03, the

authors, from the National Institute of Standards and Technology
(NIST), formerly the National Bureau of Standards (NBS), furnished a
proposed statistically designed experiment for determining the
significance of four precracking variables on results of tests
conducted with fatigue precracked Charpy specimens. This proposal
included three levels for each of the four variables (here called
factors). The factors and their levels were later modified at
meetings of the Task Group before test specimens were prepared. In
addition, the proposal specified two methods of testing, slow-bend and
impact; together, these proposed tests comprised what the Task Group
called Phase I of their testing program.
This is an analysis of the results of the slow-bend tests

conducted for Phase I. At the time that this work was being planned,
a report published by the National Materials Advisory Board [i]
recommended... "that the fatigue-precracked Charpy-size specimen,
tested in slow bending to measure the ratio of specimen strength to

either the yield strength or the ultimate tensile strength of the
material (ASTM E 399) be utilized, when applicable, for establishing
correlation with plane-strain fracture toughness and minimum
acceptance standards in quality-control programs. To foster
implementation of this recommendation, the Committee urges that the
test method be standardized as soon as practicable..." The Phase I
test program was distinguished from other extensive research programs
[2,3] I that have used precracked Charpy specimens, as the objective of
this program was to establish the effects of precracking variables.
The proposed Phase II effort was to be conducted by many laboratories
to establish a lab-to-lab variability for precracked Charpy test
results.
Charpy test specimens used in this program differ from the standard
ASTM E 23 type A, V-notch Charpy specimen: (i) Charpy specimens for
this program contain a fatigue precrack; in this respect they are
similar to valid plane-strain fracture toughness specimens (ASTM E
399), while the standard Charpy specimen is not precracked; and (2)
the standard V-notch root-radius of 0.25 mm (0.010 in) is here
modified in various ways, as shown below, to facilitate crack
initiation under fatigue loading.
TABLE 1 -- Three Notch Preparation Used Before Precracking
Machined Notch Final

Root Radius Preparation Special Root
(mm) (in) Code _ _ Preparation Radius
.250 .010 1 razor scratch - .05 mm (.002 in)
.125 .005 2 none .125 mm (.005 in)
.250 .010 3 EDM with razor - .05 mm (.002 in)
electrode
2. Test Matrix
In the proposed Phase I program, each of three materials are

designated to be tested after being precracked in the various ways
specified in the proposed test matrix. These specified precracking
variables are notch preparation (NP), stress-intensity factor at the
start of precracking (Kf maximum), and crack size (~). Each
preeracking variable is controlled at three levels, which are coded I,
2, and 3 in the proposed matrix. Thus, there are a total of four
factors and three levels per factor in the experimental design. In
addition, replicate specimens were specified for each test condition
in the proposed text matrix. The actual numbers of specimens tested
at each condition, given in Table 2 to be discussed later, differ
slightly from the proposed 2 replicate tests per condition.
2.1 Materials
Three materials included in the Phase I program are an aluminum
iFigures in brackets indicate the literature references at the

end of this paper.
o~
c,h
O
O Q.I
a.I
"O
0
~J
P
if:
0
il ~ c,h c~l cN
r ~ Q;
~n ~'O
g~ ~o
0
I~l 0
o<l
"0
"0
0
II OV I tJ~
~ rO ~
~4
~.~
.l..J , r l c~
0 ~
m~
0
";-4 0
~-~
,.-.-t 0 I
4~ ,'~
I 0
I 0 ~ ~
I 0 N 0
r
r-~ ~-I r
0
n~
0
0 r-~ 04
alloy (2419-T851) in the aged condition [4], a titanium alloy (Ti-6AI-

4V) in the annealed condition, and an 18 NiCoMo steel [AL MAR 18
(200)] in the maraged condition. These materials are referred to here
as aluminum, titanium, and steel, and they are coded 4, 5, and 6,
respectively, for purposes of computer sorting and analysis and for
presentations made in the Appendices. They are also coded A, T, and S
in some data plots. Some mechanical properties of the materials are
given in Table 3. Properties listed in Table 3 are those used for
calculations made in the preparation of this report, after the
completion of all slow-bend tests.
Reference values of fracture toughness for these materials were

furnished from tests conducted with the Compact Specimen loaded in
tension, C(T). The results of these tests, given in Table 3, indicate
that not all test results meet the validity requirements of E 599.
The tests were conducted in accordance with ASTM Method E 399-74 [5];
and according to the Method, the results are either valid and are
referred to as KI= plane-strain fracture toughness, or they are not
valid and are referred to as ~ . Table 3 gives the particular
requirement that was not met for each test that failed the E 399
validity requirements. For simplicity in this writing, these
reference values of K will be referred to as Kic values, even though
some of the results are actually termed K~ test results by Method E
399. In this report, these reference values are used to assess the
accuracy of the various responses (K values) computed from results of
Charpy tests.
Specimens used for the reference values of KI= were used for
machining of the Charpy specimens tested under this program, except
for the aluminum Charpy specimens, which were taken from the same cast
as the plate used to prepare the compact specimens [6]. For aluminum,
the two reference KI= values given in Table 3 for the C(T) tests are
averaged to obtain a single reference for all aluminum Charpy
specimens. For titanium, the C(T) test results represent two specimen
orientations, longitudinal (LT) with values of 77.9 and 80.2 ksi
(in) % , which are designated R and L respectively (see Table 3), and
transverse (TL) with values of 77.4 and 85.5, which are designated W
and T respectively. In the analyses presented here, the reference for
each Charpy specimen of titanium is the Kic result of its parent C(T),
i.e., the specimen from which the Charpy specimens were machined. For
steel, three C(T) test results are averaged to obtain a single
reference value of 120.5 ksi (in) % for steel Charpy specimens.
2.2 Laboratory Variabilfties
Various task were undertaken by each of eight laboratories that

participated in the Phase I program [6]. One objective of this
program was to conduct the Charpy tests as though only a single
laboratory had done the work. In keeping with this objective, only
one laboratory can be involved for each step of the procedures of
preparation and testing of specimens, and analysis of the data, except
for those steps for which ASTM prescribed methods are applicable. In
practice, this objective was largely met even though some
discrepancies did arise [61 as described in Section 2.4. Other
procedures with potential for giving rise to undesirable lab-to-lab
procedural effects were generally controlled closely enough so that
.I.I
0 0
~ no v
m ~
[-,.i ~
~4
v 0
0 II
9N r-~
v t~
o~ ~ ..~ r~.
,.~ r/l
~0
9 9 ~
~ 4.1
0
II
P-~ ~
0
v
' 0 ke3 ke3
0
v
d
cq ,~
0 4~
m 0
1-1 t~
~ ~ ~ ~ o,~, ~'~ ~
m ~..~~
r~O r~u'3
r~ v ~J
L~ O'3
f',l r
"~ b v I
9i-I o ~ -el ,.~
0 r
0
b vI
I
I
0
I ~ ,~ :
tl) "4"
~ . ~
C o p y r i g h t b y A S T M I n t ' l ( a l l r i g h t s r e
D o w n l o a d e d / p r i n t e d b y
U n i v e r s i t y o f W a s h i n g t o n ( U n i v e r s i t y
they are considered to have had no significant effects on the test

results.
2.3 Notch-Preparation Factor
The notch-preparation (NP) factor in the experimental design has

three levels, as shown in Table 1, which are coded 1, 2, and 3 for
purposes of computer sorting and analysis of the data.
2.4 Stress-Intensity Factor at Start of PrecrackinK
The stress-intensity factor at the start of precracking (Kf

maximum) was controlled at three levels for each material in the
proposed test matrix. The three levels are dependent upon properties
of the materials, as described below. To compute the proposed levels,
first a stress-intensity-factor parameter here called "K smallest" is
computed by using three formulae:
K(I ) = (Gyl/ay2) x Kic (ksi (in) %) (I)
K(2 ) = 0.002 x E (ksi (in) %) (2)
and
K(s > = 0.57 oy1 (ksi (in)~), (3)
where GYI and ay2 , which are equal in this case, are static yield
stress in ksi at the precracking temperature and at the Charpy slow-
bend test temperature, respectively, and E is the elastic modulus in
psi. The smallest of these calculated K values is called K smallest.
Then, loads at the start of precracking are computed to give the
following proposed values of Kf maximum (the value at the start of
precracking in a constant-deflection machine): K smallest, (2/3)K
smallest, and (I/3)K smallest. These Kf maximum values are coded 3,
2, and i, respectively, for purposes of computer sorting and analysis
of data within each material. This is shown in Table 4 which gives
the fatigue precracking levels, and their codes used for sorting and
analysis of data within a material. While three levels (coded 3, 2,
and i) are indicated (in the table) for analyses within each material,
a total of 4 levels are indicated for combined materials: aluminum and
steel have levels 3, 2, and i, and titanium has levels 2, i, and 0.
This coding became necessary for combined results because the actual
K~ maximum levels used for titanium are different from (lower than)
the proposed K~ maximum levels. The highest actual level for titanium
is 20 ksi(in) % , which is nearly equal to (2/3)K smallest, the proposed
code 2; the next actual level for titanium is nearly equal to (I/3)K
smallest, which is the proposed code I; and the lowest actual level
for titanium is much lower than any other levels used in these tests.
Thus, for combined materials, the analysis requires four levels, coded
3, 2, i, and 0.
Further, as noted in Table 4, for steel each of the three actual

levels covers a range that is, in general, slightly higher than the
proposed level. In addition, for steel specimens, K~ maximum was
measured at the finish of precracking, whereas levels of Kf maximum
for aluminum and titanium were measured at the start of precracking.
.~ 0 0 o
~1"~
0 .4
o
t~ o
O
9 . o
,-I C, 0
I I I
m O ,-~ o o r
O -~-I O
O ~
~ q.~ .IJ
4S~ SoS A A
~J
t4~ Q)
O O O
N-N
9 9 ~
4-~ ~ ~
m m O
,-,
o
`4
q4
22 TM
.m &
m O
-e
4~ O
-r4 O
6J 13..,
r-4
u ~ -,4 ",4
03
2~g
H ,.=
,I.J
m
, ~ cN ,--~ C'h
i ,-.-t , N
!l ~ 4J ~
O ,--I
c.4 1'4
U) ~ ~ O0
c,'h c~l ~.-I o

ko -~1
~o
In the precracklng machines used for this Phase I program, K decreases

slightly as crack length increases, so that Ks at the finish of
precracklng is slightly lower than that at the start. Thus, for a
short crack length, Kf at the start of precracking would be slightly
greater than Kf "maximum" tabulated for the steel; and for longer
cracks, K s maximum at the start of precracking is greater still.
Thus, actual Kf maximum levels for each material represent different
fractions of the K smallest computed for the proposed levels. These
fractions are called the Kf ratio and they are greatest for the steel
specimens and least for the titanium specimens, and they are grouped
into 3 levels for each material and into 4 levels for combined
materials.
2.5 ~rack-Size Factor
The test matrix gives three levels for the crack-size factor (~)
used in calculations. These levels of a are coded as follows:
Code i: ~ = 2.5 to 3.6 mm (0.097 to 0.140 in)

Code 2: ~ = 3.6 to 4.6 mm (0.1405 to 0.180 in)
Code 3: ~ = 4.6 to 6.1 mm (0.1805 to 0.242 in)
The lower bound for Code 1 was established as follows. After

testing had been completed, it was observed that a crack was never
initiated in some specimens fatigue cycled at the lowest level of K s
maximum; further, during a preliminary screening of the data, results
for specimens with ~ less than 2.48 mm (0.097 in) were found to be
more highly variable and in this respect inconsistent with the data
for specimens with ~ between 2.48 and 6.15 mm (0.097 and 0.242 in).
Data for specimens below this limit are not included here. No upper
limit was warranted for ~ values to be included in this analysis.
2.6 6ctual Number of Tests Conducted
The actual number of tests conducted for each test condition is

given in Table 2. Difficulties encountered in the process of
precracking at the lowest levels of the stress-intensity factor, Kf,
are responsible for the differences between the proposed and actual
number of test specimens tested. Deficiencies in the numbers of tests
for each of these materials are mainly in the K s code I level, and the
analysis of the results for Kf code 1 are adversely affected by these
deficiencies.
3. ~est Procedures
Precracked Charpy specimens were tested in three-point bend

tests, using a bend test fixture with the geometry and dimensions
recommended in ASTM Designation E 23-72 for Charpy impact testing. No
movable support pins are used in this fixture, which uses two anvil
blocks to provide support at a fixed span. The cross-head speed was
2.5 mm/min (0.I0 in/min). Load and displacement were measured and
plotted. The load was taken from the load cell of the test machine
and the displacement was measured using a transducer (LVDT) placed
between the top and bottom plates of the bend test fixture.
Load P, displacement 6, and energy absorption E (area under the P-6
plot) were measured using a digitizer to trace along the P-6 plot of
the test record. From the raw data, various fracture-toughness test
responses were computed. These responses are here designated ~ ,
KQ_pM , ~ , K~, I<3, K~, and R,b , and are computed by methods described
here, in section 4. As is shown in Figure i, each response is based
on one of three principal measurements, total energy (ET), energy to
maximum load (EM or E~) or a single value of load (F M or FQ).
To compute the response ~ , the total energy E r (Fig. IA)

is measured. To compute the responses Kj, K~, and K~ (Fig. IB),
several measurements were taken: maximum load PM, compliance DI/PI,
and energy to maximum load, E s . The responses Rsb , KQ_FM, and KQ
(Fig. IC) require measurement of either the load PM, or the load PQ
measured by the 5-percent-secant method, as described in Method E 399.
For all aluminum and titanium specimens, and for most steel
specimens, the load-displacement plots were observed to be of the type
(general shape) shown in Figure IA, in which there is no indication of
a cleavage initiation event. For some steel specimens the plots were
bimodal, indicating a cleavage fracture with rapid machine unloading.
During the unloading of the first mode of the bimodal type, the energy
of the test machine is released to the specimen and it was not
recorded on the test record. Thus, while the area under each of the
two modes is included in the measurement of ET, this energy of
unloading, which contributes to the fracture process, is not included
in this measured value. The bimodal load-displacement plot is
observed only for steel specimens with crack size less than 3.56 mm
(0.140 in), i.e. small crack lengths. The ET values used in this
analysis were not corrected to take machine unloading into account; as
the analysis had been nearly completed when this was discovered, time
constraints precluded this correction.
4. Calculation of Response from Charpy Test Results
Raw data taken in precracked slow-bend Charpy tests are not

directly used to assess material properties. Rather, data are
converted to responses of fracture toughness, designated by the symbol
K, or specimen strength ratio, Rsb. These responses are calculated by
various methods given below. The computed K and Rsb results for all
specimens included in this analysis are given elsewhere [6].
Relationships used to compute responses of K are as follow:
K 0 = (PQS/B W ~Iz) f(a/W), (4)
where S = 1.574 and

