Michael P. Monahan (Editor) Thomas A. Siewert (Editor) - Pendulum Impact Testing - A Century of Progress (ASTM Special Technical Publication, 1380) (2000)

STP 1380
Pendulum Impact Testing:

A Century of Progress
Thomas A. Siewert and Michael P. Manahan, editors
ASTM Stock Number: STPI380
ASTM
100 Barr Harbor Drive
West Conshohocken, PA 19428-2959
Printed in the U.S.A.

Library of Congress Cataloging-in-Publication Data
Pendulum impact testing : a century of progress / Thomas A. Siewert and Michael R

Manahan, editors.
p. cm.--(STP; 1380)
ASTM Stock Number: STP1380
ISBN 0-8031-2864-9
1. Materials--Dynamic testing. 2. Impact. 3. Notched bar testing. 4. Testing-machines.
I. Siewert, T.A. I1. Manahan, Michael P., 1953- II1. ASTM special technical publication;
1380.
TA418.34 .P463 2000
620.1' 125--dc21
00-038123
Copyright 9 2000 AMERICAN SOCIETY FOR TESTING AND MATERIALS, West Conshohocken,
PA. All rights reserved. This material may not be reproduced or copied, in whole or in part, in any
printed, mechanical, electronic, film, or other distribution and storage media, without the written
consent of the publisher.
Photocopy Rights
Authorization to photocopy items for internal, personal, or educational classroom use, or

the internal, personal, or educational classroom use of specific clients, is granted by the
American Society for Testing and Materials (ASTM) provided that the appropriate fee is paid
to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923; Tel: 508-750-
8400; online: http://www.copyright.com/.
Peer Review Policy
Each paper published in this volume was evaluated by two peer reviewers and at least one edi-
tor. The authors addressed all of the reviewers' comments to the satisfaction of both the technical
editor(s) and the ASTM Committee on Publications.
To make technical information available as quickly as possible, the peer-reviewed papers in this
publication were prepared "camera-ready" as submitted by the authors.
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 the peer reviewers. In keeping with long-standing
publication practices, ASTM maintains the anonymity of the peer reviewers. The ASTM Committee
on Publications acknowledges with appreciation their dedication and contribution of time and effort
on behalf of ASTM.
Printed in Baltimore,MD
May 2000
Foreword
This publication primarily consists of papers presented at the Symposium on Pendulum
Impact Testing: A Century of Progress, sponsored by ASTM Committee E28 on Mechanical
Testing and its Subcommittee E28.07 on Impact Testing. The Symposium was held on May
19 and 20, 1999 in Seattle, Washington, in conjunction with the standards development
meetings of Committee E28. The Symposium marks the 100 year anniversary of the inven-
tion of the pendulum impact test by an American civil engineer named S. Bent Russell, and
the research and standardization efforts of G. Charpy during the early part of the 20 th century.
This book includes 21 papers that were presented at the Symposium and two others
submitted only for the proceedings (one with lead author Yamaguchi and the other with lead
author Hughes). The papers are organized into four sections by topic: Background of Impact
Testing; Reference Energies, Machine Stability and Calibration; Impact Test Procedures; and
Fracture Toughness Assessment from Impact Test Data. In addition, the background section
includes reprints of two landmark papers, one published in 1898 and one in 1901, that
describe significant achievements in the development of the test equipment and procedures.
The symposium was chaired jointly by Tom Siewert, of the National Institute of Standards
and Technology, and Dr. Michael P. Manahan, Sr., of MPM Technologies, Inc.
Contents
Overview vii
B A C K G R O U N D OF I M P A C T T E S T I N G
The History and Importance of Impact TestingmT. A. SIEWERT,

M. P. M A N A H A N , C. N. M c C O W A N , J. M. H O L T , F. J. M A R S H , A N D E. A. R U T H
Experiments with a New Machine for Testing Materials by I m p a c t -

s. BENTRUSSELL,Transactions of the American Society of Civil Engineers,
Vol. 39, June 1898, p. 237. 17
Essay on the Metals Impact Bend Test of Notched Bars--G. CHARPY,Soc. Ing.
de Fran~ais, June 1901, p. 848 46
REFERENCE ENERGIES, MACHINE STABILITY, AND CALIBRATION
International Comparison of Impact Verification Programs---c. N. McCOWAN,

J. P A U W E L S , G. REVISE, A N D H. N A K A N O 73
European Certification of Charpy Specimens: Reasoning and Observationsm

J. P A U W E L S , D. G Y P P A Z , R. V A R M A , A N D C. I N G E L B R E C H T 90
Stability of a C-type Impact Machine Between Calibrations--M. SUNDQVIST

A N D G. C H A I 100
Indirect Verification of Pendulum Impact Test Machines: The French

Subsidiary from Its Origins to the Present, Changes in Indirect
Verification Methods, Effects on Dispersion, and PerspectivesmG. GALBAN,
G. REVISE, D. M O U G I N , S. L A P O R T E , A N D S. L E F R A N ~ O I S 109
Maintaining the Accuracy of Charpy Impact Machines--D. e. VIGLIOTTI,

T. A. SIEWERT, AND C. N. M c C O W A N 134
Characterizing Material Properties by the Use of Full-Size and Subsize

Charpy Tests: An Overview of Different Correlation Proceduresm
E. L U C O N , R. C H A O U A D I , A. F A B R Y , J . - L . P U Z Z O L A N T E , AND E. V A N W A L L E 146
Effects of Anvil Configurations on Absorbed Energy--Y. YAMAGUCHI,
S. T A K A G I , AND H. N A K A N O 164
The Difference Between Total Absorbed Energy Measured Using an

Instrumented Striker and That Obtained Using an Optical E n c o d e r - -
M. P. M A N A H A N , SR. AND R. B. STONESIFER 181
On the Accuracy of Measurement and Calibration of Load Signal in the

Instrumented Charpy Impact Test--T. KOBAYASHI,N. INOUE,S. MORITA,
AND H. T O D A 198
Evaluation of ABS Plastic Impact Verification Speeimensnc. N. McCOWAN,

D. P. V I G L I O T T I AND T. A. SIEWERT 210
I M P A C T T E S T PROCEDURES
Results of the ASTM Instrumented/Miniaturized Round Robin Test

Program--M. P. M A N A H A N , SR., F. J. MARTIN, AND R. B. STONESIFER 223
European Activity on Instrumented Impact Testing of Subsize Charpy

V-Notch Specimens (ESIS TC5)--E. LUCON 242
Dynamic Force Calibration for Measuring Impact Fracture Toughness using

the Charpy Testing Maehine--K. KISHIMOTO, H. INOUE, AND T. SHIBUYA 253
Low Striking Velocity Testing of Precracked Charpy-type Specimens--

T. V A R G A AND F. L O I B N E G G E R 267
In-Situ Heating and Cooling of Charpy Test SpeeimensnM. P. MANAHAN,SR. 286
The Effects of OD Curvature and Sample Flattening on Transverse Charpy

V-Notch Impact Toughness of High Strength Steel Tubular Products m
GEORGE W A I D AND H A R R Y ZANTOPULOS 298
Electron Beam Welded Charpy Test Specimen for Greater Functionality--

ROB H U G H E S AND BRIAN D I X O N 310
F R A C T U R E T O U G H N E S S A S S E S S M E N T FROM I M P A C T T E S T D A T A
Application of Single-Specimen Methods on Instrumented Charpy Tests:

Results of DVM Round-Robin Exercises--w. BOHMEAND H.-J. SCHINDLER 327
Relation Between Fracture Toughness and Charpy Fracture Energy:

An Analytical Approach--H.-J. SCHINDLER 337
Use of Instrumented Charpy Test for Determination of Crack Initiation

Toughness--H.-W. VIEHRIG,J. BOEHMERT,H. RICHTER,AND M. VALO 354
On the Determination of Dynamic Fracture Toughness Properties by
Instrumented Impact Testing--G. B. LEN~Y 366
Estimation of NDT a n d Crack-Arrest Toughness from C h a r p y Force-

Displacement Traces---M. SOKOLOV AND J. G. MERrO-E 382
Indexes 395
Overview
Overview
ASTM Subcommittee E28.07 (and its predecessor E01.7) has sponsored six symposia on
impact testing, published in Proceedings of the Twenty-Fifth Annual Meeting (1922), Pro-
ceedings of the Forty-First Annual Meeting (1938), STP 176 (1956), STP 466 (1970), STP
1072 (1990), and STP 1248 (1995). These symposia covered a broad range of topics and
occurred rather infrequently, at least until 1990. The period before 1990 might be charac-
terized as one in which the Charpy test procedure became broadly accepted and then changed
very slowly. However, the last two symposia, "Charpy Impact Test: Factors and Variables"
and "Pendulum Impact Machines: Procedures and Specimens for Verification," were driven
by new forces; a recognition within ISO Technical Committee 164-Subcommittee four (Pen-
dulum Impact) of some shortcomings in the procedure; and a growing interest in instru-
mented impact testing. These STPs, 1072 and 1248, proved to be of interest to many general
users of the test, but were of particular interest to the members of ASTM Subcommittee
E28.07 (the subcommittee responsible for Standard E23 on the Charpy test). During the past
ten years, the data presented at those Symposia have been the single most important factor
in determining whether to change various requirements in Standard E23. The data have also
been useful in supporting tolerances and procedural details during the reballoting of ISO
Standard 442 on Charpy testing, and in the refinement of instrumented impact test
procedures.
Several years ago, the E28 Subcommittee on Symposia suggested that it was time to
schedule another symposium on Charpy impact testing that would bring together impact test
researchers from around the world to share their latest discoveries and to provide input for
further improvements in the test standards. The test was also near its Centenary, and a
symposium to mark this anniversary seemed appropriate. Of course, this fact led to our very
striking title. However, the choice of the date for the symposium was complicated by the
fact that the inventory of the pendulum impact test is S. Bent Russell, while the test bears
the name of G. Charpy. Details concerning the history of the test are reported in the first
paper of this STP. While G. Charpy did publish a landmark paper in 1901 (translated and
reprinted in this volume) and later led the international committee that proved the value of
pendulum impact testing, an 1898 paper by Russell (also reprinted in this volume) was the
first to both describe the mechanics of the pendulum impact machine design and report
impact data obtained using such a machine. The 1898 Russell paper also offers an excellent
tutorial on the contemporary knowledge of the effect of loading rate on impact resistance
(then known as resilience), important variables in machine calibration, and representative
data on common construction materials. The date of the symposium was chosen to honor
the contributions of both Russell and Charpy. As can be seen from a review of the early
papers in this field, it seems as though the turn of the last century marked the time of the
most rapid development and use of impact testing.
As was the previous symposium, the 1999 symposium was successful in attracting con-
tributions from many countries. In fact, the majority (thirty-seven) of the fifty authors and
coauthors are from outside the U.S., a broader international participation than previous
symposia.
ix
X PENDULUM IMPACT TESTING
The future of pendulum impact testing appears bright, as it continues to be specified in

many construction codes and standards. Additional details on the economic importance of
pendulum impact testing were included in an earlier version of our review of the history and
importance of impact testing (the first paper in this STP). This earlier paper can be found
on page 30 of the February 1999 issue of Standardization News, where itwas recognized
as winning third place in the ASTM Impact of Standards Competition. The early history of
impact testing which led to the recognition of Russell as the inventory of the Charpy impact
test was reported in October 1996 issue of Standardization News.
Even after 100 years of use, new aspects of the test continue to be discovered, and of
course, any test can be improved as technology reveals new ways to reduce the scatter in
the test variables. The symposium also reflects the beginning of a new research thrust to
obtain fracture toughness from the Charpy test. It is expected that fracture toughness research,
particularly in relation to the Charpy test, will continue over the next 100 years. We anticipate
many more symposia on impact testing in the future.
Acknowledgments
We appreciate the assistance of Subcommittee E28.07, its Chairman, Chris McCowan, and
its members, many of whom helped by chairing the sessions and by reviewing the manu-
scripts. We also appreciate the assistance of E. Ruth (U.S. Delegate to ISO Committee 164-
TC4 for a number of years) and J. Millane (Secretary of ISO Committee 164-TC4) who
helped to encourage international participation. We would also like to thank the ASTM staff
who helped with symposium arrangements and the other myriad of details that are necessary
for a successful symposium.
Tom A. Siewert
NIST, Boulder, CO;
symposium co-chairman and editor
Michael P. Manahan, Sr.

MPM Technologies, Inc.
State College, PA;
symposium co-chairman and editor
Background of Impact Testing
T. A. Siewert, 1 M. P. Manahan,2 C. N. McCowan,3 1. M. Holt,4 F. 1. Marsh, 5 and E. A.
Ruth6
The History and Importance of Impact Testing*
Reference: Siewert, T. A., Manahan, M. R, McCowan, C. N., Holt, J. M., Marsh,

E J., and Ruth, E. A., "The History and Importance of Impact Testing,"
Pendulum Impact Testing: A Century of Progress, STP 1380, T. A. Siewert and M. E
Manahan, Sr., Eds., American Society for Testing and Materials, West
Conshohocken, PA, 2000.
Abstract: Charpy impact testing is a low-cost and reliable test method which is commonly
required by the construction codes for fracture-critical structures such as bridges and
pressure vessels. Yet, it took from about 1900 to 1960 for impact-test technology and
procedures to reach levels of accuracy and reproducibility such that the procedures could
be broadly applied as standard test methods. This paper recounts the early history of the
impact test and reports some of the improvements in the procedures (standard specimen
shape, introduction of a notch, correlation to structural performance in service, and
introduction of shrouds) that led to this broad acceptance.
Keywords" absorbed energy, Charpy impact testing, history, impact testing, pendulum
impact
Without uniformity of test results from day to day and from laboratory to laboratory,
the impact test has little meaning. Over the years, researchers have learned that the results
obtained from an impact test can depend strongly upon the specimen size and the
geometry of the notch, anvils, and striker: To a lesser degree, impact test results
also depend upon other variables such as impact velocity, energy lost to the test machine,
and friction. The goal of those who have written and modified ASTM Standard Test
1 Supervisory Metallurgist, Materials Reliability Division, National Institute of

Standards and Technology, Boulder, CO 80303.
z President, MPM Technologies, Inc., 2161 Sandy Drive, State College, PA 16803.
3 Materials Research Engr., Materials Reliability Division, NIST, Boulder, CO 80303.
4 Alpha Consultants & Engineers, Pittsburgh, PA.
5 Retired (Bethlehem Steel), San Marcos, CA.
Tinius Olsen Test Machine Co., Willow Grove, PA.
* Contribution of NIST; not subject to copyright. Further details on the economic
impact of Charpy impact testing are included in a previous version of this
report published in Standardization News, February 1999.
3
Copyright9 by ASTM International www.astm.org
4 PENDULUM IMPACT TESTING: A CENTURY OF PROGRESS
Methods for Notched Bar Impact Testing of Metallic Materials (E 23) has over the years
been to standardize and control the variables associated with impact testing. This report
looks at the history of impact testing, with emphasis on the key advances in understanding
and application of the impact test, as reflected in the evolution of the test standard.
Impact Testing: 1824 to 1895
The earliest publication that we could find on the effects of impact loading on materials
was a theoretical discussion by Tredgold in 1824 on the ability of cast iron to resist
impulsive forces [1]. In 1849, the British formed a commission to study the use of iron in
the railroad industry, which began by considering practical approaches to impact testing
[2]. Apparently, failures of structures in the field were leading some researchers to
speculate that impact loads affected materials far differently than static loads, so tensile-
strength data (from slowly applied loads) was a poor predictor of performance under
dynamic loads.
In 1857, Rodman devised a drop-weight machine for characterization of gun steels,
and over the subsequent 30-year period, his machine was widely used to test railroad
steels and for qualification of steel products [2]. Many of the early experiments with
impact tests were performed on final product forms, such as pipes or axles. Thus they
served as proof tests for a batch of material, or yielded comparative data for a new
product design, or basic reference data on the impact resistance of different construction
materials (such as the comparison of wrought iron to ductile iron). Instrumentation was
poor for the early impact tests, so the data is often only as break or no-break for a mass
dropped through a certain distance. These early drop weight tests were conducted using
smooth (no notch or crack starter) rectangular bars. While the test worked well for brittle
materials, where crack initiation is easy, specimens of ductile materials often just bent.
LeChatalier introduced the use of notched specimens while conducting drop-weight tests
in 1892 [3]. He found that some steels that showed ductile behavior (bending without
fracture) in a smooth rectangular bar, would exhibit fragile behavior when the test
specimen was notched. While the addition of a notch was a major improvement in the test
method, a test procedure was needed that would provide a continuous, quantitative
measure of the fracture resistance of materials. Also, substantial work was needed to
develop test procedures that produced consistent data, and to answer the objections of
those who doubted the value of impact testing.
1895 to 1922
This period saw the establishment of a number of national and international standards
bodies, which took up the causes of developing robust test procedures and developing
consensus standards for many technologies, including impact testing. One of these
standards bodies was The American Society for Testing and Materials, established in
1898. Another was the International Association for Testing Materials, officially
established in 1901, but this association grew out of the good response to two previous
International Congresses that had been held a number of years before. These two
standards bodies seem to have had a good working relationship, and the President of
SIEWERT ET AL. ON IMPACT TESTING 5
ASTM, Prof. H. M. Howe, also served on the Board oflATM during this time [4].
In 1902, only four years after the founding of ASTM, the ASTM "Committee on the
Present State of Knowledge Concerning Impact Tests" published a bibliography on impact
tests and impact testing machines in the second volume of the Proceedings of ASTM [5].
This bibliography listed more than 100 contemporary papers on impact testing published in
the U.S., France, and Germany. Many of these papers contained information that was also
known to the members of IATM. In fact, some of the papers had been presented and
discussed at the IATM Congresses.
Among the references is a report by Russell (published in 1898 and reprinted in this
STP) that shows remarkable insight into the needs of the design engineers of the time and
introduces quantitative measurement to the test [6]. He pointed out that none of the
machines of the time, typically of a drop-weight design, had the ability to determine any
data beyond whether the specimen broke or remained intact. Therefore, he designed and
built a pendulum machine which "would measure the energy actually absorbed in breaking
the test bar". His report shows a test machine that is based on the same swinging
pendulum coficept as those in common use today and mentions his careful analysis of the
mechanics of the test, including corrections for friction losses and calculation and
comparison of the centers of gravity and percussion. Since this was before the time of
compact, standardized test specimens, the machine was vary large and massive, and was
capable of breaking many full-size products. Besides showing a prototype of the machines
used today, this report is valuable in that it includes data on over 700 tests of typical
construction materials, and emphasizes the effect of the rate of loading in evaluating
materials for different service conditions. Russell's pendulum impact machine finally
provided a means for quantifying the energy absorbed in fracturing a test specimen for a
wide range &materials and conditions. His paper nicely summarizes the test-machine
technology and knowledge for material performance at the end of the past century, and so
served as a benchmark for future research. To the best of our knowledge, Russell was the
first to develop and demonstrate the advantages of the pendulum design for impact testing
machines.
The members of IATM Commission 22 (On Uniform Methods of Testing Materials)
continued to conduct research that addressed the shortcomings in the impact testing
techniques, until they had developed a knowledge of most of the important factors in the
test procedure. Even though many of these early machines and reports are simplistic by
today's standards, they provided previously unknown data on the impact behavior of
materials. France seems to have been an early adopter of impact testing for infrastructure
construction standards, and so French researchers provided much data on the effects of
procedure variables and were the most prolific contributors to the IATM Proceedings
between 1901 and 1912. Incidentally, it was a representative from France, G. Charpy,
who became the chair of the impact testing activity after the 1906 IATM Congress in
Brussels, and presided over some very lively discussions on whether impact testing
procedures would ever be sufficiently reproducible to serve as a standard test method [7].
Charpy's name seems to have become associated with the test because of his dynamic
efforts to improve and standardize it, both through his role as Chairman of the IATM
Commission and through his personal research [8]. He seems to have had a real skill for
recognizing and combining key advances (both his and those of other researchers) into
continually better machine designs and consensus procedures. For example, Charpy
acknowledges the benefits of Russell's pendulum design in his 1901 paper [8] by stating:
"Russell described in a paper presented in 1897 at the American Society of Civil Engineers
some 'experiments with a new machine for testing materials by impact.' The machine he is
using is designed to determine the work absorbed by the rupture of a bar, for this, the ram
used appears in the form of a pendulum arranged in such a way so that when it is released
from its equilibrium position, it meets the test bar in passing through the vertical position,
breaks it and afterward rises freely under the influence of the acquired speed. The
difference between the starting height and the finishing height of the pendulum allows
evaluation of the work absorbed by the rupture of the bar."
By 1905, Charpy had proposed a machine design that is remarkably similar to present
designs and the literature contains the first references to "the Charpy test" and "the Charpy
method". He continued to guide this work until at least 1914 [7,9-10]. A number of
other standard machine designs and procedures were also under consideration at this time,
and in 1907 the German Association for Testing Materials adopted one developed by
Ehrensberger [10]. Because the pendulum machine had not achieved dominance yet,
impact machine designers and manufacturers offered three major types; Drop Weight
(Fremont, Hart-Turner, and Olsen), Pendulum Impact (Amsler, Charpy, Dow, Izod,
Olsen, and Russell), and Flywheel (Guillery).
This was a period during which the configuration and size of specimens closely
approached what we use today [7]. Originally, two standard specimen sizes were most
popular. The smaller had a cross section of 10 by 10 mm, a length of about 53 mm (for a
distance of 40 mm between the points of support), a notch 2 to 5 mm deep, and a notch
tip radius near 1 mm. The larger and initially more popular of these specimen sizes was
scaled up by a factor of three in all these dimensions. The group favoring the larger
specimen pointed out the advantage of sampling a larger cross section of the material (for
reduced scatter in the data) and the difficulty of producing the small notch radius on the
smaller specimen. However, the group favoring the smaller specimen eventually won
because a more compact and lower-cost machine could be used, and not all structures
were thick enough to produce the larger specimen. Besides specimen dimensions that are
very similar to what we use today, the Commission proposed features for a standard
impact procedure that included:
- limits for the velocity of the striker,

- rigid mounting to minimize vibration losses,
- a minimum ratio of anvil mass and rigidity to striker size, and
- recognition of the artificial increase in energy as ductile specimens deform around
the edges of a wide striker [7].
One report at the 1912 meeting [7] included the testimonial from a steel producer of how
the improved impact test procedures had allowed them to tailor the refining processes to
produce less brittle steel. The report describes a reduction by a factor of 20 in the number
of production parts that were rejected for brittle performance.
SIEWERT E f AL. ON IMPACT TESTING 7
1922 to 1933: The Beginning of ASTM Method E 23
ASTM Committee E-1 on Methods for Testing sponsored a Symposium in 1922 on

Impact Testing of Materials as a part of the 25th Annual Meeting of the Society, in
Atlantic City, New Jersey. The Symposium included a history of the developments in this
area, a review of work done by the British Engineering Standards Association, several
technical presentations, and the results of a survey sent to 64 U.S. testing laboratories
[11]. Twenty-three respondents to the survey offered detailed information on topics such
as the types of machines in use, the specimen dimensions, and procedures. In addition,
many responded positively to a question about their willingness to develop an ASTM
standard for impact testing.
Based on the information in this survey, an ASTM subcommittee began to prepare a
standard test method for pendulum impact testing in 1923. This effort took until 1933,
when ASTM published "Tentative Methods of Impact Testing of Metallic Materials,"
ASTM designation E 23-33T. (An ASTM specification of"Tentative" indicated that it
was subject to annual review and was a work in progress. The tentative designation is no
longer used by ASTM.) (Other countries also developed their own standards; however,
we found it difficult to find their records and to track their developments.)
ASTM E 23-33T specified that a pendulum-type machine was to be used in testing and
"recognized two methods of holding and striking the specimen", that is, the Charpy test
and the Izod test (where the specimen is held vertically by a clamp at one end). It did not
specify the geometry of the striking edge (also known at the time as the "tup") for either
test. It stated that "the Charpy type test may be made on unnotched specimens if indicated
by the characteristics of the material being tested, but the Izod type test is not suitable for
other than notched specimens". Only a V-notch was shown for the Charpy test. Although
the dimensions for both types of specimens were identical with those currently specified,
many tolerances were more restrictive. The units were shown as English preferred, metric
optional. The committee pointed out many details that influence the test results, but
because they did not have the knowledge and database needed to specify values and/or
tolerances for these details, the document was issued as a tentative. The original
document contains an appendix with general discussions of applications, the relation to
service conditions, and comparisons between materials. As our understanding of the
variables in Charpy testing has grown, ASTM E 23 has been revised repeatedly to
incorporate the new knowledge.
1934 to 1940
The first revision of E 23 was issued in 1934 and it added a dimension for the radii of
the anvil and specifically stated that "these specimens (both the Charpy and the Izod) are
not considered suitable for tests of cast iron" referencing a report of ASTM Committee
A3 on Cast Iron. The method retained the "tentative" designation.
The geometry of the Charpy striking tup, specifically the radius of the tup that
contacted the specimen, was not specified in the 1934 revision. However, the minutes of
the 1939 and 1940 meetings for the Impact Subcommittee of E1 state that this item was
discussed and a survey was made of the geometries used in the United Kingdom and in
France. Those countries had been using radii of 0.57 mm and 2 mm, respectively. For
reasons that were not recorded, the members of the Subcommittee agreed to a radius of 8
mm at the 1940 meeting and ASTM E 23 was revised and reissued as E 23-41T. Two
other changes that occurred with this revision were that metric units became the preferred
units, and keyhole and U notches were added for Charpy-test specimens.
1940 to 1948
Impact testing seems to have been a useful technique for evaluating materials, but was
not a common requirement in purchase specifications and construction standards until the
recognition of its ability to detect the ductile-to-brittle transition in steel. Probably the
greatest single impetus toward implementation of impact testing in fabrication standards
and material specifications came as a result of the large number of ship failures that
occurred during World War II. These problems were so severe that the Secretary of the
U.S. Navy convened a Board of Investigation to determine the causes and to make
recommendations to correct them. The final report of this Board stated that of 4694
welded-steel merchant ships studied from February 1942 to March 1946, 970 (over 20%)
suffered some fractures that required repairs [12]. The magnitudes of the fractures ranged
from minor fractures that could be repaired during the next stop in port, to 8 fractures that
were sufficiently severe to force abandonment of these ships at sea. Remedies included
changes to the design, changes in the fabrication procedures and retrofits, as well as
impact requirements on the materials of construction. The time pressures of the war effort
did not permit thorough documentation of the effect of these remedies in technical reports
at that time; however, assurance that these remedies were successful is documented by the
record of ship fractures that showed a consistent reduction in fracture events from over
130 per month in March 1944 to less than five per month in March 1946, even though the
total number of these ships in the fleet increased from 2600 to 4400 during this same
period [12].
After the war, the National Bureau of Standards released its report on an investigation
of fractured plates removed from some &the ships that exhibited these structural failures
and so provided the documentation of the importance of impact testing [13]. The NBS
study included chemical analysis, tensile tests, microscopic examination, Charpy impact
tests, and reduction in thickness at the actual ship fracture plane. A notable conclusion of
the report was that the plates in which the fracture arrested had consistently higher impact
energies and lower transition temperatures than those in which the fractures originated.
This was particularly important because there was no similar correlation with chemical
composition, static tensile properties (all steels met the ABS strength requirements), or
microstructure. In addition, the report established 15 ft-lb (often rounded to 20 J for
metric requirements) as a minimum toughness requirement, and recommended that "some
criterion of notch sensitivity should be included in the specification requirements for the
procurement of steels for use where structural notches, restraint, low temperatures, or
shock loading might be involved", leading to a much wider inclusion of Charpy
requirements in structural standards.
1948 to Present
By 1948, many users thought that the scatter in the test results between individual
machines could be reduced further, so additional work was started to more carefully
specify the test method and the primary test parameters. By 1964, when the ASTM E 23
standard was revised to require indirect verification testing, the primary variables
responsible for scatter in the test were well known. In a 1961 paper, Fahey [14]
summarized the most significant causes of erroneous impact values as follows: (1)
improper installation of the machine, (2) incorrect dimensions of the anvil supports and
striking edge, (3) excessive friction in moving parts, (4) looseness of mating parts, (5)
insufficient clearance between the ends of the test specimen and the side supports, (6)
poorly machined test specimens, and (7) improper cooling and testing techniques. While
the machine tolerances and test techniques in ASTM E 23 addressed these variables, it
was becoming apparent that the only sure method of determining the performance of a
Charpy impact machine was to test it with standardized specimens (verification
specimens).
Much of the work that showed that impact tests did not have inherently high scatter,
and could be used for acceptance testing, was done by Driscoll at the Watertown Arsenal
[15]. Driscoll's study set the limits of 1 ft-lb (1.4 J) and • 5%, shown in Figures 1 and 2.
The data superimposed on these limits in Figures 1 and 2 are the initial verification results
gathered by Driscoll for industrial impact machines to evaluate his choice of verification
limits. In Figure 1, the verification results for the first attempt on each machine are
shown: only one machine fell within the • 1 ft-lb (1.4 J) limit proposed for the lower
energy range. Results for retests on the same machines after maintenance are shown in
Figure 2. Driscoll's work showed the materials testing community that not all machines
in service could perform well enough to meet the indirect verification requirements, but
that most impact machines could meet the proposed requirements if the test was
conducted carefully and the machine was in good working condition. With the adoption
of verification testing, it could no longer be convincingly argued that the impact test had
too much inherent scatter to be used as an acceptance test.
Early results of verification testing showed that 44% of the machines tested for the first
time failed to meet the prescribed limits, and it was thought that as many as 50% of all the
machines in use might fail [16]. However, the early testing also showed that the failure
rate for impact machines would drop quickly as good machines were repaired, bad
machines were retired, and more attention was paid to testing procedures. It was
estimated that approximately 90% of the machines in use could meet the prescribed limits
o f • 1 ft-lb (1.4 J) or • 5%. Recently acquired verification specimen data, shown in
Figures 3 through 5, confirm these predictions. Failure rates for verification tests at low,
high, and super-high energy ranges are currently estimated to be 12, 7, and 10%,
respectively [17].
Overall, the incorporation of verification limits in ASTM E 23 has greatly improved the
performance of impact machines, so that data collected using ASTM E 23 machines can
be compared with confidence. ASTM E 23 is still the only standard in the world, to our
knowledge, that requires very-low-energy impact specimens (between 15 and 20 J) for
verification, and as shown by the data in Figure 1, results obtained using machines in need
40 I I I I I
30
~
o
r
20
$ 9
i:5 lO
e
I I I I I
-10
0 20 40 60 80 100 20
Energy, J
Figure 1 - The deviation and energy values obtainedfor the first round of tests on
industrial machines. The deviation is calculated as the difference between the results of
the Watertown Arsenal machines and the industrial machines. These data were
originally published by D.E. Driscoll, Reproducibility of Charpy Impact Test, AS134 STP
176, 1955.
of maintenance can vary by more than 100% at this energy level. In effect, the limits
imposed by ASTM E 23 have produced a population of impact machines that are arguably
the best impact machines for acceptance testing in the world.
While ASTM E 23 is used around the word, there are other forums for the
development of global standards. One of these, the International Organization for
Standardization, ISO, allows qualified representatives from all over the world to come
together as equal partners in the resolution of global standardization problems [18]. ISO
Committee TC 164 handles the topic of Mechanical Testing, and its Subcommittee SC 4
handles toughness testing. While this subcommittee has developed and maintains ten
standards on toughness testing, perhaps the most pertinent is 1S0 StandardR 442:1965
Metallic Materials -lmpact Testing - Verification of Pendulum Impact Machines. This
standard covers the Charpy test and is presently undergoing balloting for revision. An
important feature of this document is that it recognizes Charpy testing with both the 2-mm
and 8-mm radius striker. There are other regional and national standards that specify
impact testing procedures, such as the Japanese standard, JIS Z2242, Method for Impact
Testfor Metallic Materials.
10 I I I I I
8
6
"
9 |
9 .... : : ' ' _ , .... ,

0. 9 9
-4
-6
I I I I I
-8
0 20 40 60 80 100 120
Energy, J
Figure 2 - The deviation and energy values for the second and third rounds of tests on
industrial machines. The data shows that all but two of the machines testedwere able to
pass the 1.4 J or 5% criteria after appropriate repairs were made. These data were
originally published by D. E. Driscoll, Reproducibiligy of Charpy lmpact Test, ASTM
STP 176, 1955.
Typical Applications Today
Nuclear
Since it is impractical to measure the fracture toughness of large specimens throughout

the life of a nuclear power plant, surveillance programs use Charpy and tensile specimens
to track the embrittlement induced by neutrons. The economic importance of the Charpy
impact test in the nuclear industry can be estimated by noting that most utilities assess the
outage cost and loss of revenue for a nuclear plant to be in the range of $300,000 to
$500,000 per day. If Charpy data can be used to extend the life of a plant one year
beyond the initial design life, a plant owner could realize revenues as large as
$150,000,000. Further, the cost avoidance from a vessel related fracture is expected to be
in the billion-dollar range. To date, the NRC has shut down one U.S. plant as a result of
Charpy data trends. It is important to note that this plant's pressure vessel was
constructed from a one-of-a-kind steel and is not representative of the U.S. reactor fleet.
600 I'1'1'1'1'1'1'1'1'1'1'1'1'
500
400
I - 0.2
"13
B
"o
o
m .
o
g 300
O
-0.1
200 130
100
0 t ,~r-[ I~_L,~ I, I, I, I, '0 0

-5 -4 -3 -2 -1 0 2 3 4 5 6 7 8 9 1 0 "
Difference, J
Figure 3 - Distribution of low-energy verification data. Data for

1995-1997. Approximately 2400 tests; each test is an average of five
specimens. The vertical lines at ~:l.4 J represent the acceptance
criteria.
400
, , , , , ' -I0.16
10"14
300
o= 200 ] 0.08 =
O
[0o,
100 0.04 "~
0.02
0 ' 0.0
-20 -10 0 10 20
Difference, %
Figure 4 - Distribution of high energy verification data. Data for

1995-1997. Approximately 2400 tests. Each test is an average of five
specimens. The vertical lines at • represent the acceptance
criteria.
150 ' I ' I ' I i
- 0.2
"13
100 B
o
:Z
E
o
0.1 -a
50
0 '~J-~ ~ L ' 0.0

-20 -10 0 10 20
Difference, %
Figure 5 -Distribution of the super-high energy verification data.

Data for 1995-1997. Approximately 650 tests. Each test is an
average of five specimens. The vertical lines at • represent the
acceptance criteria.
Nonetheless, with decisions like this based on the Charpy test, the importance of ASTM E
23 and the restraints it applies cannot be overemphasized.
Steel
The Charpy V-notch (CVN) test specimen and associated test procedure is an effective
cost-saving tool for the steel industry. The specimen is relatively easy to prepare, many
specimens can be prepared at one time, various specimen orientations can be tested,
and relatively low-cost equipment is used to test the specimen. In many structural steel
applications, the CVN test can be used: (1) as a quality control tool to compare different
heats of the same type of steel, (2) to check conformance with impact requirements in
standards, and (3) to predict service performance of components. Also, CVN test
information can be correlated with fracture toughness data for a class of steels so that the
results of fracture-mechanics analyses can be compared with the material toughness.
CVN data have many uses, such as during the design and construction of a bridge or an
offshore oil platform. Before full-scale production of the steel order can begin, the
supplier needs to demonstrate to the buyer that the steel plate is capable of meeting certain
design criteria. The process begins by making the steel grade and then testing a portion of
the plate to determine if all required criteria are met. Also, steel mill equipment imposes
limitations on plate size; therefore, individual steel plates need to be welded together in the
field to produce lengths which can reach deep into ocean waters. Small sections of the
sample plate are welded together, and fracture mechanics tests are conducted to determine
the crack tip opening displacement (CTOD) toughness in the heat affected zone ffIAZ)
and in areas along the fusion line where the weld metal meets the base metal. Then, a
steel supplier might correlate the CTOD test results with CVN 50% ductile-brittle
transition temperature (DBTT). By agreement between the customer and supplier, this
correlation can allow the steel supplier to use the Charpy test instead of the more
expensive and time-consuming CTOD testing.
Continuing Standardization Efforts
Even after 100 years, the Charpy impact test procedures still have room for
improvement. The ASTM E 23 standard has recently been redrafted to provide better
organization and to include new methods such as in-situ heating and cooling of the test
specimens. Two new related standards are also under development through ASTM Task
Group E 28.07.08, "Miniature and Instrumented Notched Bar Testing", which was
formed a little more than two years ago. The first standard covers miniature notched bar
impact testing and relies on many of the existing practices related to test machine
requirements and verification as specified in existing standard E 23. The second standard
is focused on instrumented testing, where strain gages attached to the striker provide a
force-deflection curve of the fracture process for each specimen. Research is focused on
using these data to obtain plane strain fracture toughness as well as other key test
parameters. Upon acceptance of the standard by ASTM, both the existing E 23 standard
and the new miniature notched bar standards would reference the instrumented impact
standard.
The state of the art in impact testing continues to advance in other parts of the world
also. ISO is balloting a standard (14556) on instrumented impact testing, there is work in
Europe on miniature Charpy specimens, and ESIS is investigating the use ofpre-cracked
Charpy specimens for determining fracture toughness. It can be expected that
harmonization efforts will bring some of this work into E 23 in the future.
Conclusion
The ASTM E 23 standard is a document that continues to improve as our technical

knowledge increases. Several years ago, at the ASTM Symposium on "The Charpy
Impact Test: Factors and Variables" [19], a bystander was overheard to say: "I see that
there is a Symposium on the Charpy Test; what can be new there?" Since then, the
document has been updated twice and is currently being revised to reflect new
developments and to make it more "user friendly." Although ASTM E 23 has been a
useful standard for many years, it continues to be a "work in progress," a work used
extensively to help evaluate existing and new materials for products and structures -- a test
to ensure safety as well as to reduce the initial and lifetime costs for structures.
Knowledge which will help make the test more accurate and reliable is continually being
gained. New technologies such as miniaturization of the test, instrumenting the striker to
obtain additional data, and developing mechanics models to enable extraction of plane
strain fracture toughness will be areas of development over the next 100 years. We
anticipate that the benefits from the application of E 23 during the next 100 years will
overshadow the benefits from those in the past 100 years.
References
[1] Tredgold, T., Strength of Castlron, 1824, pp. 245-268.
[2] White, A.E. and Clark, C.L., Bibliography of Impact Testing, Department of
Engineering Research, University of Michigan, 1925.
[31 LeChatalier, A., "On the Fragility After Immersion in a Cold Fluid", French
Testing Commission, Volume 3, 1892.
[4] "Report on the Work of the Council from the Budapest to the Brussels Congress -
1901-1906," Proceedings of the International Association for TestingMaterials,
Brussels Congress, 1906.
[5] Hatt, W.K. and Marburg, E., "Bibliography on Impact Tests and Impact Testing
Machines," Proceedings ASTM, Vol. 2, 1902, p. 283.
[6] Russell, S. B. "Experiments with a New Machine for Testing Materials by Impact,"
Transactions ASCE, Vol. 39, June 1898, p. 237.
[7] Proceedings of the Sixth Congress of the International Association for Testing
Materials, New York, 1912.
[8] Charpy, M.G., "Note sur l'Essai des Metaux a la Flexion par Choc de Barreaux
Entailles, Soc. Ing. Francais, June 1901, p. 848.
[9] "Impact Testing of Notched Bars," The Engineer, Vol. 99, March 10, 1905 pp.
249-250.
[10] Whittemore, H. L., "Resume of Impact Testing of Materials, with Bibliography,"

ProceedingsASTlv[, Vol. 22, Part 2, 1922, p. 7.
[11] Warwick, C. L., "Resume on Notched Bar Tests of Metals," Proceedings of

ASTAI, Vol. 22, Part 2, 1922, p. 78.
[12] The Design and Methods of Construction of Welded Steel Merchant Vessels: Final
Report ofa (U.S. Navy) Board of Investigation, WeldingJournal, Vol. 26, No. 7,
July 1947, p. 569.
[13] Williams, M. L. and Ellinger, G. A., Investigation of Fractured SteelPlates