3(a/W) % [I.99-(~/W)(I-~/W(2.15-3.93 ~/W 2.7 Az/WZ)]
f(a/W) =
2 (I+2~/W) (I-~/W) 3 ! z
after ASTM Method E 399 [5].
KQ_pM = (PNS/B w3/2) f(a/W), (5)
which is the same as the equation for K but maximum load PH,
rather than PQ, is used;
K~ = [0.5 E Er/A (I-~2)] %, after Ronald [7], (6)

Kj = [2 E EM/A (i-~2)] % after Rice [8], (7)
K~ = [2 E E~ /A(I-~2)] %, after Rice [8], (8)
Figure IA
Measurement --,--Response
I
ET---~-K Q
LOAD,P
DISPLACEMENT,
Figure IB
PM EM--~K J
EM Kj and K~!
TP1
i
Figure IC
PQ-'- KQ
PM--~'Rsb and KQ.pM

PQ
l
Figure i. Fracture toughness responses and the principal measurements
used to compute them.
where F~ is EM corrected for the compliance of the test machine:

E~ = F~ - P~ Cm/2 ; where C m is machine compliance, after EPRI
procedures [9,10]; C m = C - C a . The C and C,, respectively,
representing experimental and specimen (theoretical) compliance
values. Measured values of C are given as DI/PI elsewhere [6].
6 Y* (a) ~ P*,
K~ = B W (9)
where
Y* = 1.93-3.07(~/W) + 14.53(~/W) z - 25(a/'W) a + 25.8(a/-W) 4
P* = [2 E~ ICs]%
It is noted that K~ can also be computed using P* in place of PQ in

Equation 4. The lower-bound (or equivalent-energy) procedure (K~)
arises from concepts developed by Witt [ii].
Responses of strength ratio in slow-bend testing, Rsb were

computed using the equation
Rsb = 6 PM W /B (W-~)2ay, (lO)
after Method E 399 [5].
5. ~atistical Tests
Responses computed using equations (4) through (i0) are used here
to determine statistically whether or not the level for each of the
four factors significantly affects the test result. The accuracy of a
Charpy test response is assessed, as described in Section 5.1.3 for
each of the various fracture toughness measures of having the symbol
K. This is done using an appropriate reference value of Kit for each
material. The reproducibility is assessed, as described in Section
5.1.2, using replicate responses.
The controlled variables, called factors in this analysis, are

notch preparation (NP), fatigue-load at the start of precracking (K~
maximum), and original crack size (~) and material. The goals of this
experiment do not include between material differences. Material-to-
material differences are known to exist and in this analysis this
variable is not considered as a factor in the analysis. In the
proposed test matrix, each factor is tested at three levels only;
however, for the factors ~ and Kf maximum, many more than three levels
were actually tested. A variety of analytical methods are used here
to determine whether or not statistically significant difference
exists among the levels of a factor. These include (i) the Kruskal-
Wallis (KW) test, (2) multiple linear regression (MLR) analysis, and
(3) graphical analysis of variance. The results of both the KW test
and the MLR analysis is a Cumulative Probability Value (CPV) which
corresponds to a percent point of the null distribution. The two
tests are not identical and were run because they are sensitive to
different aspects of the same problem [6]. To determine whether a
factor is significant, the MLR analysis is conducted using the

individual responses. The KW test used ranks of these responses and
each factor is subdivided into three levels to test whether the levels
are significant. These differences are expected to give rise to
differences in the CPV results for the two tests. In this analysis,
when either of these tests indicates a factor or its levels to be
significant, then graphical analysis is used to further describe the
effect.
5.1 Kruskal-Wallis Test
One procedure used here to carry out the test of significance is

based on the Kruskal-Wallis [12] test statistic, H. Calculated values
of H which fall out in the extreme regions of the null distribution
are deemed to be indicative of a false null hypothesis--thus, the
levels within a factor are concluded to be significantly different.
Associated with any given value of the Kruskal-Wallis test statistic,
H, is a cumulative probability value (CPV). If the null hypothesis H o
is true, one would expect CPVs generally between 0.0 and 0.90. If H o
is false, one would hope to obtain CPVs larger than 0.90. In the
present analysis, all cases with CPV > 0.90 will be discussed and the
values are expressed in percentage points (i.e., 90 percent rather
than 0.90). This CPV is reported here as the result of a test for the
equality of the levels within a factor. The CPV is rather simply
related to the probability of erroneously concluding that the
difference between levels is significant.
5.1.1 Significance -- For each factor, it is of interest to test

whether the various levels of that factor give the same result or a
significantly different result. If the different levels do not
(within random error) give the same value, the factor is said to be
statistically significant. In this report, a test of significance is
applied independently to data sets representing each of the materials
tested and to data representing all materials (of Phase I) combined
into a single set. The result of the test of significance is a
determination of whether the levels of a factor are significantly
different.
The parameters used to conduct the KW statistical test are the

standard deviation for replicate responses, s, the individual test
responses, Yi, and the mean response, Y. Responses derived from these
three parameters are the specimen strength ratio R,b and the percent
relative deviation (%RD) which is computed relative to the reference
Klc value given for each material, as described below. The %RD values
are given for each of several measures of fracture toughness (Section
4) designated K6, KQ_pM , ~ , K~, Kj, and KS.
When the test of significance indicates that significant

differences exist among the responses for the various levels of a
factor, the question becomes which level is best. For the responses
given above, low values of s and I%RDI indicate, respectively, better
reproducibility and more accuracy. Thus, they are considered better
than high values of s and I%RDI; therefore, the best level can be
determined for computed responses of K by selection of the level with
the lowest values of either s or I%RDI. However, for the computed
values of specimen strength ratio, R,b , the best level can be
determined only from the reproducibility parameter s, because no

reference values of R,b (from which to make a determination of
accuracy) are available for this analysis. Hence, for R,b responses,
the parameters Yl and Y are used only to determine of the significance
of differences (in the Rsb responses) among the levels of each factor,
and the question as to which level is most accurate cannot be
addressed without reference values of R,b.
5.1.2 Reproducibility Reproducibility of the test results is

- -
estimated from replicate responses. Let Yl denote the individual

response for a group of replicate specimens. A mean response, Y, and
a standard deviation, s, are computed for each set of replicates in
accordance with the following formulae:
N
= : (Yl)/N, (11)
1
and
N
s = [7. (y~ - Y) 2/(N-1)]~ (12)
1
where N = the number of responses in the group (see Table 2). For the
test of significance, a CPV is computed from a data set for each
factor. The set includes the standard deviation, s, responses and
their corresponding levels for one of the factors.
In this way, reproducibility evaluations are made for each

material and for data combined for more than one material. The
response parameter in each case is the value of s, converted to rank
within a material. For combined data, this same rank (of s within a
material) is the response parameter. The variance of data for each
material to be combined is different, i.e. the pooled s for all
responses of a material differ for the various materials evaluated.
In the test of significance, each s response is assigned a rank that
depends performance within one of the materials, so that the effects
of differences among the variances of the three materials are
effectively eliminated, when the CPV is computed.
5.1.3 Accuracy -- The test of significance is an estimate of the

accuracy of the responses for each level of a factor whenever the CPV
is obtained from computed estimates of K for which reference Klc
values are available. The estimator of accuracy used here is called
the percent relative deviation (%RD). This estimator of accuracy is
based upon the difference between a value of fracture toughness, K,
computed by one of the described methods (Section 4) used for
precracked Charpy specimens, and the reference value (Klc) computed in
accordance with ASTM Method E 399 [5] for large compact tension
specimens. The %RD is this difference expressed as a percentage of
Kz= , or (as was noted earlier) of KQ that is here called KIr for
simplicity. The %RD has an advantage in studies of this type. It can
sometimes be used as a measure of the accuracy of the response. The
formula for relative deviation is
%RD = (K - Kzc I x i00. (13)

KIr
The test of significance is conducted independently for each of

two response parameters: Yl, the individual response, and Y, the mean
response for replicates. This was done s o as to assure that the
conclusions of the analysis were not dependent on a single approach.
This follows a general principle of data analysis which states that
perturbations in the analysis should be introduced whenever possible
so as to assess the sensitivity of the conclusion to various facets of
the analysis.
The accuracy of responses for combined materials is given a test

of significance in a manner similar to that described earlier for the
reproducibility responses, except that both replicated and
nonreplicated responses are included in the data set and two tests of
significance are made. The response for one test is the raBk of Yi
within a material and the response for the other is the rank of Y
within a material,
5.2 Multiple Linear Regression
A procedure based on linear regression is used to complement the

Kruskal-Wallis analysis. It is important to carry out alternative
analyses (such as the regression procedure discussed in this section
and the graphical procedure to be discussed in the following section)
so as to assure that conclusions are not approach-dependent.
Conclusions which do not hold up over all three different approaches
would lead one to suspect their validity; alternatively, a consistency
of conclusions over the three separate approaches is highly supportive
of their validity.
The first step in the regression approach consisted of fitting

the response versus ~ as described by the model:
Y = ~0 + ~I ~" (14)
Due to results from the Kruskall-Wallis test, ~ was chosen as the

first factor of interest. After this linear fit was performed, the
residual standard deviation was computed and noted. A simplest such
t e s t is to n o t e whether the slope HI is significantly nonzero. A
second test is to compare the l-factor residual standard deviation
with the residual standard deviation gotten by fitting the model:
y = ~0+e. (15)
If a significant reduction has occurred, then ~ is interpreted as

being significant. The next step was to augment the l-factor model to
a 2-factor model as follows:
Y = ~o + Jgz~ + ~zKf 9 (16)
The K~ factor was chosen again from Kruskall-Wallis test results. The
residual standard deviation for this 2-factor model was computed. The
appropriate test of significance was then carried out to determine if
a significant reduction occurred in the residual standard deviation in
going from the l-factor model to the 2-factor model. Such a
significant reduction would be interpreted as the second factor (Kf)
being significant. Likewise a final step was to similarly augment the
2-factor model to a 3-factor model:
Y = ~0 + ~1~+~2Kf + ~3 NP- (17)
As before, the residual standard deviation was computed and compared

to the 2-factor residual standard deviation. A significant reduction
would imply the significance of the notch preparation factor.
5.3 Graphical Analysis
The graphical approach is a valuable complement to the Kruskall-

Wallis and regression procedures. The rationale behind the graphics
approach is multifold [6], but one obvious advantage of the graphical
approach is that of communication: Whereas the use of the Kruskall-
Wallis and regression procedures of the analysis may not be fully
understood by some researchers, a properly constructed plot to
emphasize the significance of a factor is easily understood by all.
6. Results
The results indicate the following: (i) All seven computed

responses are linearly related to crack size and the sensitivity to
crack size varies with the choice of response parameter and with
material. (2) Precracking at either very high or very low levels of
stress-intensity factor, Kf, are to be avoided. (3) For the three
methods of notch preparation used in this study, no significant
effects (of notch preparation) on the responses were observed, except
for razor-scratched steel specimens. This presentation describes
results for R,b in detail and results for various responses of K more
succinctly. Results are presented for each of the factors: crack
size, Kf maximum, and notch preparation. Finally, the sensitivity to
crack size is discussed for the responses. More detailed discussions
are available [6].
6.1 Anomalies in the Data
A preliminary analysis of the data indicated that responses

computed for test specimens with crack size less than 2.46 mm (0.097
in) should be excluded from the analysis because their variability was
greater than that for specimens with larger crack sizes. As the depth
of the machined notch is about 2.0 mm, and cut-off establishes a
crack-extension of 0.5 mm as a limit below which variability of the
response increased greatly.
Anomalies in the data set [6] required that selected specimens be

omitted from the analysis, e.g. (i) responses computed using total
energy to fracture for steel specimens with the smallest crack-size
factor (~ code i), (2) the titanium code W specimens, which Kf Code 3
(highest level) were excluded.
While the original experimental design proposed three discreet K~

ratios, a wide range of actual Kf ratios (Kz maximum/K smallest) were
used in this test program. This range is illustrated elsewhere [6].
For the steel specimens the three proposed "discreet" levels of the
factor Kf maximum at the "start" of precracking--which were used for
the aluminum and titanium specimens--were not used; rather, the finish
of precracking was varied over a wide range of ratios from 0.33 to a

level of 1.3, which is well above the proposed Code 3 level of 1.0.
Thus, the range of stress-lntensity ratios used in precracklng the
three materials was varied (inadvertently) from less than 0.20 to more
than 1.3, and the levels for only two of the materials are discrete
and are measured at the start of precracklng.
6.2 Results of KW Test of Significance for R,h Responses

The CPV results of the KW tests of significance conducted for the
Rsb responses are presented in Table 5 and described here. Included
are the results for aluminum, "all" titanium (including the Code W
specimens), titanium (including only the R, T, and L specimens), and
steel. In addition, results for the combined materials, aluminum,
titanium (R,T,W,&L), and steel are presented. The results related to
crack-size factor are discussed later. In summary, these cumulative
probability results indicate that the three levels of the factor NP
are not significantly different from one another on the basis of
either the level of the response, Yl or Y, or the reproducibility of
the response, s. Thus it is concluded, on the basis of a
distribution-free analysis of variance, that the levels of the factor
NP do not significantly affect the R,b response for the materials
tested.
The level of the factor Kf maximum is significant for steel and

possibly significant for titanium. This conclusion is supported with
the CPV (Table 5). For steel, the response parameters y• and Y are,
respectively, 98% and 96%, indicating that the effect on the response
Rsb is significant at the 5% confidence level. The CPV for the
reproducibility parameter s is only 25%; thus, reproducibility of R,b
responses for steel is considered to be not affected significantly by
the level of the factor Kf maximum.
For titanium, the Yl and Y results clearly indicate that the

response is not affected significantly by the level of the factor Kf
maximum, but the value of 90% for s is marginal, and required
clarification from a more thorough analysis [6]. The results for
"all titanium" data, which include W specimens, contradict these
conclusions for titanium and this points up the impropriety of the use
of the Code W data. Code W data for titanium were excluded from the
analyses and findings of this report of the Phase I program.
6.3 Results of KW Test of Significance for All Responses
The results of the KW tests conducted for each of the

responses Rsb , K~, KQ_pM, KQ, K~, Kj, and K~ are presented in
Appendices I to III given for factor Kf maximum, notch preparation
factor, and crack-size factor, respectively. Included are data for
each of the materials and for the combined materials.
In general, the results of the KI/ tests of significance for all

responses indicate the following:
(I) The level of the crack-slze factor (E) is significant for all
three materials and for almost every computed response. This is
evident from the results of both the individual response parameter Yi
and from the mean response parameter Y.