Removedfrom WeldedShips, National Bureau of Standards Report, December 9,
1948.
[14] Fahey, N. H., "Effects of Variables in Charpy Impact Testing," Materials Research
Standards, Vol. 1, No. 11, Nov., 1961.
[15] Driscoll, D. E., "Reproducibility of Charpy Impact Test," Impact Testing, ASTM
STP 176, 1955.
[16] Fahey, N. H., "The Charpy Impact Test - Its Accuracy and Factors Affecting Test
Results," Impact Testing of Metals, ASTM STP 466, ASTM, 1970.
[17] McCowan, C. N., Wang, C. M., and Vigliotti, D. P., "Summary of Charpy Impact
Verification Data: 1994 - 1996," Submitted to the Journal of Testing and
Evaluation, 1998.
[18] More information is on the ISO World Wide Web site, at http://www.iso.ch
[19] Charpy Impact Test: Factors and Variables, ASTM STP 1072, J. M. Holt, Ed.,
ASTM, 1990.
S. B e n t R u s s e l l I
Experiments with a New Machine for Testing

Materials by Impact (Reprint from 1898)*
REFERENCE: Russell, S. B., "Experiments with a New Machine for Testing Materials
by Impact (Reprint from 1898)," Pendulum Impact Testing: A Century of Progress, STP
1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American Society for Testing and Materials,
West Conshohocken, PA 2000.
When stress is applied to a solid body, the material is distorted and a certain amount of
work or energy is absorbed. The work thus absorbed in the deformation of the material is
called resilience. If the stress changes from zero up to the elastic limit of the material, the
energy absorbed during the change is the "elastic resilience" of the material. If the stress
changes from zero up to the ultimate strength of the body, the energy absorbed is the "ul-
timate resilience" of the body. 2
In the study of this subject it must be borne in mind that resilience is work, and hence
depends upon two essential factors, force and distance acted through. The latter is fully as
important as the former. The word toughness, as used by engineers, is synonymous with
resilience. In fact, the latter may be defined by saying that resilience is toughness reduced
to measurement.
Having defined resilience, it is next found that, as it depends upon change of stress,
different results may be looked for when the stress is applied suddenly, from those obtained
when it is applied slowly. The resilience under impact may not be the same as the resilience
under gradual load. In this connection impact should not be confused with sudden load. The
effect on resilience of rapidity of change in stress can only be determined by actual exper-
iment. This is especially true in the case of material not perfectly elastic, or where the stress
has passed the elastic limit of the material.
Again, the resilience of solids may be studied under the four principal kinds of stress,
viz., tension, compression, torsion and bending. The relative resilience under these different
forms of stress can only be determined by experiment. A knowledge of the resilience of
materials of construction is of the greatest importance to the engineer. It is the great resilience
of the battle ship's steel armor that enables it to withstand the impact of heavy projectiles
without destruction. It is the low resilience of cast iron that makes it so inferior for railway
bridges. It is on account of the high resilience of wood that it cannot, in many cases, be
supplanted by masonry, glass or other decay-proof material. A concrete railroad tie cannot
take the place of the oak tie because it lacks resilience.
* Reprinted with the permission of the American Society of Civil Engineers from Transactions, Amer-
ican Society of Civil Engineers, Vol. 39, No. 826, 1898, pp. 237-250.
1Member of the American Society of Civil Engineers.
2 This use of the word resilience will be objected to by some as not being in conformity with the
original meaning of the word. It is sanctioned, however, by some authorities (see Thurston's "Materials
of Engineering"), and, for want of a good substitute, may be considered as a technical tenn.
17
Admitting the importance of a knowledge of resilience, a brief consideration of the dif-

ficulties to be overcome in obtaining such knowledge is naturally next in order. It is at once
found that they are of considerable proportions. To find the strength of a beam under given
conditions it is only necessary to find its weakest section and study that. To find the resilience
of the beam all sections must be taken into account. If the beam is irregular in form, the
problem becomes quite a difficult one. If the final stress exceeds the elastic strength of the
material, the difficulties are increased.
The actual measurement of the resilience of a beam has been found quite difficult. The
load must be increased gradually and the deflection measured and recorded with its corre-
sponding load. As the breaking point is neared the difficulties of accurate work become
important, especially in the more ductile materials. If the determination of the resilience by
impact or drop test is attempted, other complications arise. The mass or weight of the beam
itself now becomes a factor in the test. The work absorbed by the anvil and hammer and
that are taken up in abrasion, etc., are difficult to estimate.
To one who has a proper understanding of these difficulties in measuring resilience, it is
not surprising that the subject is somewhat neglected in the studies of practical men. At
present it may be said that the knowledge of comparative resilience of materials is "appre-
ciable, but not describable." It is known that a cubic in. of oak has more resilience than a
cubic in. of pine, but the value of either cannot be expressed in inch-pounds or foot-pounds.
What is known about resilience, and the modern methods of determining its values, will be
briefly considered.
An interesting series of experiments on the resilience of beams under impact was made
by Mr. Hodgkinson. The following quotations from a book well known to engineers 3 will
show the more important results of these experiments:
"The power of a beam to resist impact is the same at whatever part of the length it is
struck; . . . . this remarkable result has been confirmed by experiment."
"In rectangular beams of unequal dimensions the resistance 4 is the same, whether the bar
is struck on the narrow or broad dimension."
"With rectangular beams the resistance to impact R is simply proportional to the weight
of the beam between supports, irrespective of the particular dimensions."
The above laws exclude the effect of inertia.
"Mr. Hodgkinson has shown by his experiments that in resisting impact, the power of a
heavy beam is to that of a light one as the inertia of the beam, plus the falling weight, is to
the falling weight alone, or as (I + W)/W."
" I is the inertia of the beam and the load upon it."
"The inertia of a beam, uniform in cross-section from end to end, supported at the ends
and struck in the center, may be taken at half the weight between supports . . . . To this has
to be added the whole central load, if any."
In the second column of Table 20 will be found some values for the resilience of certain
materials, which were obtained from the book above referred to.s In modern practice, the
testing of materials by impact is by no means uncommon. Such tests, however, axe generally
made on the finished shape, as in the case of railway axles. In a code for testing materials,
recommended by a committee to the American Society of Mechanical Engineers, 6 it was
3 "Strength of Materials" by Thos. Box.

4 Resilience?
s Interesting matter on the subject of impact, resilience, etc., will be found in Engineering News,
August 2, 1894. See also "A Photographic Impact Testing Machine" with discussion, Journal of the
Franklin Institute, November, 1897, and January, 1898.
6 See Engineering News, March 7, 1891.
RUSSELL ON TESTING MATERIALS BY IMPACT 19
prescribed that drop tests should be made with a steel ball, weighing 1 000 to 2 000 lbs.,
having a clear fall of 20 ft. The anvil, block, frame, etc., should weigh not less than ten
times as much as the ball. Drop tests were recommended for rails, fires and axles. Again,
the Master Car Builders' Committee, 7 have recommended drop tests for railway axles. These
tests were to be made with a tup, weighing 1 640 lbs. The anvil should weigh 17 500 lbs.,
and should rest on springs. The axle should rest on supports 3 ft. apart. Cast-steel drawbars
are now regularly furnished by contract; under specifications which call for drop tests of
sample drawbars, specifying weight of tup, height of drop and number of blows. Drop tests
of steel rails have been in practical use for many years.
Besides the above tests of finished shapes, the following methods, which are used in
commercial practice, may be noted. These tests, while they do not measure the resilience so
directly, are, nevertheless, intended to prove the toughness of the material.
In testing cast-iron water pipe by hydraulic pressure, it is customary to strike the pipe
smartly with a hand hammer while the pressure is on. In inspecting steel where a sample
bar is nicked and then bent with the hammer, the behavior of the bar indicates the degree
of toughness which the material will have under impact. A high percentage of phosphorus
in steel is believed to reduce its ability to withstand shocks, while its strength and percentage
of elongation remain unchanged. 8 So that it may be said that the specified chemical deter-
minations of phosphorus in structural steel, which are now in use, are really indirect tests
of resilience under impact.
Users of structural steel will readily see the necessity which now exists for a definite
physical test for the ultimate resilience of steel under impact. It was this special necessity
which led the author into the study of the subject, and suggested the experiments described
in this paper.
If, instead of limiting the percentage of phosphorus in the steel, a certain ultimate resilience
per cubic inch of the metal, when tested by impact, could be called for, a step would be
made in advance. If a definite resilience under impact could be specified, just as a definite
strength and ductility are now called for, the proper inspection of steel would be much more
simple and satisfactory.
The difficulties of making impact tests have already been suggested. Some machines which
have been used for making such tests are of a type greatly open to criticism. For example:
In some machines the supporting parts are either so light or so yielding that an important
part of the energy of the blow is absorbed by them, and the test piece appears to sustain a
much heavier blow than it would in fact on the proper rigid supports.
Two general forms of testing machine have been used in recorded tests. In Mr. Hodgkin-
son's experiments the hammer used was in the form of a pendulum striking with a horizontal
blow. The weight of the hammer was concentrated in the head or ball, and the effect of the
rod or radius arm was probably neglected. The most common form of impact testing machine
is doubtless the heavy weight falling vertically, somewhat after the fashion of the common
pile-driver. In none of these machines is there any means for measuring how much energy
is left in the hammer after breaking the piece.
The Impact Testing Machine

The machine used in making the experiments given herewith was devised by the author
and has some special features.
7 See Railroad Gazette, June 26, 1896.

8 See Johnson's "Materials of Construction," pages 166 and 167.
In designing it the main idea was to make a machine which would measure the energy
actually absorbed in breaking the test bar. This was to be done by using a hammer in the
form of a pendulum, and so arranged that it would strike a horizontal blow, breaking clear
through the bar and swinging freely up to the height due to the velocity after the impact.
The difference between the height through which the hammer fell before striking, and the
height to which it rose after striking, would measure the energy absorbed in breaking the
bar. The test piece would rest against two vertical knife-edges and be struck in the middle
by the falling pendulum, thus giving the ultimate resilience of the bar under transverse stress.
In developing this idea it was found best to make the pendulum or hammer of the very
simplest form, so that the center of percussion and center of gravity could be definitely
computed. 9 The hammer adopted was a rectangular steel bar pierced by a shaft at the upper
end and provided with a suitable striking edge near the lower end.
Figs. 1, 2 and 3 show the form and dimensions of the machine used in the experiments.
Plate XII is from a photograph which shows somewhat imperfectly the general appearance
of the apparatus. The hammer used weighed 103 lbs. The fixed knife-edges were designed
so as to allow the broken bar to swing out of the way of the moving hammer, and were
secured in a manner which allowed them to be adjusted for spans of 8, 12, 16, 20 and 24
ins. The heavy anvil plates behind them were bolted to a large anvil block of concrete which
was sunk in the earth. Adjustable supports were provided to hold the test bar in position
with the axis of the bar opposite the center of percussion. The pivot blocks which support
the hammer shaft are adjustable to allow for test bars of different depths. Attached to the
hammer shaft is a registering device on which the swing of the hammer is read. The pivot
blocks, etc., are supported by a strong wooden frame. Attachments are provided for raising
and releasing the hammer. The plans for this machine were made in May, 1896. In making
the design, the author was assisted by Mr. William E Schaefer and Mr. Vernon Baker.
Figs. 4, 5 and 6 show the plans and Fig. 12 the details of a later design which it is thought
embodies some improvements in detail, although the essential features are the same. In this
design the frame will be of iron and the operator will have more room in which to work
while setting the test bars in place.
In using the testing machine the first point that comes up is the loss due to friction of the
hammer in its bearings. In practice it was found best to determine the friction anew for each
set of experiments. If the bar was to be given a blow of 6 ins., the friction loss was deter-
mined for a fall of 6 ins. If the hammer rose 2 ins. after breaking the bar, the friction loss
for a fall of 2 ins. was determined by trial. The average of the two values was called the
correction for friction.
To test the rigidity of the knife-edges and their supports, a nickel 5-cent piece was placed
on edge on the top end of one of the knife-edges. A cast-iron test bar 2 ins. by 1 in. was
then broken by a single blow. This experiment was repeated a number of times, and, in the
majority of cases, the coin was not overturned by the shock.
An effort was then made to measure the movement of the knife-edge under a heavy blow.
The movement was found to be so small that in the case of a cast-iron test bar, the energy
absorbed by the yielding of the knife-edges would be quite inconsiderable. Every impact
testing machine should be tested in this way, to see if any considerable percentage of the
energy is absorbed by the yielding of parts that support the test piece.
In this method of testing materials some energy is absorbed in overcoming the inertia of
the bar itself. The proportionate amount of this energy is probably dependent on the weight
9 The formula for finding the center of percussion will be found in Rankine's Applied Mechanics,
Article 581.
22 P E N D U L U M I M P A C T TESTING: A C E N T U R Y OF P R O G R E S S
PLAT;: Xi|.
TRANS. AM. SOC. CIV. ENGRS.
VOL. XXXIX, No, 826.
RUSSELL. ON IMPACT TESTING EXPERIMENTS,
24 PENDULUM IMPACT TESTING: A, CENTURY OF PROGRESS
of the test bar compared with the weight of the hammer, and also upon the velocity of the
hammer.
Owing to the difficulties of ascertaining how much energy is absorbed in this way, it is
best to use a test-bar whose weight is small in comparison to that of the hammer. In this
way the error due to inertia of the test piece can be reduced, if not eliminated.
In Table 5 will be found the results of tests made to determine the effect of changing the
initial fall of the hammer. The results are somewhat contradictory, but, in a general way, it
may be said that the experiments indicate that a small change in the initial fall of the hammer
will not change the amount of energy absorbed, to any great degree. This conclusion may
be regarded as important, as upon it depends somewhat the interpretation of all the experi-
ments. Table 5 will be referred to again.
The machine having been described, it only remains to present the experiments themselves.
Over 700 specimens have been broken, up to the present writing. These tests are not all
recorded here; only those which were thought to be most instructive are given. In order to
learn the possibilities of the testing machine, the study of each material was continued only
until it was thought that the principal difficulties peculiar to such material had been over-
come. It is obvious that the resilience values obtained for different materials cannot be taken
as final, and should only be used by the designer in the absence of more accurate determi-
nations. All the experiments were made by the author, with the assistance of Mr. William E
Schaefer.
Tests of Brittle Materials

The first tests were made with cast iron. Table 1 shows the resilience of cast-iron bars
tested both by impact and by gradual load. Each value given is the average of several tests.
In making the impact tests, the following values are obtained by observations:
F = the initial fall of the hammer in inches.

S = the rise after the blow in inches.
C1 = the correction for friction.
L = the distance between supports.
h = the depth of beam.
TABLE 1--Resilience by impact and by gradual load. Cast-iron bars 1 in. by 2 ins., broken flatwise.
By Impact By Gradual Load
Length Resilence per Resilence per

Lot or Number between cubic inch, in Number cubic inch, in
Melt Experiment of Supports inch-pounds of inch-pounds
Nos. Nos. tests in inches L R1 tests Ro
1 125-130 6 24 11.5 3 9.0

2 137-139 3 24 10.8 3 8.7
3 156-159 4 24 11.4 3 8.5
4 219-222 4 24 11.8 3 8.8
6 391-393 3 12 17.9 2* 11.1
7 448-449 2 12 14.8 2* 8.2
Averages . . . . . . . . . 13.03 ... 9.05
* L = 24 ins. with gradual load.

b = the width of beam.

All dimensions are in inches.
Then, by computation, when 103 is the weight of the hammer in pounds, the resilience
in inch-pounds per cubic inch of the material, or
103 [ F - (S + C1)]
Rl = Lhb
Table 19 shows a series of observations just as they were recorded by the observer, and
extended in the office.
The resilience by gradual load was obtained by breaking the bar in a standard testing
machine and accurately measuring the deflections up to the point of rupture. The resilience
was then taken as half the product of the load by the maximum deflection. The true resilience,
as obtained by a strain diagram, would be slightly greater than this, but the error is not
important as the strain diagram for cast-iron is nearly straight to the point of rupture.
Returning to Table 1 and comparing the resilience by impact and by gradual load, it will
be seen that the former exceeds the latter more than 40%. This difference is so great that it
can hardly be accounted for by losses due to inertia of bar, indentation, or movement in
supports. The bar is light compared with the hammer, so that not more than 7% could be
lost by inertia according to Mr. Hodgkinson's rule. The supports are so rigid that not more
than 1% could be lost by their movement. The indentation is so slight as to be inconsiderable
when compared with the deflection of the bar, hence there can be no great loss in this way.
The logical conclusion is that more energy is absorbed in the sudden rupture of a bar than
is the case with rupture under a gradual increase of load.
It has occurred to the author, that perhaps the causes of this difference may be traced back
to the heat which is liberated under change of stress. Under gradual increase of stress the
heat liberated has time to be conducted away from the distorted fibers. In the case of sudden
rupture, the heat has no time to escape and must produce a rise in temperature. If this be
admitted, it seems not impossible that the resilience may be affected by the rise in temper-
ature of the distorted particles. This suggestion should be taken for what it may prove to be
worth.
Table 2 needs no explanation. Bars of the same melt, but of different spans, are compared.
A bar of 12-in. span has twice the strength and one-quarter the deflection of a b a r 24 ins.
in span. With the former, then, a greater loss of energy by movement of the knife-edges and
TABLE 2--Resilience for different spans. Cast-iron bars, 1 in. by 2 ins., broken flatwise.
24-in. span 12-in. span
Resilence per Resilence per

Number cubic inch, in Number cubic inch, in
Lot Experiment of inch-pounds Experiment of inch-pounds
Number Numbers Tests R1 Numbers Tests R1
125-130 6 11.5 131-136 6 11.7

137-140 3 11.0 146-155 10 11.0
156-159 4 11.4 170-174 5 12.7
296-299 4 9.9 300-303 4 10.2
by indentation might be expected. Theoretically, the error from these sources would be about
eight times as great for the shorter span. On the other hand, the error from inertia should be
about twice as great in the longer span as in the shorter one. It will be seen by the table
that the difference in the resilience per cubic inch ranges in value from nothing up to about
10%, and that the shorter span shows the higher average resilience. It is fair to conclude
from these experiments, as far as they go, that the ultimate resilience of a bar of cast-iron
is proportional to its volume and is independent of the span.
Table 3 shows that a flat bar has about the same resilience whether broken flatwise or
edgewise. All these bars were cast from the same melt. In the case of a bar 2 ins. wide and
1 in. thick, it should have, when broken edgewise, twice the strength and half the maximum
deflection that it would have flatwise. The error from yielding supports and from indentation
should be about four times as great in the former position. The error from inertia of bar
should be the same in both cases. It would be expected that the bars would show greater
resilience when broken edgewise. The observed resilience was, however, somewhat greater
in the average, with the bars broken flatwise.
As in testing bars in this manner, it is possible for the experimenter to raise the hammer
considerably higher than is necessary to break the bar, the question naturally comes up: Will
the height to which the hammer is raised affect the results obtained? A number of experi-
ments were made to decide this point, and the results are recorded in Table 5. The experi-
ments were made in this manner: Twelve to sixteen bars were taken from the same melt of
cast-iron. Four of these bars would be broken with the hammer falling 5 ins., which would
barely break them. The resilience would be measured. The next four bars would be tested
with the hammer falling 6 ins.; the next with a fall of 7 ins., etc. The results obtained will
be seen in the last column of the table. It is evident that more experiments would have to
be made to find the true relation between the height through which the hammer falls and
the energy absorbed in the rupture. It is f a i l however, to conclude in a general way, as has
been stated, that a slight increase in the height will not materially affect the results obtained.
There seems to be a tendency for the resilience to increase as the height is increased; but
this tendency is all but concealed by variations from other causes.
TABLE 3--Resilience of cast-iron bars. Cross-section, 1 in. by 2 ins. Span, 24 ins. Melt No. 2.
Resilience per
cubic inch in
Experiment Number of inch-pounds
Position Numbers Tests R1
Flatwise 137-139 3 10.8

Edgewise 140-143 4 9.1
Average . . . . . . 9.95
TABLE 4--Resilience of cast-iron bars. Effect of planing. Melt number 4.
Weight of Resilience per

Span in Depth of Width of Bar in cubic inch in
Experiment Number of inches Beam in Beam in pounds inch-pounds
Numbers Tests L inches h inches b W RI
Rough 215-226 12 24 1 2 18 11.6

Planed 253-263 11 12 0.91 1.93 5.7 21.1
NOTE--For effect o f span, see Table 2. All bars were rectangular.