(2) The level of the stress-intenslty factor used in precracking
(Kf maximum) is significant for one or more responses for each
material, but the significance l e v e l s a r e generally not as high a s for
the crack-slze factor.
(3) The level of the notch preparation factor (NP) is not
significant.
(4) For selected cases, the reproducibility parameter (s) is
significant for the factors (~) and Kf maximum, in support (i) and (2)
above.
TABLE 5 -- Kruskal-Wallis Test of Significance and Cumulative

Probability Values for Response R,b.
Cumulative
Response Proba- Values
Type of Para- bility for
Statistic meter N Matl. NP Kf a
Single Material
Individual Responses Yl 43 A1 68.9 56.9 >99.9

Mean Responses Y 24 A1 61.2 52.8 99.4
Reproducibility s 17 A1 51.3 70.7 96.8
Individual Responses Yl 44 Ti, all 22.5 i00.0"* 93.4

Mean Responses Y 24 Ti, all 7.3 >99.9 41.9
Reproducibility s 16 Ti, all 57.2 94.3 69.8
Individual Responses Yl 28 T RTL only 57.2 49.89 96.9

Mean Responses Y 16 T RTL only 25.0 20.9 95.1
Reproducibility s Ii T RTL only 40.9 89.8 37.2
Individual Responses Yl 51 S 63.5 98.5 99.9

Mean Responses Y 27 S 35.5 96.0 99.6
Reproducibility s 21 S 77.8 25.5 75.5
Combined Materials Rank* of
Individual Responses Yl 122 A,T(alI),S 61.5 99.3 i00.0"*

Mean Responses Y 67 A,T(alI),S 92.7 99.2 >99.9
Reproducibility s 48 A,T(alI),S 53.5 3.5 99.5
Rsb = 6PmW/B(W-~)2a u
*Rank is assigned within the responses for a material.

**>99.9995
Fig. 2A
ALUMINUM
2O
R i b ond K I c
~ K~ . . . . . . . . . . . . . . .
%RD 0
KQ.pM
-20
KQ
-40 I I I I
.26 .36 .46 .66
~/w, NORMALIZED CRACK SIZE
Fig. 2B
%RD TITANIUM
~ K j
_ _ _ ._Km P_M_
KQ . . . . . ____~ . . . . . -- . . . .
--40 I I I I I I
.26 .:16 ,4t; .KS
"e/w. NORMALIZED CRACK SIZE
F~g, 2C
80 ~-"9-..,.Kj
9 ' "" " " " " " 9 - . . , .... STEEL

. . . . . JK"
"**....., "''' . .....
40 -. . . . . . . K.~ .......... " .....
....... ............... 12...........

%RO
...... KQ-pM
9 " ............. , ......... Klc
....... I~(~ .................... 9r - - . ................
-40 S Z I I !
.26 .36 .46 .66
~/W, NORMALtZEO CRACK SIZE
Figure 2. Results of seven regression analyses for aluminum (Fig.

2A), titanium (Fig. 2B), and steel (Fig. 2C) data showing the
relationship between percent relative deviation (RD) and normalized
crack size (~/W).
6.4 Effects of Crack Size and Material
Responses are shown to be linearly related to crack size. With

increases in normalized crack size, responses of R,b increase and
responses of K decrease. The relationship between response and crack
size depends on the choice of response parameter and on the material.
The CPV values of the KW test of significance highlight the

effect of crack-size factor, ~. For each material, both Yl and Y are
significant at the 5-percent level. The predominant effect of crack
size on the test responses is illustrated by selected plots presented
as Figures 2 and 3. In general, the responses of K (or %RD of K)
decrease roughly linearly with increases in normalized crack size and
response Rsb increases in normalized crack size. The plots and the
results of regression analyses [6] both support the conclusion that
the slopes and the magnitudes of the responses are functions of the
response parameter. In the figures, R,b is the only response with a
positive slope. Further, the magnitudes of the responses of K are
shown to decrease roughly in the order Kj, K~, Kd, ~ , ~ - P M , and ~ ,
but the magnitude and the order are dependent on the material and the
crack size. The sensitivity of the response to crack size is a
function of both the residual standard deviation (RSD) and the slope
(~1) of these plots and it is discussed elsewhere [6]. Sensitivity of
the response to crack size is shown [6] to be a function of the choice
of response parameter and of the material with the steel being most
sensitive to crack size and the aluminum least sensitive to crack
size. In addition, on the basis of the residual standard deviation
aluminum is least variable; for six of the seven responses, RSD is
smaller for aluminum than for either of the other two materials. The
effect of the material is further illustrated in Figure 2 and in
Appendix Tables IV and V. From the regression results, given in the
figures and in the tables, it is seen that, at any crack length the
magnitude of the expected response (%RD of K or R,b) is much greater
(and less "accurate" by this measure) for steel than for aluminum or
titanium, and this effect is especially marked for all of the K
responses. This result opens to question the validity of the
reference value of KI= used for this steel. As was shown earlier
among the C(T) results used to obtain Klc reference values (Table 4),
the steel failed to meet the thickness requirement, whereas the
aluminum and the titanium both passed this requirement. Hence, this
KIc for steel is unique among the KI= reference values used in this
analysis. Further, Figure 3, shows that the ~ data, which are based
on total energy, are inaccurate (as discussed earlier) for the code-i
crack sizes. Thus, they were omitted from the calculations.
6.5 Effect of K~ maximum
The results indicate that precracking at levels of Kf maximum

outside the range of 0.4 to 0,9 times K smallest is to be avoided, as
either the magnitude or the variability of computed responses of K or
Rsb may be greater than that for responses for specimens precracked
within this range. These observed effects of the level of Kf maximum
are somewhat dependent on the response parameter and are generally
dependent on crack-size factor. While conclusions of this analysis
are presumed to be generally applicable to all materials tested, this
presumption could not be completely tested for Kf maximum using the
I N T E R R A N T E A N D F I L L I B E N ON P R E C R A C K I N G V A R I A B L E S 163
available data. The implications of these findings should be explored

for their applicability to recommended precracking procedures for
fracture testing, e. 8. in Methods E-399 and E-812.
A summary of significant results, for all seven responses and for

the three materials, is given in Table 6. A CPV is shown for the F
test of the relevant MLR analysis and for each of the relevant
parameters (Yl, Y and s) of the KW test. The CPV is given for each
siznificant result and for a few others, included for comparison. In
addition, some symbols are presented in this table, to represent the
significance. The symbols Sy and S, represent those cases in which
the Kf ratio significantly affects the magnitude and reproducibility,
respectively, of the response. The symbol "?" is used in selected
cases to indicate a questionably significant result; for each of these
results graphical analysis was conducted to further establish,
illustrate, and describe the effect. The combined results of these
methods of analysis gives a final result indicated in Column B (for
the effect of Kf on the magnitude of the response) or in Column D (for
that of the reproducibility).
The conclusion that precracking at levels below a Kf ratio of about

0.4 is to be avoided comes from results for tests of aluminum and
titanium that indicate either the magnitude or the variability of the
response may be increased at lower values of this ratio. Indications
that precracking very high Kf ratios above 0.9 are to be avoided come
from results for steel specimens, for which the magnitude of the
responses increase with an increase in the K~ ratio over the range of
ratios of from 0.33 to 1.3.
60
LEGEND
ALUMINUM
--- TITANIUM
..... STEEL
40
.o -- ~ S S SSS
"" "..... S S
...... s
0 20 - I'~ I ~ s ........ s ~ - . s ................
Tr T 1' I
_ ^ T " T I
.... g TT T . I
ot .,
.2o I i i i i
I i i i i i i i , i i J i i , i i
t
O.Z 0.3, 0.4 O.fl 0.6 0.7
~/w
Figure 3. Plot of data and regression results for response %RD of K6

as a function of normalized crack size, ~/W, for three materials.
This plot shows the necessity for omission of K 6 data for steel
specimens of crack size code I (~/W < 0.356).
I I I I I I
~o o~
.~ ~ . , ~ ~
I I I I I I ~ ~ I ~ I I I ~ I
~ ~ ~ ~ 0 ~
~c
',~ Cl Ol o 9 o ~ C
I o.o. ~,'] ~*~ o-. I I I ~ I I I ~.~.~.

~ .,-t
R~
0
, ~ ~ f O~
<3",
i A A A
0%
'~i ~ ,,-i o ol
I ~ I
0.~1 ~ 0 I:~
3
.'4 i~
'.o O
4J
u3
6.5.1 Aluminum (material code 4)-- Results for aluminum specimens

indicate that for ~ codes 1 and 2, at a Kf ratio less than 0.4 the
magnitude of the response is greater than at higher K~ ratios. As is
indicated in Table 6, column B, the magnitudes for all responses,
except Rsb , KQ, and K 6 , are significant. This effect of a low Ks
ratio is illustrated in a plot given as Figure 4 for %RD of K~, with
plot characters representing coded crack length. For specimens
precracked at a ratio of 0.33 the magnitude of the response increases
for ~ codes I and 2. This plot typifies and represents four of the
five significant responses for aluminum specimens; these are KQ_pM,
K}, Kj, and ~ . For aluminum specimens, KQ responses were unique
among the seven responses, while the other significant responses
behaved similarly to that shown in Figure 4 for K~.
6.5.2 Titanium (material code 5)-- Results for titanium specimens

indicate that the factor Kf maximum is significant for all seven
responses (see Table 6 columns C and D); it is concluded that
responses for specimens precracked at a very low Kf ratio of 0.20 may
have greater variability than the variability for specimens precracked
at a ratio of 0.40. For titanium specimens, data is available only
for two levels of the Kf ratio and at the lowest level (0.20) the data
are sparse for specimens of ~ code 1 and very sparse for those of
code 3. Thus, the conclusions are somewhat tentative.
40
ALUMINUM
20
, l
v
u_
o 3 23
o
o- 3
N 3 ~
3
-20 3
-40 ,,,I .... i ,, t .... I .... I .... I .... L ....

0.5 0.4 0.5 0.6 0.7 0.8 0-9 1.0 I.I
Kf MAXIMUM / K SMALLEST
PLOT CHARACTERS REPRESENT CODED CRACK LENGTH
Figure 4. Plot of %RD of K} versus the Kf ratio, K~ maximum/K

smallest, for aluminum specimens.
The unique behavior observed for the response parameter KQ for

both the aluminum and the titanium specimens, as discussed in this
section, give the authors pause and we note that in a previous work
[15], the parameter ~ (based on the same 5%-secant intercept used in
the present work) was found to be an inappropriate parameter for
evaluations of fracture toughness in subsized specimens of a heat of
4340 steel at a yield strength of 180 ksi, It is concluded that
may be an inappropriate parameter for evaluations of the fracture
toughness using Charpy tests conducted under conditions used in this
study, i.e. conditions that give a load-displacement trace of the type
in which there is no indication of a cleavage initiation event.
6.5.3 Steel (material code 6)-- Only a limited number of steel

specimens were precracked at a ratio of less than 0.4, so the results
for steel are used here for conclusions concerning the general trends
for ratios above 0.4. In general, the data for steel indicate that
the magnitude of each of the seven responses tends to increase with
increases in the Kf ratio. This effect is most marked for steel
specimens of ~ code 2 and it indicates that precracking at Kf ratios
above 0.9 is to be avoided. In addition, for Rsb , variability of
responses may increase at either high or low levels of the Kf ratio.
The general tendency for increases in the response with increases

in the Kf ratio has been illustrated [6]. Table 6 (column B)
indicates the magnitude of the response is significant for all
response parameters. Although the reproducibility parameter in the KW
test (column C) does not indicate a significant effect, it is apparent
from a plot of Rsb data [6] that for ~ codes 2 and 3, variability is
decidedly smaller at intermediate levels of the Kf ratio. Thus, it is
concluded that both the magnitude and the variability of response Rsb
may be significantly affected by the level of K[ ratio for steel
specimens. The marked effect of Kf ratio on the responses of %RD of K
observed for ~ code 2
specimens is also illustrated for response KQ_pM [6]. Significance
tests (Table 6) indicate Sy (Kf maximum has a significant effect on
the response magnitude) for both the KW tests and the MLR analyses,
except for responses based on energy to maximum load (Kj, K~, and KS),
for which only the MLR analyses lead to a significant effect of Kf
maximum. Graphical analyses for these three responses indicate a
significant behavior only for ~ codes 1 and 2. Thus, it is concluded
that the magnitude of each of the seven responses tends to increase
with increases in the Kf ratio for steel specimens. This effect is
most marked for steel specimens of intermediate crack size (~ code 2)
and it is an indication that precracking at Kf ratios above 0.9 is to
be avoided. In addition, it was shown that for response Rsb,
variability of the response may increase at either high or low levels
of the Kf ratio, and this is another indication that these extreme Kf
levels are to be avoided.
6.6 Effects of Notch Preparation
Significant effects of the level of notch preparation (NP) on the

responses were observed only for steel specimens, for three responses
R,b, ~-PM and ~ . These effects were supported by results of MLR and
graphical analyses, but they went largely undetected by the KW test
for significant differences among coded levels of the factor NP. The
results indicate that for a hard materlal, such as the maraglng steel
used in the Phase I program, razor scratching before precracklngmay
lead to increases in either the variability or the magnitude of the
response. Figure 5 is a plot of ~ - F M data that illustrates an effect
observed in plots for each of three responses R,b, KQ_pM and K6:
Variability of responses for NP code I is greater than that for codes
2 or 3.
Figure 5 also illustrates two effects observed only for responses

Rsb and g~_Fs: (i) the mean and median responses for NP code 1 are
greater than those for codes 2 or 3, and (2) there exists along the
top of the trend band (of each plot) a set of NP code 1 data, with no
data there for the other NP codes.
These effects of NP were not observed for the softer materials,

aluminum and titanium. Our interpretation of these findings is that
some steel specimens of NP code 1 (razor scratched) were somehow
improperly prepared for the precracking process. The net result of
this improper preparation is that something (perhaps an uneven crack
front or perhaps cold work) occurred in the test specimen during
precracking. As a result of this, (i) responses that are a function
of maximum load (Rsb and KQ_pM ) sometimes have greater than expected
magnitude, (2) the response that is a function of total energy
absorbed ( ~ ) sometimes has smaller magnitude, and (3) those that are
a function of either PO or energy to maximum load do not appear to be
affected by NP.
I0 I
1 STEEL
?<
31
1
2
2
23
i 3.
o 0
2 2
%3
~-5- 2
1
13 1
-I0 a
-15L . . . . I .... J .... I .... I,, ,,I .... I ....

0.?.5 0.30 0.35 0.40 0.45 0.50 0.55 0.60
5/w
PLOT CHARACTERS REPRESENT CODED NOTCH PREPARATION
Figure 5. Plot of %RD of KQ_pM versus normalized crack size for steel
specimens.
7. Conclusions and Recommendations
7.1 From a preliminary analysis of the data it was concluded that a

minimum crack extension equal to 0.5 mm may be required. This is
consistent with (but only half the value of) the requirement (of E-399
and E-812) for a minimum crack extension of 1 mm.
7.2 Computed responses are linearly related to crack size.

With an increase in crack size, the response of Rsb increases and the
K responses decrease.
7.3 The accuracy of the response varies significantly as functions of

the crack size and the computed response parameter. Thus,
correlations with KIc , which are commonly made using slow-bend test
results [16], would be expected to vary accordingly.
7.4 The sensitivity of the response to crack size is a function of

both the response parameter and of the material.
7.5 Preeracklng at levels of Kf maximum outside the range of from 0.4

to 0.9 times K smallest is to be avoided, as either the magnitude or
the variability of computed responses of K or Rsb may be greater than
that for responses for specimens precracked within this range.
7.6 Among the three levels of n o t c h preparation tested, significant

effects of the level of NP on the response were not observed, except
for steel specimens. The results indicate that for materials similar
to those used in the Phase I program, similar responses are to be
expected from a standard notch that is either razor scratched or EDM
sharpened or from a sharply (non-standard) machined notch. However,
the results indicate that for a hard material, such as the maraglng
steel used in the Phase I program, razor scratching before precracklng
may lead to increases in either the variability or the magnitude of
the response.
7.7 The results indicate that the response ~ based on a 5-percent-

secant intercept may be inappropriate for characterization of fracture
toughness using precracked Charpy tests conducted under conditions
used in this study.
7.8 While the above conclusions are generally supportive of the

presently recommended precracking practices of E-399 and E-812, it is
recommended that these conclusions, especially those of 7.1 and 7.5,
should be carefully explored by members of Committee E24, so as to
establish whether any modifications of the precracking requirements of
these methods is warranted.
ACKNOWLEDGMENTS
The authors would like to thank two NIST (formerly NBS) workers, Mr.
David E. Schwab for extensive computations and programming assistance
in the computations of the Kruskal-Wallis Test of Significance and Mr.
Sam R. Low for making numerous plots and tablets needed for this
analysis. In addition, this analysis was made possible through the
ASTM Task Group E24.O3.03, its chairman Dr. C. Hartbower and the work
of participating members, M. W. Brennecke, A. Burnett, C. Curll, R. E.
Davies, S. Fisher, and the extensive works of members, T. Ronald and

W. Server. As a result of a study conducted by G. E. Hicho (NIST) for
Subcommittee E24.02 on Fractography for Fracture Testing, an error in
the data set used for this study was discovered and corrected.
REFERENCES
[1] NMAB Committee on Rapid Inexpensive Tests for Determining

Fracture Toughness, Washington, DC: National Materials Advisory
Board, Commission of Sociotechnical Systems, National Research
Council, National Academy of Science; 1976.
[21 Schwabe, J., et al., "Report of Working Group on Instrumented

Precrack Charpy Test for Medium Strength Nuclear Pressure Vessel
Steels (Parts i and 2), IIPC-PVRC Joint Task Group on Fracture
Toughness Properties of Materials for Nuclear Components --
Final Report," Library of Congress No. 77-88087, 1977.
[3] Wullaert, R. A. Olefield, W. and Server, W. L., "Fracture

Toughness Data for Ferritic Nuclear Pressure Vessels -- Final
Report of Research Project 232-i," Vol. I, II and III, Electric
Power Research Institute, NP-121, 1976 April.
[4] Aluminum Standards and Data, New York, Publications Dept.,

Aluminum Association, Inc. 750 Third Ave. New York, NY 10017;
1976.
[5] ASTM Designation E399-74; Standard Method of Test for Plane-

Strain Fracture Toughness of Metallic Materials, in Part I0,
Annual book of ASTM standards. Philadelphia: American Society
for Testing and Materials; 1976.
[6] Interrante, C. G. and Filliben, J. J., "Factors Significant to

Precracking of Fracture Specimens," NISTIR-89-89-4214,
U.S. Department of Commerce, National Institute of Standards and
Technology, October 1989.
[7] Ronald, T., Hall, J. A. and Fierce, C. M., " Usefulness of

Precracked Charpy Specimens for Fracture Toughness Screening
Tests of Titanium Alloys," Metallurgical Transactions 3:1-6;
1972 April.
[8] Rice, J. R. Paris, P. C. and Merkle, J. G., "Some Further

Results of J-integral Analysis and Estimates," ASTM STP 536;
1973 July. 231 p.
[91 Ireland, D. R. Server, W. L. and Wullaert, R. A., Procedures

for Testing and Data Analysis: "Task A Topical Report," Effects
Technologies Inc., Technical Report 75-43, 1975 October.
[lO] Server, W. L., "Impact Three-point Bend Testing for Notched and
Precracked Specimens," ASTM Journal of Testing and Evaluation,
1978 January.
[11] Witt, F. J., "Equivalent Energy Procedures for Predicting Gross

Plastic Fractures," Oak Ridge Nat. Lab., U.S. AEC Report ORNL-
TM-0172,: 1972.
[12] Kruskal, W. H. and Wallis, W. A., "Use of Ranks in One-Criterion

Variance Analysis," JASA, 47(260): 583-618; 1952 December.
[13] Draper, N. and Smith, H., APPlied Regression Analysis, New York,
Wiley and Sons, 1956.
[14] J. J. Filiben, "Dataplot, An Interactive System for Graphics,

Fortran Function Evaluation, and Linear/Nonlinear Fitting,"
Proceedings of the Statistical ComDutinE Section of the American
Statistical Association, 1978.
[15] Jones, M. H. and Brown, W. F., Jr., "The Influence of Crack

Length and Thickness in Plane Strain Fracture Toughness Tests,"
Review of Developments in Plane Strain Fracture Toughness
Testing, ASTM STP 463, Philadelphia: American Society for
Testing and Materials; 1970; 81 p.
[16] Succop, George and Brown, W.F., Jr. E tlmation of KI= from
, '1 S ~
Slow Bend Precracked Charpy Specimen Strength Ratios,"

Developments in Fracture Mechanics Test Methods Standardization,
ASTM STP 632, W.F. Brown, Jr., and J.G. Kaufman, Eds., American
Society for Testing and Materials, 1977, pp. 179-192
APPENDIX TABLE I
Results of Kruskal-Wallis Test of Significance,

Cumulative Probability Values, for the Factor Kf maximum.
Aluminum N R,b ~ KO.pM K0 K} Kj K~
Accuracy
(based upon Yl) 43 56.9 79.5 98.2 83.9 96.0 94.8 95.5
(based upon Y) 24 52.8 78.0 94.7 63.7 88.3 88.2 84.9
Reproducibility
(based on s) 17 70.7 38.1 72.4 43.6 60.4 87.5 53.7
Titanium R, T, & L data)
Accuracy
(Y:) 28 49.8 22.6 3.8 96.1 76.9 80.5 83.6
(X) 16 20.9 36.6 44.0 91.9 84.7 73.4 89.9
Reproducibility
(s) ii 89.8 89.8 95.9 84.7 95.9 98.6 95.9
Steel
Accuracy
(Yl) 51 98.5 52.0 95.4 96.8 52.9 37.6 68.8
(Y) 27 96.0 49.5 70.4 69.6 32.0 13.0 51.7
Reproducibility
(s) 21 25.52 43.4 81.2 40.7 85.1 90.8 74.5
N = 20
Combined I
Accuracy
(Yl) 122 99.3 84.1 98.0 99.9 56.8 50.9 62.3
(Y) 67 92.7 39.6 66.7 98.1 16.1 11.9 31.8
Reproducibility
(s) 49 3.53 31.8 i.i 59.5 20.4 10.7 40.5
M=48
11ncludes specimens listed above for all three materials.

2N = 20 for Rsb only.
3N = 48 for R,b only.
APPENDIX TABLE II

Cumulative Probability Values, for the Factor Notch Preparation, NP.
Aluminum N R,b ~ KQ_ps KQ KJ Kj K~
Accuracy
(based upon Yi) 43 68.9 36.3 68.4 0.0 41.0 50.8 62.5
(based upon Y) 24 61.2 36.7 67.8 4.6 56.9 60.0 66.1
Reproducibility
(based on s) 17 51.3 86.1 67.2 83.6 8.2 56.2 18.4
Accuracy
(Yl) 28 57.5 91.6 37.8 49.2 28.8 25.5 57.5
(Y) 16 25.0 80.1 16.7 2.5 23.9 17.6 50.7
Reproducibility
(s) Ii 40.9 47.7 52.8 6.9 16.0 38.2 2.6
Steel
Accuracy
(Yl) 51 63.5 79.8 15.3 22.6 7.7 7.6 21.7
(Y) 27 35.3 72.1 13.0 23.4 8.3 0.7 26.0
Reproducibility
(s) 21 77.8 ~ 36.9 94.1 58.2 58.2 74.4 77.6
N = 20
Combined I
Accuracy
(Yi) 122 61.3 33.3 61.4 26.8 50.6 47.5 72.1
(Y) 67 59.1 22.6 43.3 21.9 52.7 40.7 73.9
Reproducibility
(s) 49 53.33 40.2 48.0 38.2 43.6 26.1 57.6
M = 48
11ncludes specimens listed above for all three materials.

ZN = 20 for R,b only.
3N = 48 for Rsb only.
APPENDIX TABLE III

Cumulative Probability Values, for Factor Crack-Size, ~.
Al=in~ N R,b K6 KQ_p. KQ K~ Kj K~

Accuracy
(based upon Yi) 43 >99.9 100.0 >99.9 99.4 100.0 i00.0 >99.9
(based upon Y) 24 99.9 >99.9 99.9 95.7 >99.9 >99.9 99.9
Reproducibility
(based on s) 17 96.8 76.9 69.7 57.8 94,6 69.0 96.8
Accuracy
(Yi) 28 96.9 69.3 94.9 98.5 96.8 99.7 69.4
(Y) 16 95.1 76.0 94.1 92.7 96.2 97.8 83.8
Reproducibility
(s) 11 37.2 2.6 8.7 79.2 6.9 3 4 . 6 36.0
Steel
Accuracy
(Yi) 51 99.9 9 8 . 0 100.0 I00.0 100.0 1 0 0 . 0 i 0 0 . 0
(Y) 27 99.6 8 8 . 7 >99.9 9 9 . 9 >99.9 > 9 9 , 9 > 9 9 . 9
Reproducibility
(s) 21 75.52 86.9 78.2 42.0 60.3 8.2 59.5
N = 20
Combined I
Accuracy
(Yl) 122 I00.0 9 8 . 6 i00.0 i00.0 i00.0 i00.0 I00.0
(~) 67 99.9 9 4 . 9 100.0 >99.9 100.0 i00.0 100.0
Reproducibility
(s) 49 99.5473 95.4 90.0 0.0 94.6 42.1 96.3
N = 48
~Includes specimens listed above for all three materials.

2N = 20 for Rs~ calculation
3N = 48 for Rsb calculation
*>99.9995
APPENDIX TABLE IV
Residual-Standard-Deviatlon from Multiple Linear

Regression Analyses for Seven Responses and Three Materials.
(Eq.14) ~+K~ (Eq. 16) ~ K~+NP (Eq. 17)

Material Response ! N res. S,D, v res. S,D. v res. S.D. v
4 R~b 43 0.077 41 0.076 40 0.073 39

5 Rsb 28 0.118 26 0.119 25 0.105 24
6 Rsb 51 0.072 49 0.060** 48 0.059 47
4 K~ 43 2.39 41 2.41 40 2.44 39

5 K~ 28 5.94 26 6.06 25 6.03 24
6 K~ 51 6.56 49 6.49 48 6.46 47
6 KQ 2 36 6.14 34 5.65* 33 5.42 32
4 KQ_pM 43 3.03 41 3.02 40 2.97 39

5 KQ_pM 28 4.70 26 4.77 25 3.18 24
6 KQ_pM 51 3.07 49 2.52** 48 2.38 47
4 K~ 43 3.95 41 3.83 40 3.87 39

5 ~ 28 4.95 26 3.70** 25 3.67 24
6 F~ 51 4.09 49 3.72** 48 3.76 47
4 K~ 43 6.23 41 5.94* 40 6.00 39

5 K5 28 8.67 26 8.64 25 8.64 24
6 K~ 51 7.08 49 6.72* 48 6.62 47
4 Kj 43 5.28 41 4.92** 40 4.98 39

5 Kj 28 8.15 26 8.14 25 8.13 24
6 Kj 51 6.91 49 6.65* 48 6.65 47
4 Kd 43 5.82 41 5.64 40 5.70 39

5 Ka 28 7.60 26 7.57 25 7.49 24
6 Kd 51 6.10 49 5.77** 48 5.65 47
iFor the Rsb responses, for which reference values of the response are
not available, the residual standard deviation is computed from the
response Rsb. For each of the K responses the computation is made
using the % RD of K and the magnitude of the residual S.D. will
reflect this.
21ncludes results for only specimens with ~ > 140 mils.