TABLE 5--Resilience of cast-iron bars. Effect of increasing initial fall of hammer.*
Size of Bar Resilience

Number of Initial Fall of per cubic
Experiment Tests Span Depth Width Weight of Bar, Hammer, in inch, in inch-
Melt Numbers Made L h b in pounds W inches F pounds R 1
156-159 4 24 1 2 13 7.0 11.4

164-167 4 24 1 2 13 9.5 12.1
160-163 4 24 1 2 13 12.0 12.5
t--
170-174 5 12 1 2 6.5 4.0 12.7 6o
180-183 4 12 1 2 6.5 6.5 13.0 I-i1
177-179 3 12 1 2 6.5 9.0 16.8 i--
i"
175-176 2 12 1 2 6.5 12.0 15.2 O
z
I 219-222
223-226
4
4
24
24
1
1
2
2
13
13
6.0
7.5
11.8
11.6 m
215-218 4 24 1 2 13 9.0 11.5 -.q
z
I"253-255 4 12 0.9 1.9 5.7 5.0 21.2 fi)
249-252 4 12 0.9 1.9 5.7 6.0 19.1
4*
256-259 4 12 0.9 1.9 5.7 7.0 21.9
260-263 4 12 0.9 1.9 5.7 8.0 22.3 m
* Planed.
I'-
09
NOTE--All bars were rectangular.
-<
(3
IX3
Coming back to the regular order: Table 4 shows the effect of planing on the resilience
of a cast-iron bar. The results shown are somewhat remarkable. The bar, after planing off
the surface on all four sides, is much tougher than it was before. This difference cannot be
due to any fault in the method of testing, as may be seen from a comparison of this table
with Tables 2 and 3. The superiority of the planed bar is probably due to the lessening of
the shrinkage strains when the surface of the rough casting is removed. It is possible that
the same gain might be made by annealing the rough bar. The discovery of the great increase
in resilience after planing might have been prophesied, perhaps, from studies heretofore made
of the loss of strength due to shrinkage strains. This fact, however, has never before been
demonstrated by actual impact tests, to the author's knowledge. The great advantage of
finishing castings exposed to shocks should be taken into account by designers of machinery.
Table 6 gives the results of tests of paving brick. The first tests of brick, made with the
hammer, were unsuccessful on account of the great thickness of a brick compared with its
length. The broken brick would wedge between the hammer and the opposing knife-edge,
so that the hammer could not swing through. To remedy this, the author devised a knife-
edge which would be immovable when struck squarely, but which would move freely by a
side pressure. The form and dimensions of this device are shown in Fig. 13. As soon as the
brick is broken, the knife edges are thrown outward and the hammer swings freely through.
With the aid of these "free knife-edges" bricks were tested with good results.
Owing to the low resilience of a brick compared with its weight, it was found advisable
to raise the hammer no higher than was necessary to break the brick. A higher drop usually
showed a higher resilience. It is probable that the values given in Table 6 are higher than
would be obtained could the error due to inertia be entirely eliminated. It is hardly safe to
accept these results in comparing bricks, unless they be of the same dimensions.
Table 7 shows the results of a few tests of red brick. The comparative values obtained
from soft and hard bricks are as might be expected. The familiar test of striking two bricks
together in the hands is a crude impact test, and, in experienced hands, probably determines
the comparative toughness of the brick with some accuracy.
Table 17 gives a comparison of the values obtained with different materials, tested in the
manner described. They are classed as brittle materials because they can be tested in the
same way as cast iron, and do not require special treatment, as do wrought iron and steel.
The table gives a good rough idea of the comparative value of these materials under impact.
The values given in the last column are the mean of several tests in each case. They should
not be taken as typical, as the samples were taken from materials at hand and may not be
truly representative.
Tests of Tough Materials

Having now dealt more or less effectively with the brittle materials, a class that presents
greater difficulties must be considered. How, for example, shall the ultimate resilience of a
sample of wrought iron be determined? If an attempt is made to break a rectangular bar of
soft iron, it will only be bent. To break such a bar successfully, it must first be nicked. A
nicked bar can be broken, and the resilience to be overcome is but little more than that of
the metal lying close to the nick.
For want of some better method, the author adopted t h e plan of using a nicked bar for
testing soft iron and steel, and determining the ultimate resilience per square inch of cross-
section at the nick. If the nick is deep enough to cause the bar to break off short, and is
always of the same form, it would seem that the resilience should be in some degree pro-
portional to the area of the reduced section. If, furthermore, the reduced section be always
of the same depth, the resilience should be directly proportional to the area.
TABLE 6--Resilience of vitrified paving brick. All broken on a span of 7 ins.
Weight of Resilience per

brick, in cubic inch in
Number of Depth, in Width, in pounds Experiment Number of inch-pounds
Where Made Lot inches h inches, b W Nos. Tests Rl
c
60
1.43 60
Glen Carbon, Ill. 2 2.6 3.7 6.7 531-536 m
2.6 3.9 7.1 564-569 2.64 r-
Galesburg, Ill. 3 l--
Kansas City, Mo. 4 2.5 3.8 6.8 539-544 1.00 O
Galesburg, Ill. 5 2.6 4.0 6.9 545-550 1.54 z
Canton, O. 6 2.5 3.9 6.8 551-557 2.09 -t
i1"1
Alton, IlL 7 2.5 3.8 6.5 558-563 1.25
Glen Carbon, Ill. 8 3.0 3.8 8.1 570-575 2.19 T~
Athens, O. 9 3.3 4.0 9.5 576-581 3.26"
* This high value is probably due, in part, to the greater weight.

m
>
I'-
00
-<
bO
t.O
TABLE 7--Resilience of red brick. All broken on a span of 7 ins.
Dimensions
of brick Weight of Resilience per
in inches brick, in cubic inch, in
Number pounds Experiment Number inch-pounds
Kind of Brick of Lot l h b W Numbers of Tests g 1
Face brick 2 8.5 2.4 4.1 5.4 687-691 5 0.26

Common,
hard burned 3 8 2.2 3.9 5.1 692-695 4 0.30
Common, soft 5 8.5 2.2 4.2 5.2 708-713 6 0.10
Figures 7 to 11 show the different forms of nick that were used in the experiments. Each
form of nick is designated by a figure number, so that it may be referred to in the tables.
The order of these figures shows the results of the experience gained in these tests.
The first timber tests were made with bars like Fig. 7. The form shown in Fig. 8 was then
tried in order to reduce the chances of longitudinal splitting. In Fig. 9, the section is dimin-
ished by planin~ the sides. Fig. 10 was found to give better results with very tough wood
or metal. Fig. 11 is the same as Fig. 10, but with the section reduced as in Fig. 9. In the
last two forms, the hammer strikes the bar at the side of the smaller nick.
Table 8 shows the results of nicked tests made with cast-iron. The values given in the last
column show that the metal was all of equal toughness. The observed values, given in the
column next to the last, indicate that the resilience per square inch of section is not constant
for varying depths of section.
Table 9 shows the results of tests with different kinds of wood. The resilience values
shown by this table are probably somewhat high on account of loss by denting the wood.
l,. 1
Fig. 10
Fig. 11
~b
---..lg
i
: ~. L!
i It"
I-
Figure 13
TABLE 8--Nicked cast-iron bars. Resilience per square inch.
Resilience, in
inch-pounds
Resilience, in per cubic
Depth of Width of inch-pounds inch of
section section per square rectangular
at nick, at nick, inch of bars (rough)
in in section at from same
Number Experiment Fig. inches inches Number of nick melt
of Melt Numbers Number* h b tests made R2 R1
3 210-214 7 .5 1.0 5 49.5 11.4

3 205-209 7 1.0 1.0 5 83.8 11.4
4 272-275 9 .5 .9 4 81.6 11.8
4 268-271 9 .75 .9 4 91.4 11.8
4 264-267 9 1.0 .9 4 100.7 11.8
* Figure No. giving shape of nick (see Figs. 7 to 11).

N. B.--All bars 2 ins. x 1 in. All nicked bars broken edgewise, on 12-in. span. Weight of each bar
about 6.4 lbs.
TABLE 9--Nicked wooden bars; lot no. 3. Resilience per square inch. Experiments Nos. 313 to 375.
All bars shaped as shown by Fig. 8, with depth o f about 3/4 in. at nicked section. Depth 13/4 to 2 ins.
at ends. Width of bars, 7/8 to 13/4 ins., when not shown in second column. Span, 8 ins. These tests
were made without shims, to prevent denting.
Width of Resilience, in inch-pounds per

section Weight square inch of section at nick
at nick, of bar, Number R2
in inches in pounds of tests
Kind of Wood b W made Maximum Minimum Average
White pine . . . . . . 4 221 84 129

Ash . . . . . . 4 203 161 172
Cherry .9 .38 4 299 118 216
Poplar . . . . . . 4 255 168 222
Red cedar, No. 1 1.0 .49 2 235 215 225
" " No. 2 1.7 .85 2 86 85 85
Gum 1.0 .44 4 299 ... 247
Cypress, No. 1 1.7 .52 2 -.- 225 250
" No. 2 .9 .26 2 87 69 78
Chestnut 1.1 .44 6 432 199 306
Yellow pine . . . . . . 4 447 229 322
Black walnut .9 .44 4 420 235 328
Maple .9 .43 3 574 432 516
White oak . . . . . . 4 650 438 546
Oak, No. 2 . . . . . . 4 500 ... 419
Locust* 1.0 .56 6 690 566 633
Hickory* 1.1 .73 6 1 418 1 118 .,-
* The results in these remarkably tough woods are not strictly comparable with the others, on account
of tearing out of the extreme fiber.
TABLE lO--Nicked white oak bars. Effect o f shields or shims at knife-edges. All bars o f
straight-grained white oak o f same quality. Bars nicked as shown in Fig. 10. Depth of section at
nick (h) = 0.8 ins. Width o f section (b) = 1.7 ins. S&e of bar at ends = 1.75 ins. square. Weight
of bar = 0.88 lb. Span, 8 ins.
Resilience R 2
Experiment Number
Numbers of tests Maximum Minimum Average
With shims 450-455 6 430 30l 343

Without shims 456-461 6 505 278 410
TABLE 11--Nicked yellow pine bars. Resilience per square inch. All bars of same lot of straight
grain lumber. Bars nicked as shown in Fig. 10. Width of section (b) = 11/2 ins. Size o f bar at ends 2
• 11/2 ins. Weight o f bar = 0.75 lbs. Barprotected by steel shims at knife-edges: Span, 8 ins.
Resilience, in inch-pounds per square

Depth of section inch of section R 2
at nick Experiment Number
h Numbers of tests Maximum Minimum Average
.66 523-530 8 410 124 312

.88 496-503 8 525 211 447
Table 10 shows some tests made to learn how much loss of energy was occasioned by
denting. From these results it would appear that the loss in this way was considerable, and
that the wood should always be protected by shields or shims, at the knife-edges. The shims
used were thin strips of tempered steel about l/z-in, wide. They were laid flatwise between
the knife-edge and the specimen. All the later experiments were made with the specimens
protected from the knife-edges in this way.
Table 11 shows the results of tests made to determine the effect of increasing the depth
of the nicked section. The results indicate that the resilience of a nicked bar is not directly
proportional to the area of the nicked section. The variation is in the same direction as it
was in the cast-iron bars recorded in Table 8.
The nick shown by Fig. 10, which was used in the tests shown in Tables 10, 11 and 12,
was found to be the most satisfactory form for tests of wood. With this nick there is seldom
any longitudinal splitting, which would destroy the value of the test. Table 12 shows tests
of white and yellow pine and white oak, made with this form of nick. Shims were used in
these tests, so that they may be considered as made in a more approved manner than the
tests of Table 9. It is interesting to compare these timber tests with those made by Professor
Thurston (see Table 21).
Table 13 shows twelve experiments with bronze. Here, again, it will be noticed that the
resilience per square inch increases with the depth h, as it did in the case of cast-iron and
wood.
Table 14 shows a comparison of impact and gradual loading on nicked bars of plow-steel.
The gradual load tests were made in an ordinary transverse testing machine; the loads and
corresponding deflections were observed and plotted, and the resilience was taken from the
diagram. It will be noticed that the resilience by impact is about one-third greater than the
resilience by gradual load. The difference is nearly as great as was observed in rectangular
bars of cast iron (see Table 1). Table 14 shows also that the resilience per square inch does
not increase with a greater depth of section, as was observed in nicked bars of cast-iron,
wood and bronze. In the plow-steel tests it is found that the greatest depth gives the least
unit resilience, quite the opposite of what might have been expected.
Table 15 shows a number of experiments with steel and iron. In making these experiments
it was found that with very tough metal the bar should be nicked on both edges, to insure
a clean break and uniform results. Some of the lots of steel were tested for tensile strength
and elongation. The results of these tests are given in Table 15.
Table 16 gives the records of tests made with bars of aluminum. The metal was of the
kind used in making bicycle frames. An analysis showed 98.05% of aluminum. A tensile
test showed 16 750 to 19 970 lbs. per square inch ultimate strength, and 11/'2 to 31/'2%
elongation, in 8 ins. The specific gravity of the metal was 2.764. In these tests a greater unit
resilience with a greater depth of section may again be observed.
Table 18 gives a comparison of the tests made with different metals. The values cannot
be taken as typical. The numbers given in the last column for wrought iron and soft and
medium steel are thought to be fair values for an average grade of metal. In low-grade steels,
or steels low in carbon, it is a commonly accepted theory that a high percentage of phos-
phorus makes steel brittle under impact.* It may be from such a cause that some of the steel
tested gave such low results. It may be, on the other hand, more a question of the temperature
at which the metal was finished in the rolls.
Tables 20 and 21 were taken from well-known authorities, and are given for comparison
with the results of the other experiments. Both of these tables present values of resilience
by gradual loading.
* Johnson's "Materials of Construction," pp. 166 and 167.

TABLE 12--Nicked wooden bars; lot no. 4. Resilience per square inch. All bars 13/4 to 2 ins. deep at
ends. All bars nicked as shown in Fig. 10. Bars protected by shims. Span, 8 ins.
Resilience, in inch-pounds per

Depth of Width of Weight of square inch of section
section beam, in beam, in Number R2
Kind of at nick inches pounds of tests
Wood h b W made Maximum Minimum Average
White pine .88 .83 .35 7 223 90 161

Yellow pine .66 1.5 .75 8 410 124 312
White oak .80 1.7 .88 6 430 301 343
TABLE 13--Nicked bronze bars. Resilience per square inch. Bronze containing 85% of copper, All
bars from same melt (lot no. 2), and 2 ins. • 1/2 in. at ends. All broken edgewise on 12 ins. span.
Depth of Width of
section section Resilience in inch-pounds per
at nick, at nick, Weight of square inch of section at nick
in in bar, in Number R2
Experiment Figure inches inches pounds of tests
Numbers Number h b W made Maximum Minimum Average
394-399 7} 1.00 .50 4.22 3 1 305 1 192 t 252

.50 .50 4.13 3 1 302 884 1 087
473-478 9} 1.00 .38 4.31 3 1 205 1 087 1 147
.32 .38 4.00 3 769 581 673
NoTE--Ultimate strength, 27 730 lbs. per square inch. 6.9% elongation in 8 ins. Tensile resilience by
gradual load, 1 573 in.-lbs, per cubic inch.
TABLE 14--Nicked plow-steel bars. Resilience by impact and by gradual load. All bars nicked as
shown in Fig. 7, and broken edgewise on a span of 12 ins. All bars 2 ins. • 1/4 in.
Weight of one bar = 1.6 Ibs.
By Impact By Gradual Load
Resilience, Resilience,
Depth of section in inch-pounds in inch-pounds
at nick, per square inch per square inch
in inches Number of of section at nick Number of of section at nick
h tests made R2 tests made R2
.50 4 2 115 4 1 527

.75 4 1 625 3 1 133
.75 . . . . . . 1 1 460
.25 4 1 913 . . . . . .
NOTE--All from lot No. 1. Ultimate tensile strength, 83 720 lbs. per square inch, Elongation, 20.3% in
75/8 ins. Ultimate tensile resilience by stress diagram, 15 000 in.-lbs, per cubic inch, gradual load.
TABLE 15--Nicked iron and steel bars. Resilience per square inch. Bars nicked as shown by Fig. 10.
Depth o f section at nick, in inches, (h) = 0.25.
Depth and Resilience, in

Thickness Width of inch-pounds per m
z
o f Bar Section square inch Ultimate Tensile
Percentages c
Number before at Nick, o f Section, Strength, in
o f elongation c
Lot Number o f Tests Nicking, in inches at Nick pounds per
Metal Number o f Melt Made in inches b Rz square inch in 8 ins.
Iron, Norway 2 -" 4 13/4 • 1/4 .25 7 200 41 500 28.2

" Tenn. C o m 3 "- 6 11/2 x 1/4 .25 2 308 55 000 21.2 O
-H
" " charcoal 4 ..- 6 15/8 x 88 .25 3 560 52 500 27.5
~ --t
Soft steel 3 "" 4 11/2 x 3/4 .75 1 285 "" m
" " 6 743 3 15/8 x 1/4 .27 3 385 60 900 26.1 o9
-t
" " 6 743 2 15/8 x 5/16 .32 5 828 53 750 30.3
" " 6 743 3 15/8 x 3/8 .39 5 415 53 300 32.4 ~3
.i
" " 6 743 2 15/8 X 7/16 .45 3 886 54 800 29.6

>
" " 6 743 3 15/8 x 89 .50 4 919 52 250 33.1
" " 6 749 3 1% x 88 .27 3 448 63 100 27.0 O
m
" " 6 749 3 15/8 • 5/16 .32 4 709 61 250 23.9 z
" " 6 749 3 15/8 x 3/8 .39 4 366 56 100 31.6 .-I
c
" " 6 749 3 15/8 x 7/16 .45 4 065 57 000 31.5
" " 6 749 2 15/8 x 89 .50 1 821 58 600 25.8
" " 6 757 3 15/8 x 88 .27 4 836 60 600 25.4 0
77
" " 6 757 2 15/8 x 5/16 .32 6 036 56 900 30.5
" " 6 757 3 1% x 8/8 .39 4 997 56 000 26.2 "0
" " 6 757 2 15//8 x 7/16 .45 2 743 55 600 27.7
0
" " 6 757 3 1% x l/2 .50 2 981 57 600 27.5
" " 7 794 3 11/2 x x/z .50 1 523 61 800 25.1
17-I
" " 7 806 3 11/2 • 3/8 .39 4 416 58 900 29.0
" " 9 918 3 13/4 x 1/2 .52 3 810 52 720 32.4
M e d i u m steel 3 "" 3 2 X 3/4 .75 773 ""
" " 6 B 1 15/8 x I/2 .49 1 106 ""
.~
" " 6 C 1 15/8 x 5/8 .61 1 769 "-"
..~
" " 6 D 1 15/8 x 1/z .47 5 611 ""
.~
" " 6 G 1 15/8 x 1/2 .50 1 064 ...
...
" " 6 H 1 15/8 • 1/z .47 4 822 ""
Nickel " 1 "" 4 2 x 1/4 .25 2 600 72 680 27.5
Fluid comp. steel 1 "" 4 2 x 88 .25 3 300 91 260 18.9
Cast Steel 2 "" 2 2 X 8/8 .37 1 700 ""
.~
" " 2 .-. 2 2 x 88 .25 1 709 ...
TABLE 16 Nicked aluminum bars. Resilience per square inch. All bars 8 ins. long between
supports. Experiments Nos. 740-762.
Size of Section at Weight Resilience, in

Nick, in inches of Bar, inch-pounds per
Depth and Width in Number square inch of
of Bar at Ends, Fig. Depth Width pounds of Tests Section at Nick
in inches No. h b W Made R2
10 .25 1.25 1.44 4 468