* Significant at the i0 percent level
**Significant at the 5 percent level
9 ~ ~ ~ 0 . . . . ~ ~ ~ ~ 9 ~
~ ~>~
N ~ O I I I I
~ ~ 0 ~ 1 7 6 1 7 6 1 7 6 1 7 6 1 7 6 1 7 6 1 7 6 1 7 6 1
.r4
~J
~t
q~ M
,z::
I I I
o
I I I I
II
p..,
o c~
O O O O O O O O O O O ~ O O ~ ~ O ~
I I I I I I I I I I I I I I I I I I
4J
0 m
m~ ~ m
I I I I I I I I I I I I I I
rB ~ O
% A
I 0 I~
~ ~
9H r-~ -,.q 4J
-~ II II II
The Specimen: Size
David J. Alexander and Ronald L. Klueh
SPECIMEN SIZE EFFECTS IN CHARPY IMPACT TESTING
REFERENCE: Alexander, D. J., and Klueh, R. L., "Specimen Size

Effects in Charpy Impact Testing," Charpy Impact Test; FaGtors
and Variables, ASTM STP 1072, John M. Holt, Editor, American
Society for Testing and Materials, Philadelphia, 1990.
ABSTRACT: Full-size, half-size, and third-size specimens from

several different steels have been tested as part of an ongoing
alloy development program. The smaller specimens permit more
specimens to be made from small trial heats and are much more
efficient for irradiation experiments. The results of several
comparisons between the different specimen sizes have shown
that the smaller specimens show qualitatively similar behavior
to large specimens, although the upper-shelf energy level and
ductile-to-brittle transition temperature are reduced. The
upper-shelf energy levels from different specimen sizes can be
eompared by using a simple volume normalization method. The
effect of specimen size and geometry on the ductile-to-brittle
transition temperature is more difficult to predict, although
the available data suggest a simple shift in the transition
temperature due to specimen size changes. The relatively
shallower notch used in smaller specimens alters the
deformation pattern, and permits yielding to spread back to the
notched surface as well as through to the back. This reduces
the constraint and the peak stresses, and thus the initiation
of cleavage is more difficult. A better understanding of the
stress and strain distributions is needed.
KEYWORDS: Charpy, fracture, size effects, constraint,

cleavage, critical tensile stress, ductile-to-brittle
transition temperature, upper-shelf energy, slip-line field
theory, finite-element analysis.
Dr. Alexander and Dr. Klueh are in the Metals and Ceramics Division
of Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee,
37831-6151.
179
Copyright 9 of
INTRODUCTION
The Charpy test is widely used for the study of ferrltlc

materials. It is a rapid, inexpensive, and simple test which provides
a qualitative measure of toughness. The large body of data and
experience gained with the use of this test over many years gives added
confidence to interpretation of test results, whether the test is used
for alloy development or for monitoring the effects of irradiation on
the mechanical properties of nuclear pressure vessel steels.
These two areas of research have created an impetus for a

reduction in size of the Charpy specimen. Smaller specimens permit the
measurement of mechanical properties during alloy development when only
limited material is available, yet retain the advantages of simplicity
and convenience of the traditional Charpy specimen. However, the major
reason for considering smaller specimens is the fact that many more
specimens can he irradiated in the space available in radiation
facilities. Approximately eight half-slze specimens or eighteen third-
size specimens can be located in the same space that a conventional
full-size specimen would require. This provides a tremendous advantage
for irradiation effects studies.
The use of smaller specimens raises a number of important issues.

It is well established that these smaller specimens show behavior which
is qualitatively similar to the full-size specimens [1-4]. At higher
temperatures ductile modes of fracture occur and the energy absorbed
tends toward an upper-shelf level. As the temperature is reduced, a
brittle mode of fracture occurs with a concomitant decrease in the
energy absorbed. Thus these specimens show a ductile-to-brittle
transition similar to that observed for full-slze specimens. However,
due to the reduction in size of the specimens, the stresses and strains
which develop in the specimens differ with specimen size, and so the
transition in fracture mode will occur at different temperatures for
different specimen geometries. In addition, the energy absorbed will
obviously vary with specimen size. Therefore, it is not clear how data
generated with various specimen geometries can be compared and related.
The subsize specimen geometries have not been standardized, with
different researchers using different notch geometries for specimens
having the same nominal dimensions. These slight differences may have
significant effects on the stresses and strains, and thus the fracture
process. Finally, it may be possible to analyze these impact tests to
determine the values of material properties such as dynamic yield
stress (ayd) or the critical tensile stress required for cleavage
fracture, the cleavage fracture stress (o~) [4,5]. This requires an
accurate knowledge of the stress and strain distributions in these
specimens, which will certainly vary with specimen size and geometry.
The aim of this research is to compare a large number of data sets

which have been generated with different specimen sizes to see if the
data can be normalized or adjusted to allow different specimen sizes
to be compared directly. Most of the data given below have been
generated at Oak Ridge National Laboratory (ORNL) through alloy
development programs sponsored by the Fusion Energy Program. These
efforts have been aimed at designing steels with improved resistance
to irradiation, both through a reduction in radiation-induced
embrlttlement and an increase in the rate of decay of radiation-
induced radioactivity. Different models proposed in the literature for
ALEXANDER AND KLUEH ON SPECIMEN SIZE EFFECTS 181
normalizing the upper-shelf energy (USE) will be compared. The shift

in the ductile-to-brittle transition temperature (ADBTT) as a function
of specimen size and other material parameters will also he considered.
RESULTS
The subsize specimens were tested on a semiautomated Charpy impact

machine modified for testing small specimens [2,6]. The full-slze
specimens were I0 x i0 x 55 mm with a 45 ~ notch 2 mm deep, notch radius
0.25 mm. The half-size specimens were 5 x 5 • 25.4 m m w i t h a 30 ~ notch
0.76 mm deep, notch radius 0.075 mm, and the third-size specimens were
3.33 x 3.33 x 25.4 mm with a 30 ~ notch 0.51 mm deep, notch radius
0.075 mm. Note that the subsize specimens are not geometrically
similar to the full-size specimens, since the notch is relatively
shallower (notch depth/thickness - a/W - 0.15 for the subslze
specimens, while a/W - 0.2 for the full-slze specimen) but sharper (30 ~
for the subsize vs 45 ~ for the full-size specimen).
The impact data were fitted to a hyperbolic tangent function which

allowed the upper-shelf energy level and the transition temperature to
be determined. The transition temperature was taken at the midpoint
between the upper- and lower-shelf energy levels. Some investigators
[2] have used half of the upper-shelf energy as the transition point:
the difference between these definitions is very small, since the
lower-shelf energies are very low.
The results of the tests are shown in Tables i and 2, which

compare full-slze specimens to half-size and thlrd-size specimens,
respectively [7-12]. Mechanical property data are included. Similar
data from the literature which compare half-size and third-size
specimens to full-slze specimens are given in Table 3 [3,13]. These
investigations employed subslze specimens with notch geometries
identical to those described above. Some additional data [5] from
subsize specimens have been included, although that investigation used
thlrd-size specimens with notches which were wider (45 ~ ) and deeper
(a/W - 0.2) than those of the ORNL specimen (30 ~ and a/W - 0.15). In
addition, the span was reduced from 20 to 13.3 mm [4]. The values
given in Table 3 were read from the figures [5].
DISCUSSION
The effect of specimen size on the USE can be considered by

normalizing the energy by some factor related to the specimen
dimensions. Various researchers have used different normalization
factors [2,3,4] and a "volume" approximation in which the energy is
divided by the nominal volume of the deformed zone beneath the notch
has been shown to give the best results [2,3]. The nominal volume is
given by (Bb) 3/2 where B is the specimen width and b is the remaining
ligament thickness beneath the notch. This procedure gives better
results than using Bb 2 as the nominal volume [4] or using an area
normalization (Bb) [2-4].
The results of using this volume normalization are shown in

Fig. I, which compares the normalized data for subsize specimens to the
full-size specimens. The solid lines in Fig. I indicate a i:i
correspondence between the subsize and the full-size specimen data,
9- |
i i ~ i i ~ i i i i
I=
la
o " | ~ ~ ~ ~ | | ~
i i i m
_~o,,
g~
University of Washington (University of Washington) pursuant to License Agreement. No further reprodu
~. ~o ~ o~ ~ o.-- ~ ~ o~ ~ ~o~
@ @
p~
~o
I~. P~. ~4~ ~ P~ P~. r~ ~ P~ ~-
,4
@
o
g.
i
o
i~. p~ ~ r~ r~ P~ I ~. o P~
University of Washington (University of Washington) pursuant to License Agreement. No furthe
~ v
,,-4
~o
~v
'i ,-;/ ,/, . . . . ,
IJ
p eO
i
:>
,,,I,,4
o~ ,.4
o ~ ~ o~ o o o
m
i i
500 I l ' ~ f /
/
[] 12 Cr-1 No STEELS /
O 9 Cr-I No STEELS
O ABE 9 Cr-2 W STEEL
E 400
\
E - A
_ LOUDEN~ i
_
--)
E
w
co 3 0 0
D
w
N
I
L 200
_J
<
-1-
100 l I -,I
I00 200 300 400 500
FULL-SIZE USE (m J / r a m 3)
500
0 2.25 Cr STEELS /
0 5 Cr STEEL /
E 0 9 Cr STEELS . /
E 400 - r ] 12 Cr STEELS <,~/ -
\
E
j -
w
oO/~ooq3
D 300
W 7 ~ / ~
N
I
O 200
I ~ O ABE 9 Cr STEELS
/ A LOUDEN12 Cr STEEL
~7 LUCAS LOW ALLOY STEELS
100 I I I __
100 200 300 400 500
FULL-SIZE USE (m J / r a m 3)
Fig. i. Normalized upper-shelf energies for subsize specimens vs

full-size specimens. Top: half-size specimens. Bottom: third-size
specimens. The solid line indicates a i:i correlation rather than a
fit to the data.
rather than a fit to the data. As the figures show, this simple
normalization process provides a good means for comparlngthe data from
different specimen sizes for several different steels.
Other methods have been proposed for accounting for the specimen
sizes. Louden et al. [13] have developed a model which normalizes the
USE by a factor which incorporates the specimen width, ligament
thickness, and span, as well as an elastic stress concentration factor
which will depend on the notch depth, angle, and root radius. Thus,
all of the specimen dimensions are included. However, the use of an
elastic stress concentration factor for the upper-shelf regime, where
fracture is occurring only after extensive plastic deformation, and by
a mechanism which is more likely strain controlled than stress
controlled, is difficult to Justify. The results of their
normalization [13] give a correspondence similar to the much simpler
volume normalization used here.
Kumar et al. [14] have developed a model to predict the USE of

full-slze specimens by using both notched and fatigue precracked
subslze specimens. This allows the energy for crack initiation and
crack propagation to he separated. Good agreement for a ferritic
12Cr-IMo-V-W steel (HT-9) was observed. This procedure imposes the
added complexity of testing precracked specimens. Although further
testing is needed, the model is expected to be useful for a wide range
of alloys and the study of irradiation effects also.
A p o s s i b l e problem due to the smaller specimen dimensions may

arise when testing tough materials with high USE levels. At higher
temperatures extensive deformation may occur without fracture
intervening. If the material is sufficiently tough, the Specimen will
bend to such an extent that it will be squeezed out between the anvils
rather than fracturing. This behavior has been observed when testing
stainless steel specimens. The shallow notch and reduced thickness of
the subslze specimens increases the likelihood of this behavior, while
the deeper notch and greater thickness of the full-slze specimen favor
the occurrence of fracture. This may affect the correlation of USE
data, if these different behaviors are present. In addition, some
investigators [3,13] use specimens which are shorter than that
described above, i.e., 23.6 vs 25.4 mm. If the same span (20 mm) is
used in both cases, the shorter specimens may be squeezed through the
anvils more readily than the longer specimens, and thus give a lower
USE. The width and radius of the tup may also play a role, as well as
the span length. Despite these differences, the data from Lucas et al.
[5] can be normalized quite well, as Fig. l(a) shows.
The effect o f specimen size on the DBTT is more difficult to

account for. There is no obvious effect of material parameters such
as the yield strength o n t h e ADBTT caused by a change in specimen size,
as Fig. 2 shows. Abe et al. [3] have noted a qualitative trend that
brittle alloys show larger size effects, although considerable scatter
was observed. In Fig. 3 the subsize specimen transition temperature
is plotted as a function of the full-slze specimen transition
temperature. The data suggest that the subsize specimen transition
temperature is related to the full-size specimen transition
temperature. The solid lines in Fig. 3 have been drawn with a slope
of i, and are not fits to the data. However, these lines do suggest
a reasonable correlation. If the slope is i, then one can write:
HALF-SIZE SPECIMENS
50 I I I ] I
0 9 Cr STEELS
[] 12 Cr STEELS
4O
0
0
F-
b_
30
I
CO
[]
Z
0
F- 20 []
H
co
z OD 0
<
~- 10 [] []
0
0 I I I I I m I
500 550 600 650 700 750 800 850
YIELD STRENGTH (MPa)
THIRD-SIZE SPECIMENS
80
I I
o•60v 0
b- 9
-I- [] 0
CO
z 40
0
0 0
co 0
z 0
n< - 20 -
F.- 0 2.25 Cr STEELS
0 5 Cr STEEL
0 9 Cr STEELS
0 12 Cr STEELS
0 I I I I
500 550 600 650 700 750
YIELD STRENGTH (MPe)
Fig. 2. Shift in d u c t i l e - t o - b r i t t l e t r a n s i t i o n temperature as a

f u n c t i o n of y i e l d strength. Top: half-size specimens. Bottom:
third-size specimens.
25 I I I I / ]
/ I
0 9 C r - 1 No STEELS [] / I
o
_
[]
0
12 C r - 1 M o STEELS
ABE
LOUDEN ~ /
/
_1
q
I
I--
I.- -25
m ~P / -
0
w
N
U) -50
I
d
.<
-r-
-75
-100 I I I
-7.c -50 -25 0 25 50
F U L L - S I Z E DBTT (~
/
50 I
o2 . 2 5 Cr STEELS
o5 Cr STEEL
o9 Cr STEELS
[] 12 Cr STEELS
o
-<> ABE
A LOUDEN
H-
F- ~7 LUCAS
ro
n
9
uJ -50
N
u)
I
O
IZ
.~- -IO0
1
I.-
-150 I I I
-100 -50 0 50 100
F U L L - S I Z E DBTT (~
Fig. 3. Transition temperature of subsize specimens vs transition

temperature of full-size specimens. Top: half-size specimens.
Bottom: third-size specimens. The solid llne has a slope of i, and
is not a fit to the data.
DBTTI/~ - DBTT F + C 1 , (i)

or
DBTT1/3 - DBTT F + Cz , (2)
where DBTTF, DBTT1/2, and DBTTI/3 are the ductile-to-brittle transition

temperatures for full-, half-, and thlrd-slze specimens, respectively,
and C I and C2 are constants. This relationship is very similar to that
suggested by Louden et al. [13] but differs in that the transition
temperatures have not been normalized.
It follows from this formulation that ADBTT will be a constant for

any fixed change of specimen geometry. The existing data suggest that
the shift in transition temperature is roughly 15~ for full- to half-
size specimens, and 50~ for full- to thlrd-slze specimens. However,
it must be emphasized that this approach is strictly empirical. In
addition, irradiation or alloying effects may result in different
shifts rather than merely changing the specimen size. Although the fit
shown in Fig. 3 is encouraging, more testing and analysis is necessary
to examine the validity of this simple relationship. A more rigorous
model of size effects will require a better understanding of the
cleavage process in subsize specimens.
It is generally agreed that cleavage fracture will occur when the

peak tensile stress beneath the notch exceeds the cleavage fracture
stress arc [4,5,13]. The peak stress will be located some distance
beneath the notch root surface, as analysis of notched bars has shown
[15,16]. Abe et al. [3] have presented convincing fractographlc
evidence that fracture initiates at particles some distance beneath the
surface in full-, half- and third-slze specimens. Thus, the smaller
specimens still show fracture at a critical stress level of=. However,
how Ofc might be determined is unclear. A complete understanding of
the stress and strain distributions in these small specimens is
essential to determine the plastic stress concentration factor. At
present, empirical results are used [4,5,13]. However, these
procedures are based on sllp-line field theory, which assumes elastic-
perfectly plastic flow behavior, and plane strain conditions. In
addition, general yielding is assumed to occur at the first deviation
from linearity in the load-displacement trace as the specimen is
loaded. Full-slze specimens of materials which exhibit pronounced
Luders deformation on yielding may approach these conditions [17] which
may justify this approach, but this will clearly not be satisfactory
for smaller specimens of smoothly yielding materials. Analysis of
these specimens requires a better understanding of the constraint and
stress distributions. Full three-dimensional flnite-element
calculations are required for these subsize specimens. Such
calculations are being performed at ORNL, and the results will be
reported separately.
The need for this type of calculation is emphasized by recent

sllp-line field analyses of three-point bend specimens with shallow
notches [18]. These results indicate that deformation from the notch
will spread back toward the notched surface, which relieves the
constraint and thus reduces the peak stresses beneath the notch root.
The critical notch depth for three-polnt bend specimens for fully
constrained yielding through the specimen to the back face rather than
to the notched surface has been shown to be a / W - 0.18 [19]. Note that
the full-size specimen exceeds thls critical depth, while the subsize
specimens do not. Thus, the deformation patterns will be much more

complicated than for the deeper notch. The edge effects for the
smaller specimens will only increase the difference between the actual
behavior and that predicted by sllp-llne field theory. Therefore,
three-dimensional finite element analyses are needed.
CONCLUSIONS
Subsize Charpy specimens offer important advantages for alloy

development and irradiation effects studies through their reduction in
size. However, this size reduction raises concerns about the analysis
of test data. Upper-shelf energies from different specimen sizes can
be compared quite well by using a simple volume normalization of the
energy absorbed during fracture. Understanding the shift in the
ductile-to-brlttle transition temperature as a function of specimen
size requires a better understanding of the stresses and strains in
these specimens, which may be provided by finite element analyses.
ACKNOWLEDGMENTS
The authors would like to thank R. K. Nanstad for helpful

discussions, W. R. Corwln and M. L. Grossbeck for reviewing the
manuscript, and J. L. Bishop for preparation of the manuscript.
This research is sponsored by the Office of Fusion Energy, U.S.

Department of Energy, under contract DE-ACO5-84OR21400 with Martin
Marietta Energy Systems, Inc.
The submitted manuscript has been authored by a contractor of the

U.S. Government under contract DE-AC05-84OR21400. Accordingly, the
U.S. Government retains a nonexclusive, royalty-free license to publish
or reproduce the published form of the contribution, or allow others
to do so, for U.S. Government purposes.
REFERENCES
[1] Grounes, M., in Effects of Radiation on Structural Materials,

ASTM STF 426, American Society for Testing and Materials,
Philadelphia, 1967, pp. 224-59.
[2] Corwin, W. R., and Hougland, A. M., in The Use of Small-Scale

Specimens for Testing Irradiated Material, ASTM STP 888,
W. R. Corwin and G. E. Lucas, Eds., American Society for Testing
and Materials, Philadelphia~ 1986, pp. 325-38.
[3] Abe, F., Noda, T., Arakl, Ho, Okada, M., Narul, M., and
Kayano, H., Journal of Nuclear Materials, Vol. 150, 1987,
pp. 292-301.
[4] Lucas, G. E., Odette, G. R., Sheckherd, J. W., McConnell, P., and
Perrln, J., in The Use of Small-Scale Specimens for Testing
Irradiated Materials, ASTM STP 888, W. R. Corwin and G. E. Lucas,
Eds., American Society for Testing and Materials, Philadelphia,
1986, pp. 305-24.
[5] L u c a s , G . E . , Odette, G.R., Sheckherd, J . W . , and

Krlshnadev, M. R., Fusion Technology, Vol. I0, 1986, pp. 728-33.
[6] Alexander, D.J., Nanstad, R . K . , Corwln, W . R . , and

Hutton, J. T., in Applications of Automation Technology to Fatigue
and Fracture Testing, ASTM STP (to be published).
[7] Klueh, R. L., Vitek, J. M., and Grossbeck, M. L., Journal of

Nuclear Materials, Vol. 103-104, 1981, pp. 887-92.
{8] Klueh, R. L., Vitek, J. M., and Grossbeck, M. L., in Effects of

Radiation on Materials: Eleventh Conference, ASTM STP 782,
H. R. Brager and J. S. Perrln, Eds., American Society for Testing
and Materials, 1982, pp. 648-64.
[9] Klueh, R. L., Metallurgical Transactions A, Vol. 20A, 1989,

pp. 463-70.
[lOl Corwin, W. R., Klueh, R. L., and Vitek, J. M., Journal of Nuclear
Materials, Vol. 122-123, 1984, pp. 343-48.
[11] Klueh, R. L., and Corwin, W. R., Journal of Materials Engineering,

Vol. II, 1989, pp. 169-75.
[12] Klueh, R. L., and Alexander, D. J., Journal of Nuclear Materials

(to be published).
[13] Louden, B. S., Kumar, A. S., Garner, F. A., Hamilton, M. L., and
Hu, W. L., Journal of Nuclear Materials, Vol. 155-57, 1988,
pp. 662-67.
[14] Kumar, A. S., Garner, F. A., and Hamilton, M. L., in Effects of

Radiation on Materials: Fourteenth International Symposium,
(Vol. II), ASTM STP 1046, N . H . Packan, R . E . Stoller, and
A. S. Kumar, Eds., American Society for Testing and Materials,
Philadelphia, 1990, p. 487-95.
[15] Grlffiths, J. R., and Owen, D. R. J., Journal of the Mechanics and
Physics of Solids, Vol. 19, 1971, pp. 419-31.
[16] Norris, Jr., D. M., EnEineering Fracture Mechanics, Vol. ii, 1979,
pp. 261-74.
[17] Alexander, D.J., Lewandowski, J.J., Sisak, W.J., and

Thompson, A. W., Journal of the Mechanics and Physics of Solids,
Vol. 34, 1986, pp. 433-54.
[18] Wu, S.-X., Cotterell, B., and Mai, Y.-W., International Journal
of Fracture, Vol. 37, 1988, pp. 13-29.
[19] Knott, J. F., Fundamentals of Fracture Mechanics, Wiley, New York,

1973, p. 41.
The Test Technique
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
Randy K. Nanstad, Ronald L. Swain, and Reynold G. Berggren
INFLUENCE OF THERMAL CONDITIONING MEDIA ON

CHARPY SPECIMEN TEST TEMPERATURE
REFERENCE: Nanstad, R. K., Swain, R. L., and Berggren, R. G.,

"Influence of Thermal Conditioning Media on Charpy Specimen
Test Temperature," ~harpy Impact Test: Factors and Variables,
ASTM STP 1072, John M. Holt, Editor, American Society for
Testing and Materials, Philadelphia, 1990.
ABSTRACT: The Charpy V-notch (CVN) impact test is used

extensively for determining the toughness of structural
materials. Research programs in many technologies concerned
with structural integrity perform such testing to obtain Charpy
energy vs temperature curves. American Society for Testing and
Materials Method E 23 includes rather strict requirements
regarding determination and control of specimen test
temperature. It specifies minimum soaking times dependent on
the use of liquids or gases as the medium for thermally
conditioning the specimen. The method also requires that
impact of the specimen occur within 5 s of removal from the
conditioning medium. It does not, however, provide guidance
regarding choice of conditioning media. This investigation was
primarily conducted to investigate the changes in specimen
temperature which occur when water is used for thermal
conditioning. A standard CVN impact specimen of low-alloy
steel was instrumented with surface-mounted and embedded
thermocouples. Dependent on the media used, the specimen was
heated or cooled to selected temperatures in the range -i00 to
IO0~ using cold nitrogen gas, heated air, acetone and dry ice,
methanol and dry ice, heated oil, or heated water. After
temperature stabilization, the specimen was removed from the
conditioning medium while the temperatures were recorded four
times per second from all thermocouples using a data
acquisition system and a computer. The results show that
evaporative cooling causes significant changes in the specimen
temperatures when water is used for conditioning. Conditioning
in the other media did not result in such significant changes.
The results demonstrate that, even within the guidelines of
E 23, significant test temperature changes can occur which may
substantially affect the Charpy impact test results if water
is used for temperature conditioning.
Dr. Nanstad and Mr. Swain are in the Metals and Ceramics Division
of Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee,
37831-6151; Mr. Berggren is a consultant, retired from Oak Ridge National
Laboratory.
195
Copyright 9 1990
of Washington International www.astm.org
KEYWORDS: acetone, air, Charpy V-notch, cooling rate,

evaporation, methanol, nitrogen gas, oil, thermal
conditioning, water.
INTRODUCTION
The Charpy V-notch (CVN) impact test is used extensively for

determining the toughness of structural materials. For example, it is
required by the American Society of Mechanical Engineers Boiler and
Pressure Vessel Code [I] for both nuclear and nonnuclear applications;
by Title i0, Part 50 of the Code of Federal Regulations [2] for nuclear
plants; by the American Association of Highway and Transportation
Officers [3] standard for bridges; and by similar international codes
and standards. In the case of commercial light-water nuclear reactor
pressure vessels, CVN specimens are tested prior to operation to verify
acceptable as-fabricated toughness and during operation to monitor
changes in toughness due to neutron irradiation. For both the
preirradiated and postirradiated testing, full Charpy impact energy vs
temperature curves are obtained and used to determine the effects of
irradiation on fracture toughness of the reactor vessel. Research
programs in many technologies concerned with structural integrity perform
similar experimental studies.
The American Society for Testing and Materials (ASTM) Method

E 23-88, "Standard Methods for Notched Bar Impact Testing of Metallic
Materials," includes rather strict requirements regarding determination
and control of specimen test temperature. It specifies minimum soaking
times dependent on the use of liquids or gases as the medium used to
thermally condition the specimen. For liquids, the specimen is required
to remain in the bath at the desired temperature within •176 for at least
5 min. For gases, the soaking time is 30 min. Whatever method is used
for heating or cooling the specimen, E 23 requires that impact of the
specimen occur within 5 s after removal from the medium. The method does
not, however, provide guidance regarding choice of conditioning media,
except to note that temperatures up to 260~ may be obtained with certain
oils. Commonly used media within the testing community include air,
nitrogen gas, acetone, oil, and water. The primary objective of this
experimenta ! study was to compare the effects of these different
conditioning media on the temperature of the test specimen between the
time of removal from the medium and impact. A second objective was the
comparison of test results of the same heat of steel from two
laboratories which showed consistent differences in reported energy
values, especially in the ductile-to-brlttle transition region, one
laboratory using heated air and the other using heated water.
Figures l(a) and l(b) are photographs of the computer-automated test

system used for testing standard and subslze CVN specimens. The system
includes a conditioning chamber where the test specimen is heated with
hot air from a controlled heat gun, or cooled with cold nitrogen gas from
a pressurized liquid nitrogen supply. More detailed descriptions of the
testing system are provided in refs. 4 and 5. Another objective for this
study was the overall characterization of the testing system performance
regarding the use of heated air and cold nitrogen gas.
NANSTAD ET AL ON THERMAL CONDITIONING MEDIA 197
Fig. i. Photographs of (a) Computer interactive Charpy impact test

system and (b) Charpy transfer device, conditioning chamber for heating
and cooling test specimens, and temperature control system.
PROCEDURES
A standard C~'q" impact specimen of a low-alloy steel was instrumented

with five chromel-alumel thermocouples. Figure 2 schematically shows the
locations of one surface and four "buried" thermocouples. The buried
thermocouples were located approximately symmetrically relative to the
notch, at mid-width, and along the longitudinal axis of the specimen.
The thermocouple ends were beaded (welded), then welded onto the bottoms
of 4.76-mm-deep drilled holes 4.76 mm in diameter; ceramic cement was
used to fill the holes and allowed to harden. The surface thermocouple
was tack-welded to the surface directly opposite the notch and at
mid-thickness (Fig. 2). The thermocouples were connected to a Hewlett-
Packard Model 3497 Data Acquisition/Control Unit, containing a
5 I/2-digit integrating voltmeter, which was connected to a
Hewlett-Packard Series 200/300 computer. The maximum error of
temperature measurement for a given thermocouple reading is estimated to
be about 0.5~
For all the conditioning media investigated, comparisons were made

of the temperature changes in the specimen after removal from the medium
into laboratory air, and after removal from the medium directly to the
anvil of the Charpy machine. For the tests in heated air and cold
nitrogen gas, the instrumented CVN specimen was placed in the
conditioning chamber of the testing system. For the tests in liquids,
the instrumented specimen was placed in the liquid bath already
stabilized at the target temperature. In all cases, temperatures were
monitored during conditioning, and withdrawal did not take place until
all five thermocouples had stabilized at the target temperature. The
system was programmed to read all thermocouples at an interval of 0.25 s
and provide a hard copy of the results. The thermocouples were read
sequentially, and the acquisition rate resulted in about a 0.05-s time
difference between readings.
Figure 3 shows the results of conditioning the specimen with heated

air from room temperature to 100~ The spread in temperature readings
from the five thermocouples does not change appreciably over the 8-min
heating cycle, indicating that, for the heating rate used, the surface
thermocouple reading is a reasonable representation of the temperature
in the interior of the specimen. Figure 4 shows similar results of
cooling with cold nitrogen gas to -100~ although the spread in
temperature readings increased somewhat throughout the cooling cycle of
about 4 mln. During CVN impact testing with this system, the specimen
is kept at the target test temperature for i to 2 min to allow for
complete stabilization. To track temperature changes following removal
from the conditioning media, thermocouple 2 was chosen because it is one
of the buried thermocouples located near the region of the specimen where
fracture occurs.
Figure 5(a) shows the results of heating with air to target