188 x 188 10 .37 1.25 1.45 4 600
11 .37 .75 1.41 3 513
10 .25 1.25 2.06 4 530
2•188 10 ,50 1.25 2.13 4 579
11 .50 .75 2.07 3 519
NOTE--Ultimate tensile strength, 16 750 to 19 970 lbs. per square inch. Elongation, 1V2 to 31/2% in 8
ins. Ultimate tensile resilience by stress diagram, 15 000 in.-lbs, per cubic inch, gradual load.
TABLE 17--Resilience of brittle materials. Resilience, in inch-pounds per cubic inch. All tests made
with rectangular beams, struck in the center and broken by a single blow.
Resilience R 1 Resilience R
Material Max. Min. Av. Material Max. Min. Av.
Cast-iron rough 18 10 11.5 Common brick soft ...... .10

" " planed 22 19 21 Fire . . . . . . . . .44
Vitrified paving brick 3 1 1.6 Terra cotta, red ...... .33
Face brick, red ...... .26 "Granitoid"* .30 .15 .20
Common brick, hard ...... .30
* A composition of Portland cement and crushed granite, much used for sidewalks.
TABLE 18--Resilience of tough materials. Resilience, in inch-pounds per square inch of section at
nick. All tests made with rectangular beams, nicked at the center and broken by a single blow.
Fig. No. 10, h = 0.25 in.
Resilience R 2 Resilience R2
Material Max. Min. Av. Material Max. Min. Av.
Aluminum 530 468 500 Soft-steel 6 000 1 300 3 000

Wrought-iron 7 200 2 300 2 000 Cast-steel 1 770 1 709 ...
Medium-steel 5 600 770 2 000
fao
oo
-o
m
z
c
r-
e
TABLE 19--1mpact test. St. Louis, February 24, 1897. Office of Water-Works Extension. Specimen o f cast iron taken from Shickle & Harrison.
Specials--Pump Main No. 8. Tested f o r cross-breaking resilience, with results herewith appended. Lot No. 6.
Volume Effective Resilience

Rise of Bar, Correction Fall of per cubic m
r.-q
Initial after Correction Length in Weight for Inertia Hammer Total inch, in
Fall of Blow, for between Depth Width cubic of Bar of Bar, in H = F Resilience, inch-
O
Hammer, in Friction, Supports, of of inches in inches - (S + in inch- pound
Experiment Bar in inches inches in inches in inches Beam Beam V= pounds C2 = .011 Ca + pounds R1 103H
Number Nttmber F S C] L h b Lhb W W.S. C2) R = 103 H = T Remarks O
m
Z
582 1 6 3.24 0.17 12 1.05 2.07 26.082 7.75 .28 2.31 237.93 9.12
c
583 2 4 0.64 0.13 12 1.09 2.02 26.420 "" .05 3.18 327.54 12.40
584 3 4 0.85 0.13 12 1.05 2.05 2~830 "" .07 2.95 303.85 11.76
Small flaws O
-n
586 4 4 1.38 0.13 12 1.02 2.05 25.092 "" .12 2.37 244.11 9.73 on tension -0
and comp. side 30
O
587 5 4 0.81 0.13 12 1.04 2.00 24.960 "" .07 2.99 307.97 12.34
588 6 4 1.55 0.14 12 1.00 2.03 24.360 "" .13 2.18 224.54 9.22
m
Av. 10.76 o)
o)
TABLE 20--Resilience o f beams. R = Resilience, in inch-pounds o f a beam 1 in. square and 12 ins.
between supports.
Kind of Material Value Kind of Material Value
Cast-iron 81 Oak, English 78.4

Slate 3.2 " Canada 71.5
York paving 0.96 Pine, pitch 70.7
Ash 127.6 " red 58.7
Cedar 100.0 . . . . . .
NOTE The above values were taken from Table No. 67 of Box on "Strength of Materials."
TABLE 21---Relative torsional resilience.
Kind of Wood Value Kind of Wood Value
White pine 1.00 Yellow pine 3.87

Spruce 1.50 Black walnut 3.95
Red cedar 1.61 Locust 5.80
Spanish mahogany 1.65 Oak 6.60
Ash 2.25 Hickory 6.90
Chestnut 2.40 . . . . . .
NOwE--The above table was taken from Thurston's "Materials of Engineering." These tests were made
with gradual load.
Conclusions
The conclusions are: First, in the case of brittle materials, definite values for resilience
may be obtained.
Second, in the case of tough materials, like wrought iron, definite relative values for
resilience of materials of the same class may be obtained by the use of a test bar of standard
form and size.
This latter conclusion indicates that it may be specified that steel shall show a certain
ultimate resilience per square inch, with a given form of nicked test bar. Should this require-
ment prove satisfactory in practice, it may eventually be possible to dispense with chemical
tests of steel for structural purposes.
It may also be concluded from the tests that the resilience of cast-iron bars is greatly
increased by planing.
One more important deduction may be made from the tests, and that is, that metals show
a higher ultimate resilience under impact than they do under gradually applied loads.
When the proper values of resilience under impact have been determined for structural
materials, designers will be able to act with more intelligence in planning structures exposed
to live loads and to shocks. They will be able to substitute iron or stone for wood in certain
cases with greater assurance of safety. The study of resilience will also lead to better de-
signing in other ways. Useless material in a structure or member will generally decrease the
resilience, which fact is already known but frequently lost sight of. The general use of
resilience tests would serve to keep such facts in mind, and make them more commonly
understood.
It is with the idea of encouraging the practical use of impact tests that the results of these
experiments are offered to the Society.
DISCUSSION
L. L. Buck, M. Am., Soc. C. E . - - W h i l e the New York and Brooklyn bridge was being
built, the speaker saw some interesting experiments on resilience. A wire about 100 ft. in
length was suspended from the land span on the New York side, and on the lower end there
was a nut and a washer. Above the washer, and sliding on the wire, there was a weight of
about 50 lbs. The weight was first raised about 2 ft. and allowed to drop. It was then dropped
from a height of 4 ft., then 6 ft., etc., the distance being gradually increased for each drop.
The wire was broken by the weight falling from a height of 36 ft. The elongation of the
wire before it broke was remarkable.
A rod, about 94 ft. long and 1 in. in diameter, to the lower end of which was attached a
wire 6 ft. long, was used in another experiment. At the lower end of the wire there was a
nut and a washer, and the same weight was made to slip over the rod. The weight was
dropped from a height of 1 ft., then from a height of 2 ft., etc. The wire, although cut from
the same piece as in the first experiment, was broken by the falling of the weight from a
height of 6 ft. This showed the effect of the length of time taken in arresting the motion of
the falling weight. When the length of the wire was only 6 ft. there was very little elongation,
and it was broken apparently very easily. This can be shown by using a rubber strap to arrest
the motion of a weight which would on falling the same distance break a wire considerably
stronger than the strap.
In the experiments mentioned the elongations of the wire under the different falls were
not measured. The wire was of No. 7 steel, having a tensile strength of 160 000 lbs. per
square inch. The long wire broke some 6 or 8 ft. from the bottom; the short wire about 2
ft. from the bottom.
Joseph Mayer, M. Am Soc. C. E . - - T h e resilience of a material is the quantity of work
consumed per pound before breaking, if tested as a prismatic body.
Prismatic bodies of a given material consume a certain amount of work per pound which
is independent of the shape and size of the piece.
If nicked pieces are used, the same material will absorb, before breaking, a different
amount of work per pound, according to the size and shape of the nick and the size and
shape of the piece. The amount of work absorbed per pound, of a given material is not a
constant, but depends on various factors independent of the quality of the material. It is
therefore entirely improper to call the amount of work absorbed under such conditions its
resilience.
Unless the term resilience means the amount of work per pound consumed by prismatic
bodies before breaking, it has no definite meaning at all, and each experimenter will obtain
a different resilience for the same material according to the shape of the piece tested; and
confusion reigns supreme and brings discredit on all the tests and on the whole idea of
resilience.
J. B. Johnson, M. Am. Soc. C. E., has made a number of valuable tests,* in which the
resilience of cast iron was measured by gradually applied loads.
These tests do away with the inaccuracies unavoidable in tests by blows, because in the
latter an unknown amount of the work consumed is absorbed by the yielding of supports,
the inertia of the test piece, and the development of heat.
* Transacfion& Vol. xxii, p. 91.
40
These tests by gradually applied loads give results similar to those by blows, and they
could be applied to soft materials without nicking them, so as to obtain their real resilience.
Correspondence
James Christie, M. Am. Soc. C. E . - - A determination of resilience, especially if obtained
by impact tests, would supply desirable knowledge of the practical value of materials. When
the material is susceptible to fracture, and has no well-defined elastic limit, the ultimate
resilience alone can be depended upon; but in ductile materials, or those that do not fracture
readily, a determination of the elastic resilience would be more useful. The author might
obtain this in his impact machine by introducing another pendulum, bearing on the reverse
side of the specimen; the energy conveyed to this pendulum, through the deflection of the
specimen, being used as basis for computations of elastic resilience.
The results obtained from nicked specimens are of doubtful utility. Possibly in wood or
similar material, which is formed by an assemblage of strands or fibers, the resistance of the
nicked specimen may bear some constant ratio to that of a specimen with parallel sides, but
this is not true in the case of a material of a crystalline character. The well-known weakness
of a nicked bar of high-grade steel is confirmed by the results obtained by the author, which
place tool-steel below the hard woods in resilience. The theory of elasticity indicates that
when deflection occurs in such nicked specimens, the stresses at the interior angle are infinite,
or that destruction has begun at the lowest possible stress. Therefore, nicking the specimen
is only partially destroying the material, in order to insure rupture in material that otherwise
would not break by a bending process.
J. C. Meem, Assoc. M. Am. Soc. C. E . - - T h e r e are one or two points in this interesting
paper concerning which the writer begs to offer the following in the line of suggestion or
of asking for more information. In Tables 1 and 14 the author refers to the discrepancies
noted between the resilience of metals broken by impact and by gradual loading. It appears
to the writer that this discrepancy may be explained by the fact that the difference between
the initial fall and rise of the hammers does not seem to be correctly expressed in the formula
R 1 = 103 IF - (S + C1)]/L h b and that it should rather be R 1 = 103 (F2/a - {$2]al +
C } ) / L h b in which a and al are the arcs, respectively, through which the hammer falls and
rises, and F and S are the actual fall and rise, as in the first formula, i.e., the versed sines
of the arcs a and al; and, as it follows always that a~ is larger with respect to S than a with
respect to F, it will be found that the impact results will probably be reduced in any case
by not more than 89 or 1/2. The author's formula may perhaps embody this without expressing
it in detail, but if it does not, it may account, as stated, for the discrepancy.
The author proves that it is necessary to know only the weight of the hammer and its
relative rise and fall, and that an increase or diminution of this fall does not materially affect
the result. This is especially interesting as doing away with the necessity for further consid-
eration of the force and velocity of impact.
It is further borne out by the fact that a gradually applied load gives relatively the same
results as one by impact.
It is instructive also to note that the length being the same, the resilience is the same for
a piece broken along the breadth b as for one broken along the depth d; and that an increase
of length proportionally increases the ultimate resilience. Both of these conclusions are borne
out by an inspection of the theoretical formulas for resilience under a gradual load. As some
of the author's results are expressed in terms of resilience per cubic inch, and others per
square inch, the writer ventures to suggest that it might be well to express them always in
terms of resilience per square inch. For it would seem to convey a more practical idea to
say that the resilience of an inch bar was twice as great per square inch for a length of 24
ins. as for a length of 12 ins., than to say that the resilience per cubic inch was the same in
each case; i.e., it would seem better to express the variable (resilience per square inch) in
terms of the variable (length). This point may likewise have been considered by the author,
and discarded for good reasons in the valuable paper he has contributed to the opening of
an apparently new field for the investigation of a very interesting subject.
J. B. Johnson, M. Am. Soc. C. E.--This paper is a substantial contribution to the literature
on the strength of materials. The machine described is the first that has come to the notice
of the writer, which will indicate the true shock-resistance of any engineering material. The
theory of the machine is very simple and natural, and it is therefore the more remarkable
that it should not have found an earlier embodiment. There is no portion of the vast field of
testing the strength of materials which has been so grossly mismanaged and misunderstood
as this matter of testing resistance to shock. As the writer has repeatedly asserted elsewhere,
tests of shock resistance by repeated blows give no absolute data which can be used for
comparative purposes. Only when all the conditions of the test and test-specimen are exactly
duplicated can any comparison be instituted, and then only with the greatest caution and the
most intelligent discrimination. The author's results, however, are absolute in their character,
especially on brittle materials, and since it is only with such materials that engineers are
especially concerned regarding the resistance to shock, his apparatus is all that practice really
demands.
The paper shows the great necessity there is for a single word denoting resistance to
shock. The author, in the absence of such a word, follows some other writers by using the
term resilience, or total resilience, as indicating this property; but there is no doubt that this
is a misuse of the term. It has hitherto been restricted by careful writers, and by the highest
authorities, to that definition given by the Century Dictionary (quoting Thomson and Tait),
this being also the original meaning of the term as used by Young in 1807. The term
resilience should, therefore, be made to mean only the energy given back by the body in
returning to an unstressed condition. Within the elastic limit this is sensibly the same as the
energy absorbed by the body in deforming it. If the return path (on a stress diagram) were
identical with the deforming path, these two would be precisely equal, but as a matter of
fact this return path (on the load coordinate) is always a little below the deforming path, the
difference representing the small amount of heat generated and dissipated, even in elastic
deformation. Beyond the elastic limit the resilience proper becomes a very small part of the
deforming energy absorbed, so that two terms are needed, one indicating the energy put into
the body in deforming it and the other the energy given back by the body in the act of
recovery. The latter only should be called resilience. The former may be called resistance to
shock, or shock-resistance, but in the opinion of the writer it should not be called resilience.
In the absence of such a machine as the author here describes the writer has been accus-
tomed to determine shock resistance from the total area of the static stress diagram. That
this gave too small a result he also knew, since it has been shown that for equal deformations,
made statically and by impact, the resistance, or stress, is very much greater under a quick
action than under a slow one, and hence the impact stress diagram, if it could be obtained,
would be very much larger than the static diagram. Thus, for soft iron wire it has been shown
that the actual resistance to shock is some 30% greater than would be inferred from a static
stress-diagram.* For brittle materials it has been supposed there would be less difference.
The author shows, however (Table 1), that there is at least this difference in the case of cast
* Materials of Construction, p. 79.

iron. In other words, the time element does effect the ratio of stress to deformation with
brittle materials the same as with ductile materials. It follows, therefore, that a greater actual
shock-resistance may always be expected than would be computed from static stress-
diagrams, this excess being, perhaps, from one-third to one-half the computed shock-
resistance. This in itself is a very important discovery.
The superior resistance of the planed bars (Table 4) is doubtless due, mostly, to the removal
of the rough exterior rather than to the relieving of internal stress. Mr. W. J. Keep has just
shown* that a similar increase of strength under a static test results from the smoothing and
peneing action of the innumerable blows received in a rattler. He undertakes to show, and
apparently succeeds in showing, that this increase in strength is but slightly due to the shocks
received [as had been proved (?) by Outerbridge in 1895], and that it is almost wholly due
to the smoothing down of the rough exterior. It has long been known that test specimens
show an increase of strength due to such smoothing away of all irregularities, even though
these be very small, so that it is now common to require tests specimens to be finished with
a fine file, and, perhaps, polished, rather than to take them directly from the lathe or planer.
Very small indentations furnish favorable conditions for the starting of a crack, or permanent
deformation, when these would be considerably delayed without such starting points.
The necessity of rigid supports in shock tests has been well brought out in the paper, and
the author has evidently very successfully mastered this problem in his design.
It may still be doubted whether or not ductile or tough materials can be successfully tested
on such a machine. If the author would try the experiment of varying the sharpness of the
base of the notch at the center, as shown in Figs. 7 to 11, he would find that the slightest
change here makes a great difference in the result. These results also would seem to have
no absolute meaning, and comparisons could only be instituted between specimens which
were identical in size and shape in every particular. "Energy absorbed per square inch of
cross-section" is a meaningless phrase, since no energy can be absorbed on a true mathe-
matical plane or section. Some length dimension must be included to give a volume on
which the energy spends itself, but what this length dimension is, in the case of nicked bars,
cannot be determined. It would seem, therefore, that the only true field for such tests as here
described is with brittle materials. For these the author seems to be the first to show how an
absolute shock-resistance modulus, which is characteristic of the material and independent
of the form and dimensions of the specimen, can be obtained, and it is this which gives to
the paper a very great significance and value.
S. Bent Russell M. Am. Soc. C. E . - - T h e first point raised by Mr. Meem (a question of
formulas) does not seem to be well taken, as the following illustration will show:
Assume a pendulum weighing 100 lbs., swinging in a vacuum on frictionless supports so
that its center of gravity rises and falls through a vertical height of 1 ft. The energy contained
in the pendulum will be 100 ft.-lbs, at any instant, whether taken at its highest point, when
it has no velocity, or at its lowest point, when it has the greatest velocity, or at any inter-
mediate point.
Having determined the energy of the pendulum to be 100 ft.-lbs., it must be capable of
doing 100 ft.-lbs, of work. Now interpose a test-bar at any point in the path of the pendulum,
where it will be broken. After the bar is broken by the pendulum it is found that the latter
is still swinging back and forth, but that it now swings through a smaller arc. Suppose that,
on measuring, the center of gravity is now found to rise and fall through a vertical height
of 1/2 ft. The pendulum now contains but 50 ft.-lbs, of energy and hence the difference, or
* In a paper before the Am. Soc. Mech. Engrs., New York meeting, December, 1897.
50 ft.-lbs., has been absorbed in breaking the test bar, as is known from the law of conser-
vation of energy. It is not necessary to know the position of the test bar or the length of the
arcs, or even the length of the pendulum.
The vertical height through which the hammer is raised determines the energy in it as
soon as it is released. Hence, in the testing machine described (where the hammer weighs
103 lbs.) 103 F is the energy before striking, and 103 S is the energy after rupture. Add to
the latter a small quantity, or what is believed to have been lost in friction, and take 103
(S + C) as the true energy after the rupture of the bar. The difference between these values,
or 103 [F - (S + C)], must correctly express the value of the energy absorbed in rupturing
the specimen.
In reference to the point made by Professor Johnson on the resilience of planed cast-iron
bars, the author cannot agree with him in thinking that the experiments of Mr. Keep are
conclusive in the matter, and would still incline to the opinion that it is largely a question
of shrinkage strains.
There is another statement made by Professor Johnson to which the author cannot fully
subscribe, viz.: "It is only with such (brittle) materials that engineers are especially con-
cerned regarding the resistance to shock." Engineers are concerned with the resistance to
shock of all structural materials, and it may certainly be allowed that they are especially
concerned with the resistance to shock of the higher grades of structural steel. The greater
the strength of steel, the greater the danger of low resistance to shock, hence the more
complete the knowledge of the shock resistance of the metal, the higher the strength that
can be used with assured safety and the more economical the design. If impact tests have
been found a practical necessity in the case of rails, railway axles and cast-steel drawbars,
it can scarcely be denied that a knowledge of the resistance to shock of tough material is
eminently desirable.
Referring to Mr. Meem's last point, as to expressing resilience in terms of its value per
square inch, it may be said that there would seem to be no advantage in giving the resilience
of rectangular bars in terms of the area of the section for a stated length. In the case of
nicked bars the use of the term resilience per square inch is open to objection and is only
excusable where the width and depth are nearly uniform for all the sections to be compared,
and where the material is of the same character, so that the field of distortion may be
presumed to be the same in all sections. Even under these conditions the expression should
be regarded as a temporary expedient, so that the true conditions of each experiment are not
lost sight of. In using this expedient it is assumed that the unknown length dimension is the
same in all bars to be compared.
In discussing this point Professor Johnson remarks that resilience per square inch "would
seem to have no absolute meaning." Granting this, it may be in order to note that "percentage
of elongation" has likewise no absolute meaning. The length measured must be given, or
no definite knowledge is conveyed. Given the percentage of elongation in 8 ins. of a sample
of steel, and who can say what the percentage of elongation will be in a length of 2 ins. or
in a length of 8 ft.? And yet it is the common practice to give the elongation in percentage,
stating the length measured. In something the same way resilience per square inch means
nothing unless the form and dimensions of the bar are given, but it seems easier to make
comparisons if the results are reduced to a common area.
Mr. Christie remarks that "The results obtained from nicked specimens are of doubtful
utility," and in this position is sustained by Messrs. Mayer and Johnson. In considering this
point it is, perhaps, in order to note that in all physical tests of material there is more or
less difficulty in obtaining results that are "characteristic of the material and independent of
the form and dimensions of the specimen." For instance, in the case of cement, experimenters
have been working for years to obtain such results, and with but indifferent success. There-
fore, we need not be discouraged if only comparative values can be obtained in the case of
ultimate resilience of steel.
It is also in order to note that in actual practice structural material is not always used in
prismatic forms subject to uniform stress.
Wherever two members are joined together, there are changes of section more or less
abrupt. At every seam in a boiler shell a somewhat abrupt reduction of section is made by
the row of rivet holes. The screw thread on a bolt causes a nicked section. In timber con-
struction the pieces have sharp re-entrant angles where they are framed together.
Keeping these points in mind and admitting that the nicked tests are merely comparative
and are so only when the specimens are identical in size and shape in every particular, it
seems fair to conclude that tests of nicked specimens may yet prove of some practical value,
in the absence of a better method of determining the toughness of ductile materials under
shock. The value of nicked tests could best be determined with a suitable testing machine
located in a steel mill, where specimens could be obtained of any desired thickness and
composition and worked at any desired temperature.
Professor Johnson's objections to the use which has been made of the word resilience in
the paper are doubtless well taken. The term shock-resistance, however, does not appear to
be free from objection. For example, the title of Table 1 is "Resilience by Impact and by
Gradual Load." Shock-resistance by gradual load would be an inconsistent expression, hence
the term shock-resistance cannot be substituted here to convey the idea in mind. It would
be desirable, then, to find some expression for energy absorbed which would avoid the use
of the word shock.
Mr. G. Charpy*
Essay on the Metals Impact Bend Test of Notched Bars (Reprint from 1901)
Reference: Charpy, G., "Essay on the Metals Impact bend Test of Notched Bars
(Reprint from 1901)," Pendulum Impact Testing: A Century of Progress, ASTAqSTP
1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American Society for Testing and
Materials, West Conshohocken,
PA, 2000.
*This paper is printed with permission of CNIFS. Originally published in Soc. Ing. Civ.
de Frawais, June 1901, translated by E. Lucon, p. 848 - 877.
It has been maintained for a long time that the ultimate resistance of a metal to
failure, defined as the work needed to cause the rupture of a tensile sample, fully
classifies such metal from the point of view of its resistance to the most variable forces.
Numerous circumstances have shown that this opinion is erroneous, and particularly that
there seem to be no correlation between the resistance of a metal to dynamic stresses and
its resistance to static stresses. In his work on The Metals at the 1878 Exhibition, Mr.
Lebasteur has written to this regard:
"We conclude the following with Jern Kontor.

It is impossible to achieve complete results over the strength of materials without impact
tests.
On the other hand, this remark does not claim to establish a new jurisprudence in metal
testing.
It simply confirms an already ancient practice, which has not always been observed by
the consumers and from which they seem to be moving away, since tensile tests have
become more and more popular in the industrial practice. "
Impact tests are usually performed on small bars with square cross section, resting
on two supports and receiving an impact in the middle from a striker; sometimes the
small bar is clamped at one extremity and receives an impact from a striker on the free
end.
Such tests are presently only required in contract specifications in order to verify
the absence of fragility in metals. However, in the last few years there has been a
tendency to modify this procedure by favouring the rupture of the small bar by means of
a machined notch.
46
CHARPY ON METALS IMPACT BEND TEST 47
This procedure was allegedly applied for the first time by Mr. Andr6 Le Chatelier
in some experiments that he submitted in 1892 to the French Commission for test
methods and that were not the object of any specific publication; a summary may
however be found in an essay by Mr. Consid6re titled: Onfragili(y after coldworking a n d
the cleavability, which was included in the works of the French Commission on Test
Methods (volume III).
In this essay, Mr. Consid6re shows that certain steels, which provided normal
results in the tensile and bend tests, seemed conversely very brittle if the bend test was
performed on a thin bar previously cold notched on one side or locally work-hardened by
punching a hole through it.
Mr. Consid6re reports tests performed by Mr. Andr6 Le Chatelier, impact tests on
small square bars in which a notch, 1 mm wide and 1 mm deep, had been saw cut, and
similar tests carded out by Mr. Barba, impact tests on small bars in which triangular
notches of variable depth had been machined by the use of a planing tool.
Mr. Consid6re concludes in the following way:
"A remark can be drawn, with no possible objection, from the various experiences which
have been recalled. The tensile elongation and the bending after quenching, normally
regarded as providing a measure of ductility, do not give any indication on the resistance
to rupture which metals exhibit when their fibres, subject to an elongation, have been
altered by cold working or interrupted by a slight notch. It can therefore be contended
that conventional tests are absolutely insufficient to provide information about two of the
greatest dangers that threaten metals, at least in welded structures. "
This essay, with all the firmness of its conclusions, has not produced modifications
in the testing practice. The only endeavour to introduce a new acceptance method based
on the use of notched bars, of which we have knowledge, has been made by the
Establishment of the Navy of Indret, based on a report by the Engineer of Naval
Constructions, Auscher. Mr. Auscher, taking into account the remarks of Mr. Andr6 Le
Chatelier and Mr. Barba, has adopted the following test procedure:
The bars used for fragility tests have square cross section with 20 mm of side
length; they are clamped on one end and notched on all four faces, in the clamped
section, by means of a special tool which produces an equilateral triangular notch of 1
mm side length.
The bars must endure, without breaking, the impact of an 18 kg striker falling from
a height of 3 m on the extremity which is 100 mm away from the clamped section.
Such conditions have been applied since 1894 to a number of contracts for ship
masts.
Even if the tests on notched bars have not been adopted in testing practice, they
have nevertheless constituted the object of several important memoirs.
Mr. Barba submitted in 1893 to the Commission of Testing Methods a note about
the fragility of metals in which he described the following test method:
The test bars (sampled from plates) have a length of 300 mm and a width of 30 ram.
In these bars, triangular notches with 45 ~ angles are machined every 25 ram. The depth of
such notches can vary by a few tenths of a millimetre without consequences, but "the
sharpness of the angle which determines the length of the extreme fibre subject to tension
needs to be strictly controlled. This notch root has a rounded shape whose radius does not
exceed two tenths of a millimetre. In order to give the notches a uniformity that appears
essential, after rough shaping they are finished using a finishing tool carefully sharpened
by a scrupulous workman. "
Mr. Barba used notch depths which varied according to the thickness of the plate.
For the impact, he used an 18 kg striker falling from a variable height on the free end of
the bar, which was clamped with the clamped section corresponding to the first notch.
The protrusion, that is the distance from the striker axis and the plan through the notch
root, is equal to 22 mm. Fracture of the different notched sections of the bars can be
eventually produced. In order to characterise exactly the degree of fragility of the bar,
Mr. Barba wanted to find the exact falling height that causes rupture, that is in effect to
establish two falling heights which differ by a certain amount, for instance 10 cm, such
that the greater causes fracture of the bar, whilst the smaller is unable to separate the two
parts.
By submitting the different notched sections of the bar to impacts by the striker
falling from different heights, two limits of the fragility degree of the bar can be
estimated, that is: the maximum height which the bar could endure and the minimum
height below which rupture has taken place. The average of these two limits can be used
to characterise the metal under examination from the point of view of its fragility.
The procedure that we have just described seems to bear the disadvantage of having
to perform a large number of tests in order to obtain a unique result; moreover, it appears
quite difficult to achieve identical notches having a radius of two tenths of a millimetre,
something that Mr. Barba proclaims to be necessary.
The novelty of this method consists in the procedure which allows to obtain a
numerical value for characterising the fragility of a metal.
This numerical evaluation of fragility can be obtained in a more accurate way by
measuring the kinetic energy that the striker retains after producing the rupture of the bar,
thus calculating the work absorbed by the rupture. Two different solutions for this
problem have been accomplished by Mr. Russell and Mr. Fremont.
In an essay presented in 1897 to the American Society of Civil Engineers, Mr.
Russell described "some experiences with a new machine for impact testing materials".
The machine he employs has the aim of measuring the work absorbed by the rupture of a
bar; for this, the striker used has the form of a pendulum such that, when released from its
equilibrium position, it meets the test bar while passing through the vertical position,
breaks it and freely re-ascends under the effect of the velocity acquired. The difference
between the initial and final heights of the pendulum allows evaluation the-work
absorbed by the rupture of the bar.
The pendulum used is constituted by a simple metal bar of 1 m length, with
rectangular cross section, oscillating around an axle crossing it at one extremity. The
falling height could hardly exceed 0.25 m.
The height to which the pendulum rises after breaking the bar is indicated on a dial
by a needle mounted on a reduced friction pivot pin, which the pendulum trails in its
swing.
Using this device, Mr. Russel has studied many different materials, such as wood,
bricks, stones and metals. For those materials, he generally used bars with constant cross
section; only for ductile metals, he reckoned the necessity of characterising the rupture
section by notches in order to ensure the rupture. He employed the different notch shapes
shown below without recommending any one in particular (fig. 1).
Fi .t.
Mr. Fremont, in his interesting researches on material testing methods, has studied
a procedure for measuring fragility and evaluating the kinetic energy which a pendulum
retains after breaking a notched specimen. For this purpose, after the striker has caused
the rupture of the sample, it hits two springs which it compresses by an amount which is
proportional to its residual kinetic energy. An index allows reading the amount of
compression undergone by the springs. First of all, Mr. Fremont evaluates the
compressions obtained on the springs after freely swinging the pendulum from different
heights; he therefore obtains a diagram which he uses when performing tests to calculate
the residual energy which corresponds to a certain compression of the springs.
Mr. Fremont's device is described in the Bulletin of the Society of Engineers,
December 1898.
The standard specimen is 10 mm wide, 8 mm thick and 25 mm long; it rests on two
supports spaced by 20 mm and it is notched on its lower face by a saw cut 1 mm deep and
1 mm wide. The striker weighs 10 kg and falls from a height of 4 m.
In a note which recently appeared in the Bulletin of the Sociegy of Civil Engineers
(1), Mr. Barba has described some experiences performed in collaboration with Mr. Le
Blant by means of a pendulum that, similar to Mr. Fremont's, after breaking the bar goes
to strike two springs, but in which the residual kinetic energy is measured not from the
compression of the springs, but from the rebound height of the striker. As a possible
source of error, there could be concern for the friction of the striker against its guides,
which could assume a significant value particularly in the rebound of the striker, which in
such case is not forcedly moving parallel to the slides.
(1) Note about someexperiencesof impactbending on notched bars, April 1901.

Mr. Barba and Mr. Le Blant use samples 30 mm wide and 12 mm thick, which
exhibit on two opposite sides triangular notches with sharp bottom and 1 or 2 mm depth.
Such notches are initiated by a planing tool and finished by a tempered steel knife, with
sharp edge, which is placed on the bottom of the notch and is moved down half a
millimetre by means of a press.
The bars are clamped in such a way that the notched section coincides with the
clamped section; they receive on their extremity, 35 mm away from the notch, an impact
from a 25 kg striker whose falling height, in the tests described, was always lower than a
metre.
Finally, as a conclusion of this review of works related to tests on notched bars, we
must mention two memoirs recently published in the Bulletin o f the Encouragement
Society, which deal with the theoretical issues concerning the test method rather than with
the practical aspects of the test itself In the first memoir, Mr. Fremont attempts to relate
the results of the fragility tests to the relative values of the tensile and compressive elastic
limits; in the second, Mr. Fremont and Mr. Osmond study with ingenious methods the
distribution of deformations in the notched or unnotched bars subject to bending tests.
II
From the preceding exposition, it follows that all the researchers who have studied
impact tests on notched bars have applied very diversified procedures. If, in spite of all
those experiences, the tests on notched bars have not been widely adopted, this is
undoubtedly due to the drawbacks of any testing method whose peculiarities cannot be
defined in a rigorous way. This is what happens, with varying degrees, to all the
procedures we have described. Therefore, we consider it will be interesting to describe a
procedure that we have employed for several years at the Saint-Jacques Ironworks of
Montlur (ChStillon Company, Commentry and Neuves-Maisons).
The most delicate aspect, whose importance has been acknowledged by all
experimenters, is the preparation of the notch. It has been ascertained that shape and
dimensions of the notch have a remarkable influence on the test results. It is therefore
extremely important to standardise in a rigorous way shape and dimensions of the notch.
The triangular-shaped notch should not be machined using a planing tool, since the
edge can easily become blunt; the procedure which seems to yield the highest precision is
that proposed by Mr. Barba and Mr. Le Blant, which consists in finishing the notch by
pressing a sharp knife against its bottom. Experience shows that the consequent strain
hardening has no significant influence on test results; however, it would be extremely
difficult to proceed like this in practice, since, although the same knife could produce
comparable results over a reasonable period of time, it is almost impossible to prepare
two knives having exactly the same edge shape; moreover, the results obtained differ
remarkably according to the edge of the knife being more or less blunt. The same
inconvenience, even more serious, is encountered with the graver recommended by Mr.
Auscher. The fragility test is much easier to perform when the graver is slightly blunted
by usage than when it has just been re-sharpened; there ensues a source of disputes
between supplier and customer, which makes this testing method almost inapplicable.
The use of a saw cut gives slightly more regular results, but still remarkably
different according to the saw being new or used for a period of time. In the former case,
the notch has a rectangular shape with two sharp angles; in the latter, it is clearly blunted,
which obviously represents a more favourable condition for the test.
Considering all these difficulties, we have tried to find a notching method that
should not imply all the drawbacks previously mentioned. The method we have adopted,
after several attempts, consists in giving the notch root a well-defined radius which is
sufficiently large for the inevitable machining tolerances to have a negligible influence.
We subsequently proceed as follows:
By means of a helical drill, we drill in the bar a cylindrical hole and finish it by
using a reamer; the notch is obtained by joining the hole to one of the bar faces with a
saw cut.
A notch with the desired depth is thus obtained, its bottom having a p e r f e c t
cylindrical shape; it can be easily understood that such a procedure does not cause any
complaint and therefore allows to get rid of one of the difficulties caused by the use of
notched bars.
It seems possible to obtain a blunt-bottomed notch by using a planing tool; in most
cases, this procedure will give the same results as the one we have just described, but, if
the tool is slightly chipped, a scratch will be produced at the notch root, which will
behave as a sharp notch. We thus believe that it is much safer to use a tool working
perpendicularly to the notch direction, so that, if scratches are accidentally produced,
these will have no real influence on test results. Moreover, the cylindrical shape is the
easiest to define and to achieve in a perfect way by drilling.
The remaining test conditions are less rigorously defined. Following are the ones
that we adopted, together with the reasons which have motivated the choices.
We use bars with square cross section of 30 • 30 ram, and latterly bars with 20 x 20
ram, that is the most popular for impact tests, with blunt notch and affecting half the
thickness of the bar. A section of 15 x 30 tara is thus left for the bigger bars and 10 x 20
m m for the smaller. For the plates, we extract strips both 30 m m and 20 m m wide,
according to the thickness, and we machine the same notches as for the bars,
perpendicular to the rolling faces. The fracture section thus involves the whole plate
thickness.
The dimensions that we have adopted are relatively large, for two reasons. First of
all, if one doesn't want to sacrifice the accuracy o f results, it is evident that the precision
of testing and of specimen preparation should be inversely proportional to the dimensions
of the samples. I f very small bars are used, precautions which are more common to a
laboratory than to a workshop have to be adopted. Secondly, local defects assume a more
pronounced influence due to the small sample section, and it is totally useless, in
acceptance tests, to exaggerate the importance of local defects which will not be
accounted for in the applications of the metal and which, indeed, have a certainly
negligible influence on the resistance of the pieces actually used.
The bar can be clamped or just resting on two supports. We prefer this last solution,
which is simpler and safer, since clamping the bar more or less strongly could have an
influence on the results. Moreover, the clamped bar can be bent only to 60 ~, whereas the
bar resting on two supports can easily be bent as much as 90~ with the notch that we use,
it happens frequently for mild steels that rupture takes place with angles between 66 ~ and
90 ~
We have chosen 12 cm as the distance between the supports; the striker used for the
tests is the 18 kg striker prescribed for navy tests; the tup has been modified in order to
allow 60 ~ bending. The bar is subjected to a series of impacts from a constant height and
we count the number of impacts needed to provoke rupture, as well as the angle at which
rupture takes place. These two data allow a very clear ranking of the different metals. If,
in a very exceptional case, the bar is not broken when the angle reaches 90 ~ bending is
achieved by a press or a hammer and the breaking angle is recordedl
Basically, we come back to most of the test conditions adopted by Mr. Andr6 Le
Chatelier in his early tests, conditions which we considered superior from a practical
point of view to most of the ones which have been proposed later. The only modifications
concern the machining of the notch, for which a cylindrical shape allows a precise and
safe definition and whose influence is exaggerated by a increase of depth which causes
rupture for nearly any steel, even the mildest.
If the very simple test procedure that we have just described seems preferable for
routine acceptance tests, it is certainly true that there is great interest, in order to improve
our knowledge of metal properties, in the numerical determination of the work absorbed
by the rupture. For such a measurement, the pendulum procedure seemed to us clearly
superior to the use of springs, whose operation under the influence of an impact seemed
to be extremely complicated and hardly amenable to calculation.
/"
/
/"
/"
/"
,,, J
sI
1-
I"*" ,
e
i
,.s S
o
Striker arrangement in case of vertical bar. Scale 1/20.