temperatures from 52 to I02~ followed by removal of the specimen from
the chamber and immediate placement on the anvil of the machine. The
start time for temperature recording (t - 0 s) was upon removal from the
chamber. The vertical dashed line at 5 s represents the maximum
allowable time specified in E 23 for impact of the specimen following
NANSTAD ET AL. ON THERMAL CONDITIONING MEDIA 199
LEGEND:
1. TACK-WELDED SURFACE THERMOCOUPLE LOCATED CENTER OF
SPECIMEN, BEHIND NOTCH.
2 BURIED THERMOCOUPLE (3/16" DEEP (TYP.)),
7/8" FROM LEFT END OF SPECIMEN.
3 BURIED THERMOCOUPLE, 3/4" FROM RIGHT END OF SPECIMEN.
4. BURIED THERMOCUPLE, 5/16" FROM LEFT END OF SPECIMEN.
5. BURIED THERMOCOUPLE, 1/4" FROM RIGHT END OF SPECIMEN.
Fig. 2. Schematic drawing showing thermocouple locations on Charpy

specimen used for influence of thermal conditioning media studies.
removal from the conditioning medium. Very little change is apparent

even over a 10-s period. Figure 5(b) provides a plot of the temperature
change (cooling) vs time for that experiment. After 5 s the greatest
change is about I~ The accuracy of the thermocouples is estimated to
be about 0.5~ and, thus, the ordering of the temperature changes
relative to the target temperatures are likely obscured by the fact that
the measured changes are of the same order as the measurement accuracy.
Figures 6(a) and 6(b) show similar plots for cooling with cold nitrogen
gas to temperatures from 0 to -101~ After 5 s, the greatest rise in
temperature is about 1.5~ from a target temperature of -101~
The results of heating with oil are shown in Fig. 7. The greatest
decrease in temperature after 5 s is about I~ from a target temperature
of 204~ Similar results were obtained for cooling to temperatures from
0 to -750C in mixtures of methanol and dry ice, and acetone and dry ice,
respectively. After 5 s, the temperature changes were less than Ioc.
A heated bath of acetone at a target temperature of 500C was also
investigated and Fig. 8 shows the temperature changes for the experiments
conducted in acetone. For target temperatures from 0 to -75~
temperature decreases initially occur after removal from the bath; the
same result was observed for the methanol and dry ice. This is the
result of evaporation of the liquids and the resultant evaporative
cooling of the specimen. At 50~ the same phenomenon occurs but with
greater changes in specimen temperature, although still less than 2=C
after 5 s. At 50~ the acetone is near its boiling point and
evaporation occurs rapidly. At the cold target temperatures, the
evaporative cooling effect reaches a maximum at about 5 s.
I I I I I I
400
8
80
8
6o 8
HEATING IN AIR
40 8 O TC4
TC2
<> TC4
20
-u~O ] i i i L i L i-
t 2 3 4 5 6 7 8
TIME (rain)
Fig. 3. Plot of temperature vs elapsed time for all five

thermocouples during heating of the specimen with air in the conditioning
chamber. The measurements indicate that exterior and interior
temperatures of the Charpy specimen are essentially the same during
heating to 100~ in 8 min.
I I I L
-20 --
~-40
I-
,,=,
a.
uJ - 6 0
I- COOLING IN NITROGEN GAS
0 TCI
TC2
0 TC4
- 8 0 --
-t00 L I I ~ L
0 50 t00 t 50 200 250
TIME (s)
Fig. 4. Plot of temperature vs elapsed time for all five

thermocouples during cooling of the specimen with nitrogen gas in the
conditioning chamber. The measurements indicate that exterior and
interior temperatures of the Charpy specimen are essentially the same
during cooling to -100~ in 6 min.
1 ' I ' I 'I I ' I ' I

400 _V V V V V V V V V V--
i
o o o o o 9 0 0 0 0 0
<> <> <> <> O O 0 O O O
I
O [] [] D [] [] O [] D [] O
75 __ I
I
l
A A ~ A ~
1
I
I
n. 50 0 0 0 0 0 0 0 0 0 O--
t~ I
I1- INITIAL AIR TEMPERATURE I
t~ 0 52 ~ I
Z~ 66~ I
0 80% . . ~ - ASTM E - 2 3
25 TIME LIMIT
0 88 ~
I
0 94~ I
V t02"C I
I
o I ,l_,, I i I I I I i I
(a) 0 2 4 6 t0
TIME (s)
0
I I I_
V Z~ I
V A I1
V /",
-0.5 6 , § A
A A
hi
(.9
Z
g~ 0 @ o
A
n
D O
--t.0 o 1~ o
- d
Ld I 0 0
D
INITIAL AIR TEMPERATURE I
V D
~: - t .5
rr 0 52"C I 0 --
bJ Z~ 66~ I 0 D
a. I
hi [] 80~ I e
F" <> 88~ <>
-2.0 0 94% ~rASTM E-23 0--
I TIME LIMIT V
V~02~
I <>
I
-2.5 I / i t I I I
(b) o 2 4 6 8 ~0
TIME (s)
Fig. 5. Plots of (a) temperature, and (b) temperature change, from
buried thermocouple 2, vs elapsed time following removal of the specimen
from the heated air environment to the anvil of the Charpy machine for
conditioning temperatures from 52 to I02~ As shown in (b), the
temperature decrease is about I~ or less after 5 s.
1 ' I ' i 1 ~ I 1 |
0 --0 0 0 0 0 0 O O O O O --
A
0
-50
A A A L~ L~
LI.I
or" I
I
I
n" [] G [] [] I~ [] [] D n O
hi !
el
I
W I
~ --t00 --<> <> <> <> 0 0 <> 0 <> <> <>--
I
INITIAL NITROGEN GAS TEMPERATURE
I
0
A - 5 t =C
[] - 7 7 ~
O~
~ -"TAST M E-23
IME LIMIT
<~ - t 0 2 ~ I
-450 I ~ I I .,.11 I i I i I
(a) 0 2 4 6 8 10
TIME (s)
I I I I
INITIAL NITROGEN
GAS TEMPERATURE
_ 0 0~
A -51 ~
oo [ ] -77~
-102 ~
UJ
(.9
Z
O ~
<
"1- 2
0
LM
n"
6I
I--
LU o
n 1
:Z
I O O
I
O O
o 6I O
~
I
I
I I I I I
(b) 0 2 4 6 8 10
TIME (s)

from the cold nitrogen gas environment to the anvil of the Charpy machine
for conditioning temperatures from 0 to -I02~ As shown in (b), the
temperature increase is about 1.5~ or less after 5 s.
N A N S T A D ET AL. ON T H E R M A L C O N D I T I O N I N G M E D I A 203
200
_+ + + + + $ $ + $ ; ~_
t 5O --O O O O O O 0 O 0 O O--
~ too
I
I
I
O O <> <> O <> <> <> <> <> O
I
50 - - 0
I
O O O O O O O O O O--
I
INITIAL OIL TEMPERATURE L~r---" ASTM
E-23
0 48~ D ~~ 0 |Zl,8~ mTIME LIMIT
o ~ A too"c v 2~ I
I I I I I I I J I i I
0 2 4 6 8 I0
TIME (s)
Fig. 7. Plot of temperature, from b u r i e d thermocouple 2, vs elapsed

time f o l l o w i n g removal of the specimen from the h e a t e d oil b a t h to the
anvil of the C h a r p y machine for c o n d i t i o n i n g temperatures from 48 to
204~ V i r t u a l l y no cooling takes place after 5 s.
I i I i
' i I , r
, %# 8,
I n ~ 0
V V
O I 0
O O O V
V I V
O V
w v ~, v
z _~
INITIAL ACETONE
-- TEMPERATURE o m
0 50'(; I
I o
I- Z~ O~ I
13 -20~ I o
<> -46~ I
-3
0 -61 "C ~,,,~
I
ASTM E-23 0
TiME LIMIT 0 --
~7 -75=C
I i I I I
,
I I i I L
~
0 2 4 6 8 t0
TIME (s)
Fig. 8. Plot of temperature change, from b u r i e d thermocouple 2, vs

e l a p s e d time following removal of the specimen from a h e a t e d acetone b a t h
and a c o l d b a t h of acetone and dry ice to the anvil of the Charpy
machine. For c o n d i t i o n i n g temperatures from 0 to -75~ the temperature
changes are less than I~ after 5 s; for a temperature of 50~ the
temperature change is about 2~ after 5 s. Evaporative cooling of the
s p e c i m e n takes place for b o t h the hot and cold conditions.
The experiments with water as the conditioning medlumwere conducted

at six target temperatures from 52 to 100~ Figure 9(a) shows the
temperature profiles for each of the six temperatures. It is obvious
that the temperature changes are greater as the target temperature
approaches 100~ Figure 9(b) shows the temperature changes vs time and
amplifies that observation. At the two lowest target temperatures, the
cooling is only about I~ after 5 s. At the higher target temperatures,
however, the cooling effects become significant. At 100~ the
temperature decrease is over 9~ after 5 s, and about 19~ after I0 s.
The change from 5 to i0 s is noted to amplify the observation that
significant changes in temperature can occur in very little time at those
temperatures when water is used as the conditioning medium.
Figures lO(a) and lO(b) show the temperature changes which occur at
the various locations in the specimen from a target temperature of lO0~
Thermocouples 4 and 5, near the specimen ends, show greater changes in
temperature than do thermocouples 2 and 3, located near the notch region.
It is likely that greater cooling takes place near the ends because
cooling occurs through the ends as well as the side surfaces. The fact
that the specimen rests on the room temperature anvil near the ends of
the specimen likely has some effect on that observation; however, an
experiment was performed in which the specimen was removed from the IO0~
water and left in still air, that is, not placed on the anvil, and the
cooling rates were about the same as for those moved directly to the
anvil. Thus, the cooling mechanism appears to be primarily due to
evaporation of the water.
It should be noted that, even given the sequential nature of the

thermocouple readings, a IO~C change in 5 s would result in a maximum
difference of about O.I~ between the first and last thermocouple
readings during the 0.25-s cycle. Thus, the sequential procedure used
to read the temperatures does not have a significant bearing on the
observations. The average temperature change for the two buried
thermocouples in the central region of the specimen is about IO~ after
5 s. A simple heat transfer analysis was performed to compare with the
experimental results [6]. Heat losses during the experiment occur as the
result of evaporation, natural convection, and radiation. To ohtain
accurate (• results, a much more sophisticated analysis would be
required because the problem is basically three-dimensional and
transient. A correlation developed by Langhaar [7] was identified as an
appropriate simplified model that combines heat transfer by all three
aforementioned mechanisms. Prior to use of that model, however, a simple
calculation was performed to check the film thickness of water required
on the specimen to result in a temperature drop of 15~ by evaporation
alone. A film thickness of 0.056 mm (0.0022 in) was calculated and,
without considering surface tension and wettability, that result seems
reasonable as a possible film thickness. Then, using the Langhaar model
with the assumptions of the surrounding air at 22~ a wind velocity
(walking) of 0.894 m/s (2.93 ft/s), and a uniform rate of cooling, the
average time (average for specimen temperatures of i00 and 85~ the
model predicts to dissipate the 297 J (heat loss required for a 15~
temperature drop) from the surface area of the Charpy specimen is about
18 s, or about l~ That compares with the observed cooling rate of
about 2*C/s. Considering that a very simplified model was used for this
application, the model calculation demonstrates that the experimental
observations are credible. Regarding the postulated dominance of
[ ' I ~ [ I I l I ' I
I
1OO --V I
V V [
V V
0 0 0 0 0 4,
0 0 0 0 0
0 [] [] 0 [] ,[] o o w g
75 0 []
TI [] [] O
I
hi A A Z~ A A
tr A A A A A
t-
n. 50 __O O O O O 0
I 0 0 0 0 0--
W
(3_ I
INITIAL WATER TEMPERATURE I
BJ I
0 52~
A 66~
25 [] 80~ LIMIT
0 88~ I
O 93~
I
V t00%
I
I
o I J 1 1 I I I , I = I
(a) O 2 4 6 8 t0
TIME (s)
o-6 8 I
2
I I i I I
[] 0 0
[] 0 0
A A O
a a
v 0 i [] A
O" -5
o
v 0 6 []
LU
~9 V
z
<
-1- ' O []
o
-10 INITIAL WATER
TEMPERATURE o 0
0 52~ V O <>
m
O_ A 66~
[ ] 80~
V O
~" - 1 5 88~ O
V
93~
~7 100~ V
I I , 7
(b) 0 2 4 6 8 10
TIME (sec)
from the heated water bath to the anvil of the Charpy machine for
conditioning temperatures from 52 to 100~ As shown in (b), the
evaporative cooling effects increase as the conditioning temperature
approaches 100"C, resulting in a decrease of about 10~ in the interior
temperature of the specimen.
100 -@i , I ] I ' I ' I-
0 m
95 8 z~
0
o []
,~ 9o 0
0
ee O
U.I THERMOCOUPLE READINGS AFTER
&.
REMOVAL FROM tOO~ WATER
w
I-
85 - 0 TC~ I 8 []
I 0 A
TC2
I 0 0
[] TC3 I 0
O TC4 I 0 Q A
O TC5 I o 0 [~_
80
~ - ASTM E-23
TIME LIMIT O
8
I %
I L U., I L I i L I
0 2 4 6 8 tO
TIME (s)
i 1 I ' I l I
O
z~
o
-5-- 8
bJ 9 A
(9 O
Z 0
1- o
o -I0 --
tr A
o
ne A
THERMOCOUPLE READINGS AFT~ER 0 []
~ -t5 -- REMOVAL FROM t 0 0 ~ : WATER 0 Z~
ul 0 mct ] ~' []
Z~ TC2 t 0 8 A
Q
[] TC3 I A
O TC4 O
-20 -- O TC5 LIMIT O
I I I I I I I , I I
(b) o 2 4 G 8 I0
TIME (s)
Fig. i0. Plots of (a) temperature, and (b) temperature change, from
all five thermocouples, vs elapsed time following removal of the specimen
from the heated water bath to the anvil of the Charpy machine for a
conditioning temperature of 100~ Thermocouples 4 and 5, near the ends
of the specimen, experience greater cooling rates due to heat loss from
the ends as well as the sides of the specimen.
evaporative cooling, a simple calculation was performed using a lump

model for purely natural convection in air (using an average heat
transfer coefficient of 8.5 W/m2/~ for all surfaces). That calculation
indicated a cooling rate of O.080~ more than ten times less than that
indicated by the Langhaar model and twenty times less than the
experimental results. The Langhaar model calculations showed, in fact,
that only about 10% of the heat flux from the water on the specimen
surface comes from convection. Thus, the comparison demonstrates that
evaporation of water from the specimen is the primary cooling mechanism.
These experiments show that evaporative cooling can cause