' i
i ........
~. Cour/fw
........ j
We have built a pendulum for impact tests which operates in a satisfactory way.
This machine is represented in Figures 2, 3 and 4.
The striker, constituted by a plate weighing 50 kg, is designed so that the knife
which causes the impact coincides with the centre of gravity and is suspended, by a
triangle formed by very light unwelded tubes, to a metal frame; the pendulum swings
around an axle mounted on pins, whose friction is extremely low, as can be easily
verified with a free swing. The triangular shape of the suspension shaft guides very
exactly the striker, which oscillates always on the same vertical plane. The distance
from the axle to the striking edge, and therefore to the centre of gravity, is 4 m.
pi 9
Striker arrangement in case of horizontal bar. Scale 1/20.
The bar to be tested is placed on two supports against a metal plate which
constitutes the anvil-bed, weighing 1600 kg, driven into the ground and cemented in a
masonry block of 5 m 3. Two arcs of circle are placed on both sides of the frame, one
for clamping the striker at different heights by means of a clasp operated by a rope;
the second arc of circle supports a board on which a paint-brush, fixed to the
pendulum, marks the height to which the pendulum rises after producing the rupture
of the bar; this height is however easy to measure directly due to the slowness of the
oscillation. The difference between the initial and final height of the striker,
multiplied by its weight, yields the work absorbed by the rupture.
The striker can be attached to the pendulum so that the edge of the tup is
horizontal; this enables it to break bars clamped in a special support which is bolted to
the base-plate.
The loss due to passive resistances can be easily evaluated by performing a free
swing and following the reduction in the oscillations of the freely swinging pendulum.
We can thus calculate the loss due to passive resistances for any value of
amplitude and compile a table of corrections for the height observed in the rise of the
striker. Here are a few values for such correction:
Amplitude Height Correction

in degrees, observed, for the height.
30 ~ 0,53 m 0,008 m
42 ~ 1,02 m 0,017 m
52 ~ 1,54 m 0,028 m
60 ~ 2,00 m 0,030 m
68 ~ 2,50 m 0,040 m
76 ~ 3,00 m 0,070 m
The influence of friction is therefore always feeble; the work loss is, generally,
lower than 2 0/0; since this loss is known rather exactly, the error after the correction
becomes quite negligible.
Apart from this cause of error, the only one that could interfere with the
measurement of the work needed for the rupture of the bar is the kinetic energy
imparted to the fragments of the bar; this kinetic energy is always very small and
simple to calculate, in an approximate way, assuming that the fragments of the bar
have a speed equal to the residual speed of the striker, something which is very close
to reality. Since the weight of the bar is at most the fiftieth part of the weight of the
striker, it follows that the fraction of kinetic energy absorbed is very small. After
correction, the error from this side is even smaller and can be considered certainly
lower than 1 0/0 of the total kinetic energy.
In summary, the measurement of the work needed for breaking a notched bar by
means of the machine that we have just described can be achieved with an error
smaller than 1 0/0, that is with the same accuracy provided by most methods for
mechanical testing of metals.
III
It is not easy to appreciate the relative value of two procedures for testing
metals. The mechanical tests on metals provide data that are so complex that it is
impossible to infer numerical coefficients from them and to give such coefficients an
absolute value; when a tensile test indicates a resistance of 50 kg for a metal, we
cannot attribute any absolute significance to this figure, which depends on test
conditions; the only conclusion we can draw is that the metal would resist better to a
tensile stress with respect to another metal that, under the same test conditions
(specimen dimensions, tensile speed etc.), would provide a resistance value of 45 kg.
It is therefore necessary to consider mechanical tests only as a way to identify

the different metals and to classify them when respect to one another, without
claiming to learn their elementary properties.
Based on this, in order to compare two testing procedures, we consider that the
only way to proceed is to apply such procedures to several metals and to evaluate how
these metals rank with respect to one another in the two cases.
From the considerations previously exposed, we can also remark that the only
tests to be examined are those reproducing more or less closely the conditions found
in the building practice. From this point of view, the tests on notched bars hold an
undoubted interest since in any machine or construction there parts containing
weakened sections can be found.
In order to compare the test on notched bars with conventional tensile and
impact tests on unnotched bars, we have taken 25 rods from 25 different steel castings
of standard quality for rolled sections. From each of the rods the following were
extracted: 1~ a tensile bar with diameter 13.8 mm and 100 mm between reference
marks; 2 ~ one 30 x 30 mm square bar for impact test; 3 ~ three 30 x 30 mm square bars
with a notch with cylindrical bottom of 4 mm radius and 15 mm depth and that have
been tested, the first one with a 18 kg striker falling from 2 m (that is, 36 kgm each
blow), the second with a 12 kg striker falling from 2,75 m (that is, 33 kgm each blow);
finally, the third one with a 50 kg striker falling from 3 m. Here are the results
obtained from these different tests.
1o TENSILETESTS.
The bars are ranked in order of increasing resistance: such values range from
34,8 kg to 50,1 kg. Values of elongation and reduction of area are variable but always
show rather high values, all fractures are normal.
ProgressiveNo. Elasticlimit. Resistance. ElongafiorL Reductionof area.
1 22,1 34,8 33 68
2 23,1 35,2 34 67,6
3 22,1 35,4 35 67,3
4 23,4 36,1 32 68
5 22,7 36,1 36,5 70,5
6 22,7 37,4' 33 67,2
7 23,2 37,7 29 53,7
8 23,4 38,1 32,5 66,4
9 24,7 38,1 34,5 69,7
10 24,1 39,4 29 60,3
11 24,7 39,4 29 62,2
12 24,1 39,4 33 62,2
13 27,7 39,7 29 63,9
14 24,1 40,1 33 68,9
15 24,7 40,8 31 57,5
16 25,4 40,8 29 46,5
17 28,1 41,8 29 59,3
18 26,1 41,8 29 60,3
19 24,7 41,8 32 62,9
20 24,1 42,1 29 55,6
21 26,1 42,8 35 66,4

22 28,1 43,5 32 61,2
23 28,1 46,8 30 62,9
24 28,7 48,1 30 61,2
25 30,1 51,1 25,5 63,9
2 ~ IMPACT TESTS ON 30 x 30 UNNOTCHED BARS.
The bars have previously sustained 15 blows from a 18 kg striker falling from
2,75 m. None was broken. Bending was achieved by means of a steam-hammer; all
the bars could be fully bent without visible cracking.
Therefore, on the basis of the ordinary tests, all 25 metals would be classified of
good quality and substantially equivalent. On the contrary, the following tests indicate
significant differences.
3 ~ IMPACT TESTS ON NOTCHED BARS.
We assume as starting point the test with the 18 kg striker falling from 2,75 m
and we classify the bars according to the number of blows they have received before
rupture. The columns in the following table provide rupture angles and number of
blows sustained in the remaining tests. With the 50 kg striker, all the bars were broken
after the first blow and we have calculated the fracture work from the rising height of
the striker.
The examination of this table allows to express the following conclusions.
1~ There is no correlation between the results of the tensile tests and those of
the impact tests on notched bars; neither the resistance, nor the elongation, nor the
reduction of area can provide any information on the results of the bending tests on
notched specimens;
All the different metals appear of comparable quality according to the tensile
and impact bending tests; instead, the results of the impact tests on notched bars tend
to show that metals n ~ 20, 6, 4, 12, 13, 3 and 2 are clearly worse than the other and
could cause serious disappointments in service conditions.
2 ~ If we consider that:
the number of kilogrammetres that correspond to one striker blow is not the same
in both cases;
consequently, the numbers of striker blows should correspond approximately to
one unit;
on the other hand, striker tups are not identical (the pendulum tup is much sharper
than the others), and this has influence on the fracture angle;
we conclude that the various falling tests are substantially in accordance; they show
that, keeping everything else the same, the number of blows from the supported
striker is as bigger as the fracture angle is smaller, in other words the bar is more bent
before breaking; the work absorbed by the rupture of the bar, as measured by the
pendulum impact test, also increases as the fracture angle decreases.
The three quantities, bending angle, number of striker blows and fracture work
therefore vary in the same way and can all be used for the classification of metals.
Actually, the table presents a certain number of anomalies from this point of
view; these are the figures in bold characters; but if we examine them carefully, we
realise that such anomalies concern simultaneously the fracture angle and the number
of striker blows or the work absorbed; they should therefore be attributed to
differences amongst the specimens extracted from the same bar rather than to
irregularities in the tests.
As a matter of fact, we did not take special measures in order to avoid those
irregularities in the first series of tests, since these we,re aimed at demonstrating that,
for metals of current fabrication and chosen at random from a steel supplier,
significant differences could be overlooked with conventional tests and become
evident instead when testing notched bars.
But if we want to examine the influence of various details on the results of tests
on notched bars, such as the shape and depth of the notch, the falling height, etc., it is
absolutely necessary to eliminate the differences amongst the bars themselves. It is
well known that, from this point of view, bars extracted from blocks of the same
production lot cannot be considered identical, since the differences can be significant.
In order to eliminate as much as possible this cause of error, we have proceeded
as follows. A fragment of extra-mild steel (boiler plates quality), coming from the
healthiest region of a big ingot (of which it represents just barely the fiftieth part), was
hammer-drawn and rolled into bars. By taking a small fragment of a big ingot, we can
avoid segregation effects and the small bars eventually extracted from a rolled bar can
be considered identical. The big bars have been cut into fragments of 16 c m of length,
which have constituted three groups of small bars; each group has been given a
different heat treatment, in such a way that all the bars of the same group, all treated
however at the same time, were submitted identically to the same temperatures.
We have thus formed three groups of bars that we will identify with the letters
A, B, C, which are identical in terms of chemical composition, and these different
groups have been given different heat treatments; these treatments, which will not be
addressed in detail here, have been conceived in order to create differences among the
metals but without deteriorating them in a definitive way. We have made sure that, by
means of a new heat treatment,, the metal in state A could be brought to state B or C
and vice versa.
Test with 12 kg striker Test with 18 kg slriker Test with 30 kg

Falling height 2,15 m Falling height 2 m pendulum
33 kgm per blow. 36 kgm per blow. Falling height 3 m.
Number
o f the
Work
bar
Number of Number of absorbed in
Fracture Fracture Fractur
striker striker kgm per angle.
blows, angle. blows, angle. square c m o f
section.
20 1 179 ~ 1 175 ~ 1,84 173 ~

6 1 176 ~ 1 177 ~ 2,64 175 ~
4 1 175 ~ 1 176 ~ 3,46 171 ~
12 1 175 ~ 1 172 ~ 3,34 171 ~
13 1 177 ~ 1 170 ~ 4,0 173 ~
3 1 168 ~ 3 135 ~ 18,2 136 ~
2 1 172 ~ 2 162 ~ 4,1 173 ~
1 2 153 ~ 2 148 ~ 18,3 133 ~
25 3 150 ~ 2 156 ~ 14,0 156 ~

8 3 141 ~ 3 145 ~ 23,6 129 ~
17 3 139 ~ 3 140 ~ 16,8 143 ~
16 4 144 ~ 3 145 ~ 18,0 138 ~

23 4 143 ~ 3 141 ~ 16,3 146 ~
9 24 4 146 ~ 3 148 ~ 13,7 143 ~
11 94 143 ~ 3 145 ~ 15,9 144 ~
7 4 139 ~ 3 146 ~ 20,9 132 ~
18 4 137 ~ 3 142 ~ 25,2 128 ~
10 4 135 ~ 3 144 ~ 16,8 142 ~
14 4 132 ~ 3 141 ~ 18,7 131 ~
9 4 130 ~ 4 139 ~ 21,6 130 ~
19 4 144 ~ 4 130 ~ 20,2 139 ~
22 4 140 ~ 4 140 ~ 20,9 141 ~
21 5 131 ~ 5 129 ~ 23,5 131 ~
5 6 115 ~ 5 120 ~ 32,0 116 ~

We have then performed various tests aimed at characterising and

differentiating states A, B and C.
The ordinary tensile tests, performed on bars with diameter of 13,8 m m and 100
m m between reference marks, have given the following results:
Metal Elastic Maximum Reduction

designation limit resistance Elongation0/0 of area
A 22,7 33,7 37,5 77,8

B 23,1 34,5 36 76,4
C 33,9 43,6 29,5 77,1
The two states A and B are therefore substantially identical from the point of
view of the tensile test; state C is significantly more resistant.
We have also performed on the three metals compression tests using cylinders
with 26 m m length and 13 m m diameter; the following table shows the results
obtained:
Metal Elastic Residual heig~htsafter appliedloads of

designation limitper mm= 9000kg 15000 kg 20000 kg
A 36 23 18,3 15,2
B 36,7 23,6 18,3 15,1
C 54,5 24 21 17
The two met~ils A and B are still practically identical; metal C is remarkably
more resistant. If we calculate the ratio between tensile and compressive elastic limits
we find the same value for the three cases, 0,63 for A, 0,63 for B, 0,622 for C.
The three metals have been impact tested in the form of bars with 30 x 30 turn,
tested on supports 16 cm apart, with a 18 k g striker falling from 2,75 m. None of the
bars was broken and all three could be completely bent without visible cracking. The
impact bending test, therefore, does not indicate any difference among the three
metals A, B and C.
However, the impact test on notched bars has indicated a remarkable difference
among these metals.
The 30 x 30 section bars, notched for half of their thickness with a round
bottom notch (4 m m radius), and tested on supports 12 c m apart with a 18 k g striker
falling from 2,75 m, have given the following results:
Designation Numberof strikerblows

of metal before rupture Fracture angle
A 5 45 ~
B 1 166 ~
C 7 52 ~
The differences are here very evident; metals A and C break substantially under
the same angle, but more striker blows are needed to give such fracture angle to metal
C, whose resistance is therefore higher. On the other hand, metal B appears clearly
brittle and presents a grain-type fracture and a very high fracture angle, something
that none of the previous tests would suggest.
We have investigated whether this difference remained when test conditions

were modified, and we have eventually tried impacts from a lower height, static
bending, notches with round and sharp bottoms and different depths; the results
obtained are summarised in the following tables:
1~ ROUNDBOTTOMNOTCHWITH4 mm RADIUS
18 kg striker. Failing height 1,50 m.
Designation Number of blows Fracture

of metal, before rupture, angle.
A 11 60 ~
B 5 115 ~
C 10 62 ~
2 ~ TRIANGULARSHARPBOTTOMNOTCHWITH 15 mm DEPTH
18 kg striker. Falling height 1,50 m.
Defignation Number of blows Fracture

of metal, beforerupture, angle.
A 7 110 ~
B 4 150~
C 10 63 ~
3 o TRIANGULARSHARPBOTTOMNOTCHWITH3 mm DEPTH
18 kg striker. Falling height 1,50 m.
Designation Numberof blows Fracture

of metal, before rupture, angle.
A 21 20 ~
B 6 118 ~
C 30 15 ~
4 ~ ROUNDBOTTOMNOTCHWITH4 mm RADIUS.DEPTH 15 ram.
Tested under static pressure.
Designation Fracture Maximumstress

of metal, angle, per squaremillimetre.
A 18~ 6,2 kg
B 98 ~ 7,3 kg
C 36 ~ 8,2 kg
5 ~ TRIANGULARSHARPBOTTOMNOTCH. DEPTH 15 ram.

of metal, angle, per square millimetre.
A 62 ~ 7,1 kg
B 100 ~ 7,8 kg
C 31 ~ 8,6kg
6 ~ TRIANGULAR SHARPBOTTOMNOTCH. DEPTH 3 ram.

of metal, angle, per square millimetre.
A 23 ~ 10,9 kg
B 75 ~ 10,5 kg
C 50 ~ 13,1 kg
Examining these tables, we see that the difference observed among the metals
A, B, C for the first impact test described and which is evident particularly in terms o f
fracture angle, can be found in all these different tests, but with more or less
pronounced evidence, we can say that, keeping all conditions the same, the
differentiation is more evident in the impact test than in the static bending test, and
more evident for the greater falling heights than for the smaller; with the same depth,
the sharp bottom notches emphasize more the difference as compared to the round
bottom notches, and, with the same shape, deeper notches seem to be more influential.
Nevertheless, we observe that in all tests metal B breaks more easily and under a
remarkably greater angle than both metal A, that has the same tensile test results, and
metal C, which appears significantly more resistant in the tensile test.
The fact that such difference can be remarked in tests under static pressure as
well as in impact tests, seem to indicate that it is not a question o f the metal being
more or less brittle. The difference observed among metals A, B, C can be explained
by the localization of deformation caused by the notch, and not by the use of a
dynamic action. If this is true, we should be able to find again such difference in other
tests than the bending test, under conditions where deformation is localized by means
of a notch.
Indeed, by testing tensile bars in which a circular notch was machined, a
difference is clearly observed among metals A, B, C that is particularly evident in
terms o f fracture; that of metal B is grain-type, whereas fractures of metals A and C
are clearly fibrous, as it happens in bending tests on notched bars. At the same time,
the deformation o f the notched part is clearly smaller for B than for A and C;
however, the measurement o f such deformations inside the notch is too difficult to
obtain numerical values.
Before drawing conclusions from the results we have presented, we will report a
second series of similar tests, carded out on a special metal of a rather resistant
quality.
Rolled bars o f this metal, prepared with the precautions previously described,
have been subjected to two different heat treatments that allowed to obtain two states
which we will code by the letters D and E.
On these metals D and E, we have performed the same series o f tests as on
metals A, B, C. The results are summarized in the following.
TENSILE TESTSONBARS WITH 13,8 m m DIAMETER

AND 100 m m BETWEENREFERENCEMARKS.
Maximum Elongation Reduction

Metal. load. 0/0 of area.
D 97,5 13,4 57,5

E 93,7 15,0 57,5
The results obtained from the tensile tests are therefore very similar for the two
metals.
COMPRESSIONTESTON CYLINDERSWITH 13,8 m m OF HEIGHT

AND 8 m m OF DIAMETER.
The results are once again very similar for the two metals.
RESIDUAL HEIGHT UNDER A LOAD OF
METAL
3ooo 4ooo ,ooo 1~
D 13 13,0 12,8 12,4 11,2 10,1 8,0
E 15 12,8 12,4 11,6 10,7 9,4 7,8
IMPACTTESTON UNNOTCHED30 x 30 m m BARS.

18 kg striker, f a l l i n g height 2, 75 nx
The bar of metal D does not break and can be fully bent with a press without
breaking. The bar of metal E breaks when the bending angle reaches 66 ~ Therefore,
in the case of this resistant quality metal, the differentiation takes place already for the
unnotched bend bar.
Here are the results of the impact tests on notched bars.
18 kg striker.
NUMBER
OF STRIKER
FALLING FRACTURE
NOTCH METAL BLOWS
HEIGHT ANGLE
BEFORE
RUPTURE
m. degrees.
Round bottom, radius 4 D 136
m m , depth 15 m m
2,75 E 164
Round bottom, radius 4 D 6 107
m m , depth 15 m m
1,50 E 3 166
Triangular notch, sharp D 7 106
bottom, depth 15 m m 1,50 E 1 166
Triangular notch, sharp D 20 115
bottom, depth 9 m m 1,50 E 9 165
The following table gives the results of the analogous tests performed under
static pressure.
LOAD
FRACTURE
NOTCH METAL SUSTAINED
ANGLE.
PERmm2
degrees
Round bottom, radius 14 D 140 16,7
ram, depth 15 m m E 155 15,4
Triangular notch, sharp D 96 18,0
bottom, depth 15 m m E 165 17,1
The results are of the same nature as those obtained on the mild steel; the two
metals D and E, which appear substantially identical in the tensile and compressive
tests, are clearly different from the tests on notched bars, both under impact or static
pressure; metal E appears consistently inferior to metal D, both in terms of number of
striker blows sustained and in terms of fracture angle.
Two bars with a 15 mm deep notch with round bottom have been tested on the
pendulum. The work absorbed by the rupture was found to be 0,20 kgm for metal D
and 0,11 kgm for metal E. The fracture angles were respectively 143 and 160 ~
The tensile tests on notched bars have shown the difference in the texture of the
two metals, similar to the bend test and with more evidence than in the case of the
mild steel. The difference in resistance to rupture of the two metals becomes more
significant on the notched bars. Thus with a 3 mm deep notch with triangular section,
we record a strength of 127 kgfor metal D and only 109 k g for metal E.
The less ductile is the metal, the clearer seems this differentiation in the tensile
test on notched bars.
In this regard, we will mention the following experiments carried out on tWO
metals of hard quality.
METALF METALG
Elastic limit 38 kg 81kg

Tensile test M a x i m u m load 124 kg ll7kg
on unnotched bar Elongation • 100 6 6
Reduction of area 52,3 50,2
Tensile test on
bar vcith triangular M a x i m u m load 95,4 kg 57,5 kg
notch with 3 mm depth
Impact test on
notched bar Number of striker blows 6 1
with round bottom notch Frac~re angle 155 ~ 180 ~
and 15 mm depth
I f we recapitulate all the tests described in this chapter, we see that:

1~ The procedure for bend testing notched bars can emphasize differences that
the ordinary tensile test leaves completely hidden;
2 ~ Such differences are not caused by variations in the relative values of the
tensile and compressive elastic limits, since for metals A, B, C the ratios between
tensile and compressive elastic limits are rigorously equal;
3 ~ Such differences appear in the static as well as in the dynamic tests.
As we have already remarked above, we are therefore led to correlate them to a
property for which the word 'fragility' appears ill-chosen, since it seems to imply the
intervention of an impact. We believe we can satisfactorily interpret the results
mentioned above by resorting to the concept of ultimate resistance to rupture,
introduced by Poncelet; but, in order to give this concept a physical meaning which is
more precise and more independent from the test circumstances, we will consider the
ultimate resistance to rupture of an infinitely thin slice; we will propose to give this
quantity the name of resilience, which has already been employed, specifically by Mr.
Russell, with an analogous meaning. Therefore, we will name resilience of a metal the
ultimate resistance to rupture of an infinitely thin slice, or the work needed to produce
such rupture expressed in kilogrammetres per square centimetre of cross section.
How can we measure resilience? Based on the definition itself, we need to
measure the rupture work in a test in which deformation is limited to the rupture
section. We cannot expect to obtain such an indication in an absolute way, but the
different mechanical tests will provide resilience values which are closer to the exact
value, the more localized is the deformation around the rupture section. It is therefore
well understood that the ordinary tensile test would give the worst results from this
point of view, since the observed rupture work will differ the most from the
elementary work of rupture or resilience. In order to achieve more exact values of this
parameter, we will need to find test methods which can localize deformation, that is
resort to bending tests, machine deep and sharp notches, and finally use dynamic
actions with impact velocities as high as possible.
From a practical point of view, we have been forced to abandon the sharp notch
because of the difficulty in giving it a proper definition. The tests described in this
essay show that, using a sufficiently deep rounded notch, we can succeed in
classifying different metals, and we can maintain that such a classification is very
similar to what we would obtain through the exact measure of resilience.
Based on this, the test on notched bars would not be a fragility test; it would be
a test that allows the classification of metals into metals with elevated resilience and
metals with low resilience. The term "fragile" should be reserved to metals that
behave in a different way when tested statically and dynamically; among the tests
described in this paper, we can't find any example of this; the difference in the work
needed for the rupture of the same metal, both under the influence of more or less
rapid impacts and under the influence of a static pressure, can always be attributed to
the fact that, in impact tests, the localization of deformation is more pronounced than
in tests under gradual pressure; this is easy to verify if any of the tests indicated by
Mr. Fremont and Mr. Osmond are performed: for instance, by observing the amount
of the surface, previously polished, of a bar which turns dull around the notch.
Working in this manner, we have always remarked that the differences observed in
the rupture work of a same metal tested in different conditions are correlated to the
extension of the deformed area.
On the other hand, there is no point in trying to distinguish whether the metal
presents different values of resilience in the dynamic and static tests. We must
eliminate all metals that, after being notched, are likely to fracture under a small
impact, whether the same weaknesses remain or not in the static pressure tests. In
practice, the impact test on notched bars will therefore be sufficient.
However, in order to test metals from this point of view, it will not be necessary
to measure the work absorbed by rupture; the angle which produces such rupture
already provides a very important information. From the results shown in the table of
tests performed bn a series of 25 metals, we have seen that, for metals of the same
quality, the work of rupture varied regularly with the rupture angle.
This is obviously incorrect when metals of different quality are compared; the
rupture work corresponding to a given rupture angle is in this case proportional to the
resistance of the metal. This clearly follows from the following tests concerning two
series of steels:
MILD STEELS HARD STEELS

?=38t~4Okg ?=45to??kg
FRACTURE ANGLE RUFIXJRE WORK FRACTURE ANGLE RUPTURE ANGLE
139 ~ 16 148 ~ 20
146 ~ 15 160 ~ 15
156 ~ 14 170 ~ 13
162 ~ 4 172 ~ 12
176 ~ 4 175 ~ 10
IV
Conclusions.
In this essay, we have attempted to collect results which can demonstrate the
importance of the test on notched bars and to define the conditions which can lead to
the formulations of practical specifications. By comparing such tests with those
described in the memoirs which have been mentioned in the first part of this work, we
believe that the following conclusion can be drawn:
Presently, metals are classified on the basis of the their resistance to rupture in
tensile tests and their ductility, measured by the elongation before rupture; at the same
time, impact bend tests make sure that the metal does not present excessive fragility.
It seems that there could be a more rational classification method, by invoking a
quantity that we propose to call resilience, and that represents the ultimate resistance
to rupture of a thin slice. This quantity is clearly useful from a practical point of view,
since it intervenes whenever the metal undergoes a local deformation, or in other
words whenever there is a weakened section operating. In order to measure it, one
needs to perform tests in which deformation is localized in a region very close to the
rupture section; the tensile test does not provide any information to this regard; the
test which appears to be the most appropriate is the impact bend test on notched bars.
The usefulness of this test, regardless of the considerations which have been
expressed, is shown by the fact that the classification that it provides is totally
different from that provided by the tensile test, and that it allows clear d~fferenUatlon
between mild steels which appear identical from the ordinary impact test.
In order that this test be adopted in the current acceptance practice of metals, it
is absolutely necessary that all details be defined without any possible ambiguity. The
most important point is the preparation of the notch, which has a considerable
influence on the results. We propose to adopt a notch with cylindrical bottom,
prepared with a tool working perpendicularly to the notch, in order to avoid any
scratch which could act as a sharp notch. The differentiation among metals, even the
mildest, is easily achieved with the round bottom notch, provided that such notch is
sufficiently deep.
The dimensions of the bar should not be too small, in order to minimise the
influence of positioning errors and small local irregularities.
As far as the rupture mode is concerned, in principle it is preferable to use a
procedure that allows a numerical measure of the rupture work. We have described a
pendulum device that provides such a measure with a superior exactness, in our
opinion, than the spring-based systems which have been proposed for the same
purpose. However, in the practice of acceptance tests, we believe that such a
measurement is unnecessary, and that it is sufficient to employ a simpler device, such
as the striker used for ordinary impact tests. A sufficiently clear differentiation is
obtained among the metals, by counting the number of blows needed to achieve
rupture and marking the angle corresponding to rupture. Keeping everything else
constant, these two parameters vary in accordance with the resistance of the metal.
Independently from resilience, one should also consider, in the characterization
of a metal, the difficulty with which such metal undergoes early deformation; such a
difficulty is measured, in a tensile test, by the elastic limit and can be evaluated, in the
impact bend test, by the bend angle produced by a striker falling from a predefined
height. The possibility of performing such evaluation, thus avoiding the tensile test or
a similar one, makes also the method of rupture by repeated impacts from a striker
superior to methods in which rupture is achieved by a single blow.
In summary, we believe that the impact bend test on notched bars should be
used for the acceptance of metals, especially in the case of mild steels. As a
preliminary indication, the following form of specification seems to us acceptable in
practice; obviously, different conditions should be numerically established in any
particular case, based on previous experiences.
The test bars (square cross section of 30 x 30 mm or 20 x 20 mm or 30 or 20
ram, depending on the thickness of the plate) shall be notched for (one half or one
third) of its thickness; the bottom of the notch shall be formed by a cylinder having a
radius of(2 or 4 ram).
The bar, placed on two supports spaced (12 cm), shall receive an impact from a
(18 kg) striker, falling from a height of(2 m).
The edge of the striker and the bar supports shall have a shape such that the bar
can be bent to an angle of (60 ~ or 90 ~ or 125~ The bar shall sustain bending to such
angle without breaking; the number of striker blows needed to achieve such a bending
angle shall be at least (3 or 4 or 6).
Reference Energies, Machine
Stability, and Calibration
C. N. M~Cowan, 1 J. Pauwels, 2 G. Revise, 3 and H. Nakano 4
International Comparison of Impact Verification Programs
Reference: McCowan, C. N., Pauwels, J., Revise, G., and Nakano, H.,
"International Comparison of Impact Verification Programs," Pendulum Impact
Testing: A Century of Progress, STP 1380, T. A. Siewert and M. E Manahan, Sr.,
Eds., American Society for Testing and Materials, West Conshohocken, PA, 2000.
Abstract: A horizontal comparison is made between the four laboratories that certify
Charpy impact verification specimens. The participants in this study were Japan (NRLM),
France (LNE), the European Commission (IRMM), and the United States (NIST). The
exercise was conducted to show how the impact verification programs, specimens, and
test procedures compare with each other. Results for both 8 and 2 mm strikers were
compared. The study showed the following: (1) The certified energies of impact
verification specimens distributed by these four metrological authorities often agreed
within 1% of the average values determined in this study; (2) the variation in energy for
the specimens was low, typically bracketed by a coefficient of variation of between 0.02
and 0.04; and (3) the energies measured for the tests performed with 2 and 8 mm strikers
on the 4340 steel specimens were nearly equivalent, but a trend of slightly higher energy
for the 2 mm tests is indicated.
Keywords: Charpy impact verification, impact testing, verification testing, verification

specimens
Introduction
Charpy impact testing is often specified as an acceptance test for structural materials,
and companies performing acceptance tests are typically required to verify the
performance of their impact machine using certified verification specimens. To our
knowledge there are only four laboratories in the world that certify and distribute
reference materials for the verification of Charpy impact machines: (1) The Institute for
Reference Materials and Measurements (IRMM, Belgium), (2) Laboratoire National
D'Essais (LNE, France), (3) The National Institute of Standards and Technology (NIST,
1 Materials Reliability Division, NIST, 325 Broadway, Boulder, CO, U.S.A., 80303
2 Reference Materials Unit, EC-JRC-IRMM, Retieseweg, B-2440, Geel, Belgium
3 Mechanical and Equipment Testing, LNE, 5 Avenue Enrico Fermi, 78197, Trappes Cedex, France
4 Materials Measurement Section, NRLM, 1-4, Umezono 1-Chrome, Tsukuba, Ibarki, 305, Japan
73
USA), and (4) The National Research Laboratory ofMetrology (NRLM, Japan). These
four laboratories supply impact verification specimens to verify the performance of an
estimated 1800 impact machines annually.
This study provides the first horizontal comparison between these four laboratories.
We compare both the impact verification specimens and the machines (or systems) used to
certify the absorbed energies of the specimens. Our goals are to use this comparison to
better understand the details of each other's verification programs, and to consider how
any differences between our verification systems and specimens might affect the users of
our respective programs. It is only through these types of horizontal comparisons that we
can assess the equivalency of the results for impact acceptance tests made around the
world. These initial results help confirm that acceptance tests performed under one system
are equivalent to the others, making the verification systems and specimens transparent to
the user.
Background Information on Verification Programs
The four verification programs represented in this report have similar goals and much
in common, but each program is unique. To start with, there are two fundamentally
different approaches used to stabilize the certification procedures for impact verification
specimens. The European philosophy for stabilizing their certification procedures (EN
Standards) is based on traceability to "master specimens." This system is represented here
by IRMM, which currently certifies and distributes impact verification specimens for the
Community Bureau of Reference (BCR). The stability of the national impact verification
programs in the United States, Japan, and France is based on traceability to designated
impact reference machines that are maintained by the respective metrological authorities
(NIST, NRLM and LNE). This difference and details of the certification procedures
discussed below (and given in Tables 1 and 2) make direct comparisons of our systems
difficult. However, keeping these differences in mind, we can make useful comparisons of
our results.
I R M M P r o g r a m - In the IRMM program, "master batches" of impact verification
specimens are tested in a round robin for the BCR. From the results of the rotmd robin a
certified energy for the specimens is determined, and these "master specimens" are then
tested over time to track and normalize the results of the impact machine that is used to
determine certified values for BCR verification specimens. The certified values for BCR
impact verification specimens are determined as follows: (1) New batches of BCR impact
verification specimens are tested (30 samples per batch) on a single impact machine,
together with sets of "master specimens" (generally 25 or 35 specimens) of similar
nominal energy; (2) if the samples have acceptable variation and energy, the difference in
the average energies for the "master specimens" and the batch being evaluated is
determined for the machine; (3) this difference is added or subtracted from the average
energy for the batch of BCR verification specimens to determine a certified energy
value[l]. These procedures meet the EN 10045 and ISO 148-3 requirements for
McCOWAN ET A L ON IMPACT VERIFICATION PROGRAMS 75
verification specimens [2,3]. Table 1- Summary of the verification

For this study, the results of the impact requirements for national and
machine used by IRMM were reducedby 4% international standards: difference
to represent the BCR procedure, so these allowedfor the verification result and
results represent the certification system, not a the certified energy value (E~).
single impact machine. The IRMM specimens
can be used to verify impact machines in
accordance with EN 10045 and ISO 148-2 [4]. Designation ] Requirements
NRLMProgram - In Japan, there are two ISO 148-3 i 2 J for Eo g 40 J
C-type impact machines in the program, but a +5% for Ec > 40 J
single impact machine is used to determine the ISO 148-2 +4 J for E~ _<40 J
certified values for the verification specimens. +10 % for E~ > 40 J
The second machine is used as a back-up EN 10045 • J for Ec _<40 J
machine and for comparison to the machine • % for Eo > 40 J
used for certifications. Because a single ASTM E23 • J for Eo _<28 J
machine is used to determine the certified ASTM 1236 • for Eo > 28 J
value for the verification specimens, the results JIS B 7722 • J for Ec _<40 J
for the NRLM impact machine used in this • % for Eo > 40 J
study represent those that would be attained JIS B 7740 • J for Eo _<40 J
for the Japanese certification system. The • for Ec > 40 J
certified value for the NRLM specimens is
determined by testing 25 specimens. The
results are evaluated and if all statistical criteria are met, the average energy of the 25
specimens can be used as a certified value. This procedure meets the JIS B 7740 and the
ISO 148-3 requirements for verification specimens (the requirements for the JIS B 7740
and ISO 148-3 standards are identical) [5].
The impact specimens certified by the NRLM program are primarily used to verify
the performance of impact machines according to JIS B 7722, which is the national
standard for impact testing in Japan [6]. The NRLM specimens can also be used for the
ISO 148-2 machine verification (again, the requirements for the JIS and ISO standards are
identical).
NISTProgram - In the United States, the certified values for impact verification
specimens are determined using three impact machines (2 U-types and 1 C-type). The
certified value is determined as follows: (1) 25 specimens are tested on three different
impact machines, (2) the results are evaluated to determine whether the differences in the
variation for the specimens and the average energies of the three machines meet
established criteria, (3) if the specimen variation is acceptable and the comparison of
results for the three machines are within normal bounds, the results are combined and the
certified value is defined as the average value for the 75 specimens. The certified
specimens meet the ASTM E 1236 Standard for Qualifying Charpy Impact Machines as
Reference Machines and ISO 148-3 requirements for verification specimens. In this study,
only one impact machine was used, so the NIST results represent that machine, not the
certification system. However, at energies above 40 J, the results for the NIST machine
are expected to closely represent the average value of the three NIST machines. For the
low energy (<40 J) specimens, the results for this single machine are often higher than the
average of the three machines used to determine a certified value for NIST specimens.
The NIST specimens are used primarily to verify the performance of impact
machines according to ASTM E 23, the Standard Test Method for Notched Bar Impact
Testing of Metallic Materials. The specimens can also be used for ISO 148-2 verification
tests.
L N E P r o g r a m - In France, two impact machines are used to certify verification
specimens (a U-type and a C-type). The machines are located at two different laboratories
in France (LNE and CTA) which have an agreement from the French Commtmity of
Accreditation (COFRAC) for the certification of impact verification specimens. The
procedure used to develop a certified value is in accordance with EN 10045-2 and is
s i m i l a r to t h e N I S T p r o c e d u r e : 25 s p e c i m e n s are t e s t e d o n e a c h m a c h i n e , a n d t h e g r a n d
a v e r a g e (50 s p e c i m e n s ) is u s e d as t h e c e r t i f i e d v a l u e f o r t h e s p e c i m e n s i f s t a t i s t i c a l
e v a l u a t i o n s s h o w a c c e p t a b l e p e r f o r m a n c e for t h e m a c h i n e s a n d t h e s p e c i m e n s . In this
Table 2- Details on the specimens and impact machines used in the study. Under
machine information, the pendulum type (C or U), the striker radius which was
tested, the method of reading the absorbed energy (dial or encoder), and the
maximum capacity of the machine are given.
Laboratory Specimen Information Certified Values
Machines
Test Energy Steel Striker Value Details
Temperature Level Type Radius (J)
(~ (mm)
1 20 1 2 27.0 - C- type
4340 - 2 and 8 mm
8 26.8
- Dial
20 2 2 113.6 - 300 J
4340
8 112.6
2O 3 XM32 2 157.1
2 20 1 2 26.4 - U-type
E.40 - 2 and 8 mm
20 2 2 67.2
NCD7 - Encoder
20 3 2 111.8 - 350 J
3 -40 1 8 16.5 - U- type
4340 - 2 and 8 mm
-40 2 8 99.2
- Encoder
20 3 T-200 8 258.0 '- 350 J
4 0 1 2 * - C-type
4340 - 2 and 8 m m
0 2 2 *
- Dial
0 3 2 * - 500 J
* The certified values were not available because the lot size for the specimens was too
small to both certify the specimens and provide the n u m b e r o f specimens required for this
study.
McCOWAN ET AL. ON IMPACT VERIFICATION PROGRAMS 77
study only the U-type impact machine was used, so the LNE results represent that
machine, which typically has a slightly higher absorbed energy than the C-type used by
LNE. The LNE specimens can be used to verify impact machines according to EN 10045
and ISO 148-2.
Materials and Procedures
Each of the participating laboratories provided 100 verification specimens at three

different levels of absorbed energy, referred to as levels 1, 2, and 3 in Table 2. The
laboratories are coded using numbers (1-4). The specimens provided by each laboratory
are also coded (1-4) to identify the laboratory that supplied the specimens.
Information on the impact machines used for the testing is given in Table 2, along
with the test temperature specified for each group of specimens. The certified absorbed
energies for the specimen groups, which each laboratory had previously determined for the
specimens, are also given in Table 2. (These values were not made available to the
participants until the testing was completed.)
All four laboratories used a similar type of steel to produce verification specimens for
energy levels 1 and 2 (Table 2). The steel is a medium-carbon, low alloy, high-strength
steel, designated in the U.S. as AISI/SAE 4340. Although these steels are very similar in
composition, the heat treatments used by the various laboratories are assumed to vary
significantly.
The types of steel used to produce verification specimens for energy level 3 differ
between the various laboratories. So at this energy level, differences in impact properties
due to alloy content must be considered. Basically, 3 types of alloys were used for energy
level 3: (1) a XM32 steel, which is a Cr- Mo type alloy, (2) a T-200 steel, which is an 18
Ni-0.7 Ti maraging steel, and (3) the AISI/SAE 4340 type steel, which is also used for
energy levels 1 and 2.
The impact testing was conducted using the procedures normally followed by each
laboratory, and any special instructions that were provided with the specimens. From each
specimen group of 25 specimens, 15 were tested using a 2 mm striker, and 10 were tested
Using an 8 mm striker. Some outlier data were removed during testing, but most of the
statistical outliers were left in the data for our analysis. 5
Specimen Variation
The variation in the absorbed energy for the specimen groups is given in Table 3.
The results are presented in terms of the coefficient of variation (CV), which is the ratio of
the standard deviation in the absorbed energy to the average absorbed energy of the
specimen group being evaluated. This relative measure provides a convenient means of
normalizing the various energy levels being considered here. A CV of about 0.04 (or less)
represents a low level of the variation in absorbed energy for impact verification
5 Statistical outliers are defined here as those data that were identified as outliers in box-and-whisker
analysis. For the most part, data were removed during testing only due to a testing problem
(specimens placement, etc.).
specimens. This CV value Table 3 - The coefficient of variation is given for the 4
is related to sample size laboratories, organized by energy level, striker ( tup),
calculations (for a sample specimen group. The pooled CV (CVP) of the
size of 5), and would laboratories is also given. The CV is defined as the ratio
generally ensure a variation of the standard deviation and the average absorbed
small enough to conduct a energy for the group.
verification test within the CVs for Labs 1-4, and the
statistical considerations Energy TUP Specimen pooled CV (CVP)
upon which our verification Level (mm) Group
programs are designed. The CV1 CV2 CV3 CV4 CVP
ISO 148-3 requirements for
1 2 1 0.02 0.04 0.04 0.04 0.04
verification specimens allow
1 2 2 0.02 0.03 0.02 0.05 0.03
a CV of 0.05 or less.
1 2 3 0.02 0.07 0.04 0.02 0.04
The pooled CV is
considered the best 1 2 4 0.02 0.02 0.02 0.03 0.02
indicator of specimen 1 8 1 0.02 0.02 0.02 0.04 0.03
variation, due to its 1 8 2 0.03 0.04 0.03 0.06 0.04
inherently larger sample 1 8 3 0.05 0.06 0.03 0.02 0.04
size. It is defined as the 1 8 4 0.02 0.03 0.04 0,01 0.03
rms value: the square root 2 2 1 0.04 0.04 0.05 0.03 0.04
of the sum of the squares of 2 2 2 0.02 0.02 0.02 0.01 0.02
the CVs (for laboratories 1-
2 2 3 0.02 0.03 0.03 0.02 0.02
4) divided by 4.
2 2 4 0.02 0.02 0.03 0.02 0.02
At the lowest energy
2 8 1 0.04 0.03 0.03 0.03 0.03
level evaluated, level 1, the
variation for the 8 and 2 2 8 2 0.02 0.01 0.02 0.02 0.02
mm tests are similar. 2 8 3 0.02 0.02 0.03 0.02 0.02
Considering the pooled 2 8 4 0.02 0.03 0.03 0.03 0.03
CVs, no clear effect of 3