significant changes in the specimen temperature when water is used for
conditioning. Figure ii shows a plot of the temperature decrease at 5 s
after removal from the water bath vs the water bath temperature. The
magnitude of the changes increase and the rate of change increases
rapidly as the test temperature approaches 100~ At 100~ the
temperature change in 5 s is about 10~ while it is less than I~ at a
test temperature of 50~ The effects of the evaporative cooling at the
5-s limit become increasingly significant at temperatures above about
65~ The changes at 7 and I0 s following removal from the bath are also
plotted in the figure and show the same trend. As shown earlier, the
cooling changes which occur when heated air is used as the conditioning
medium are minimal. The temperature changes which occur when using
heated air will become significant at temperatures well above IO0~
however, this investigation was conducted to compare various conditioning
media with heated water and, therefore, was limited to lO0~ Within the
range studied for the other conditioning media, -I00 to 50~ temperature
0
I I I I I
-5 -- D
rr
z~
/x 9
z -10 --
w
TIME AFTER SPECIMEN
REMOVAL FROM WATER BATH
9 5 s (ASTM E - 2 5 time limit) Z~
z
-15 -- Z~ 7s
[] tOs
z
0
-~0 I t I 1 I
50 6O 70 80 9O tO0 tt0
WATER BATH TEMPERATURE (~
Fig. ii. Plot of temperature decrease at various elapsed times vs

bath temperature following removal of the specimen from the heated water
bath to the anvil of the Charpy machine. For water bath temperatures
ranging from about 52 to 100~ evaporative cooling decreases the
interior specimen temperature from 2 to 10~ after 5 s, respectively.
changes at 5 s after removal from the environment were also small, I~

or less. Two exceptions are heated acetone (not recommended) at 50~ and
nitrogen gas at -100~ both of which cooled the specimen about 2~ after
5 s.
The potential effects of these temperature changes on impact

properties is highly dependent on the material being investigated. For
a typical low-alloy pressure vessel steel tested in the mid-transition
region, a decrease of IO~ in test temperature would cause a decrease in
absorbed energy of about 12 J (8.8 ft-lb). In the determination of the
reference NDT temperature, RT~T , for nuclear reactor vessel steels, the
attainment of 68 J (50 ft-lb) at 33~ (60~ above the NDT temperature
is required for the RT~T to be equal to the drop-weight NDT. If the
50-ft-lb criterion is not met, testing must be performed at higher
temperatures until 50 ft-lb is obtained (specific requirements for number
of specimens, retests, etc. are delineated in Subsection NB, Section III
of the ASME Code[l]). Thus, the Charpy impact toughness of steels which
have a ductile-to-brittle transition in the 50 to lO0~ range could be
affected by the evaporative cooling effect; an artificially high R T ~
could be determined. The degree to which this cooling affects such
determinations is dependent on the specific material. In the same way,
the testing of irradiated surveillance specimens from commercial nuclear
power plants may result in an artificially high irradiation-induced
transition temperature shift which can influence the operation of the
reactor vessel. Finally, in the certification of materials where impact
toughness requirements are specified, the impact energy obtained at the
target test temperature would be artificially low. There are, of course,
many factors in these and other examples which complicate the simplified
scenarios described; however, the significant effects of evaporative
cooling on the test specimen temperature are certain.
Regarding the heating of specimens with air and the cooling of

specimens with nitrogen gas, the results showed that the interior regions
of the Charpy specimen achieve the target temperature at about the same
rate as the specimen surface. The need for soaking the specimen at the
test temperature for 30 min seems unnecessarily restrictive. Many
investigators have the equipment (use of buried thermocouples, etc.) to
demonstrate that the target temperature is achieved throughout the
specimen in less time than specified by ASTM Method E 23. It is
recommended that a provision be considered for inclusion in the method
that would allow such users to take advantage of those capabilities in
the conduct of Charpy impact testing.
CONCLUSIONS
A study was performed to investigate the effects of various thermal

conditioning media on temperature changes in the standard Charpy impact
specimen during the time between removal from the environment and impact.
The conclusions are:
I. Conditioning in heated water between 50 and 100~ results in

significant evaporative cooling of the specimen at 5 s after removal from
the water bath; the effects are increasingly greater approaching 100~
where the temperature decrease was 10~
2. The use of heated air up to 100~ results in temperature changes

less than I~ at 5 s after removal from the chamber.
3. The use of oil up to IO0~ results in temperature changes less

than I~ at 5 s after removal from the bath.
4. The use of mixtures of acetone or methanol with dry ice from 0

to -75~ resulted in temperature changes less than I~ at 5 s after
removal from the bath.
5. The use of cold nitrogen gas from 0 to -100~ resulted in

temperature changes less than 2~ at 5 s after removal from the chamber.
6. The use of heated water for specimens tested in the ductile-

to-brittle transition region can have a significant effect on the test
temperature, with the magnitude of the effects on the Charpy impact
toughness dependent on the specific material.
7. A warning to users of ASTM Method E 23 should be considered for

inclusion in the method regarding the potential effects of using heated
water baths for thermal conditioning.
8. A provision should be considered for inclusion in the method

which would give flexibility in the soaking time requirement for gas
environments when the user can demonstrate that the target temperature
of the specimen is achieved in less time than specified.
ACKNOWLEDGMENTS
This research is sponsored by the Office of Nuclear Regulatory

Research, U.S. Nuclear Regulatory Commission, under Interagency
Agreement DOE 1886-8109-8L with the U.S. Department of Energy under
contract DE-AC05-84OR21400 with Martin Marietta Energy Systems, Inc.
The submitted manuscript has been authored by a contractor of the

U.S. Government under contract DE-AC05-84OR21400. Accordingly, the
U.S. Government retains a nonexclusive, royalty-free license to publish
or reproduce the published form of the contribution, or allow others to
do so, for U.S. Government purposes.
The authors gratefully acknowledge I. L. Simon-Tov for heat transfer

analyses, F. M. Haggag, D. J. Alexander, and R. E. Pawel for their
reviews of the manuscript and helpful comments, and J. L. Bishop for
preparation.
REFERENCES
[I] ASME Boiler and Pressure Vessel Code, an American National Standard,
American Society of Mechanical Engineers, New York, 1986.
[2] Title i0, Code of Federal Regulations, Part 50, U.S. Government
Printing Office, Washington, D.C., January 1987.
[3] American Association of Highway Transportation Officials - Material
Specifications, Association General Offices, Washington, D.C., 1974.
[4] Corwln, W. R., and Hougland, A. M., "Effect of Specimen Size and
Material Condition on the Charpy Impact Properties of 9Cr-IMo-V-Nb
Steel," The Use of Small-Scale Specimens for Testing Irradiated
Material, ASTM STP 888, W. R. Corwln and G. E. Lucas, Eds., American
Society for Testing and Materials, Philadelphia, 1986, pp. 325-338.
[5] Alexander, D. J., Nanstad, R. K., Corwin, W. R., and Mutton, J. T.,
"A Semiautomated Computer-Interactive Dynamic Impact Testing
System," Applications of Automation Technology to Fatigue and
Fracture Testing, ASTM STP i092, American Society for Testing and
[6] Simon-Toy, I., Oak Ridge National Laboratory, "Transient Temperature
Distribution in Charpie (sic) Specimen," Letter Report to
R. K. Nanstad, Oak Ridge National Laboratory, November I, 1989.
[7] Langhaar, J. W., "Cooling Pond May Answer Your Water Cooling
Problem," Chemical Englneering , 194-98 (August 1953).
Author Index
A Lowe, A.L., 54
Alexander, D.J., 179
B Mikalac, S., 134
Barcus, S.E., 83
Berggren, R.G., 195
Bertozzi, R., 7 Naniwa, T., 67
Nanstad, R.K., 195
E
Early, J.G., 120

Por~,-o, F . , 7
F
Fields, B.A., 120

Filliben, J.J., 142 Revise, G. , 35
Fink, D.A., 94 Rogers, H.C., 134
Garagnani, G., 7 Schmieder, A.K., 20

Shibaike, M., 67
H Shiota, K., 67
Shiraishi, T., 67
Hanawa, N., 67 Swain, R.L., 195
Holt, J.M., i
I
Tanaka, M., 67
Interrante, C.G., 142 Tani, H., 67
Trippado, R., 7
K
Klueh, R.L., 179

Koester, R.D. , 83 Vassilaros, M.G., 134
Low, S.R., 120
211
Copyright 9 of1990
University by ASTM
Subject Index
indirect calibrations, 7,
35
A l l o y steels, 83, 120, 134, loading parameters, 120,
195 134
Anvils, 7, 35, 67 m e t r o l o g i c a l techniques,
A r t i f i c i a l defects, 7 20
ASTM standards r e f e r e n c e type, 35, 94
E 23 -- M e t h o d s for N o t c h e d sensitivity, 7, 142
Bar Impact Testing of slow bending, 142
M e t a l l i c Materials, 7, International
67, 94, 195 s t a n d a r d i z a t i o n , 20
Bending deformation, 67, 179 Kruskal-Wallis test, 142

Broached notches, 83
C
Lateral expansion, 83, 120
Clinometer, 20
D
N a t i o n a l Standards
Ductile-to-brittle Institute, 1
t r a n s i t i o n , 1 3 4 , 179 N a t i o n a l S t a n d a r d i z a t i o n , 20
Dynamic t e a r , 134 Notch parameters (see a l s o
Impact t e s t e r )
E a c u i t y , 134
f a b r i c a t i o n methods, 83,
Energy d e t e r m i n a t i o n s , 7, 35, 94, 142
54 loading, 120, 134
microstructural
F deformation, 83, 179
physical c h a r a c t e r i s t i c s ,
F i n i t e element a n a l y s i s , 179 83
F r a c t u r e toughness N u c l e a r r e a c t o r testing, 54
e v a l u a t i o n s , 142
Pendulums
Ground notches, 83 a b s o r b e d energy, 67
c h a r a c t e r i s t i c s , 7, 20
description, 1
elevation, 20
Impact t e s t e r f r i c t i o n loss, 20, 67
compliance measurements, 7 load system, 7
data accuracy, 54, 94 o s c i l l a t i o n period, 20
d i m e n s i o n a l parameters, 35, rod angle, 20
54 striking edge radius, 67
direct calibrations, 7, 20, P h o t o m i c r o g r a p h s , 94
35
213
R high energy, 83
irradiated, 54, 179
Regulatory requirements, 54 low energy, 83
Research programs, 1 steel, 142, 179 (see also
Resilience analysis, 35 Alloy steels)
titanium, 142
unirradiated, 54
v-notched, 83, 120, 134,
S c i e n t i f i c e v a l u a t i o n s , 54 195
S l i p l i n e f i e l d theory, 179 S t a n d a r d i z a t i o n overview, 1
Specimen parameters Steel product testing, 54,
absorbed energy, 83, 120, 83
142 S t r i k e r a b r a s i o n , 67
bending c h a r a c t e r i s t i c s ,
35, 67
c o n d i t i o n i n g media
a n a l y s i s , 195 Testing techniques, 1
d e f o r m a t i o n p a t t e r n , 179 Testing t e m p e r a t u r e
e l e c t r o n i c c o n t r o l systems, 7 parameters, 83, 179
loading rate, 120, 134 Transition temperature
m a t e r i a l s specification, 20, variables, 54, 120, 134,
142 195
stress rate, 134, 142, 179
type, I U
S p e c i m e n types
aluminum, 142 Upper s h e l f energy
fatigue pre-cracked, 120, l e v e l , 179
134, 142
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc

Charpy Impact Test - STP 1072

Uploaded by

Copyright:

Available Formats

Charpy Impact Test - STP 1072

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

Charpy Impact Test - STP 1072

Uploaded by

Copyright:

Available Formats

STP 1072

Charpy Impact Test:

John M. Holt, editor

Peer Review Policy

Printed in Chelsea, Mich.

Impact Tester Compliance: Significance, Sensitivity, and Evaluation--F. PORRO,

Comparison of Metrological Techniques for Charpy Impact Machine Verification--

Influence of Dimensional Parameter of an Impact Test Machine on the Results of a

THE SPECIMEN: NOTCHES

Quantitative Comparison and Evaluation of Various Notch Machining Methods and

The Effect of Fatigue Pre-Cracking versus V-Notching on Impact Testing of Charpy

Pre-Cracking and Strain Rate Effects on HSLA-100 Steel Charpy Specimens--

Significance of Precracking Variables for Slow-Bend Charpy Tests---

Specimen Size Effects in Charpy I m p a c t Testing---D. J. ALEXANDER AND R. L. KLUEH 179

THE TEST TECHNIOUE

Influence of Thermal Conditioning Media on Charpy Specimen Test Temperature---

Author Index 211

The Symposium on Charpy Impact Test--Factors and

At the s u b c o m m i t t e e meeting, a g r e e m e n t was reached that

The original goal of having world-wide research

Twelve of the papers presented are being published in

Because the dimensional parameters of the m a c h i n e s are

The specimen was investigated from two points of view:

The influence of the temperature c o n d i t i o n i n g media on

As a result of the various studies presented, ASTM

Prior to the Symposium, one attendee was overheard

I wish to thank the many people that helped to arrange

Thanks are also in order to the people who have been

Reference I: Tobler, R.L., R.P.Reed, I.S. Hwang, M. Morra,

IMPACT TESTER COMPLIANCE: SIGNIFICANCE, SENSITIVITY AND EVALUATION

ABSTRACT: The compliance is very sensitive to internal mechanical

KEYWORDS: compliance, impact testers, pendulum, Charpy specimens,

As pointed out by Bluhm [i] the flexibilties and the softness of

Dr.Porro is Quality engineering supervisor at Ansaldo ABB

The record of the strain of an instrumented tup actually made on

Fig.N.l: load signal from an instrumented impact tester tup

In order to minimize the influence of this vibrational energy on

The rule that governs the energy transformation during an impact

where Ep = potential energy of the hammer (weight * height) to be

Ek = kinetic energy remaining after impact;

The quantity Ee represents the energy stored and lost by the

The energy dissipated as elastic deformation of the loading

where Ee = energy stored by the system hammer/specimen/anvils by

The compliance, or displacement under a given load, can be

After substitution, the formula (2) can be written:

The first, described by Venzi [2], has only experimental

The second, used by Ireland [3], requires an instrumented impact

Following the Venzi approach [2], the pendulum-specimen system

vo--> Jk/h/k/h/k/k/~ * > x

I< equivalent system >I

M = pendulum mass (specimen mass is neglected, as Bluhm [i])

The displacement "x" is the sum of the displacements of the

and the following law relates the three stiffnesses:

where P is the load on the system (action and reaction at the

In order to solve the equation, i.e to obtain the value of Sm, it

The value of Sp for a s t a n d a r d unnotched s p e c i m e n made in AISI

The value of Se can be obtained e x p e r i m e n t a l l y with the following

the s o l u t i o n of this e q u a t i o n is:

The value of Se is d e t e r m i n e d by the equation: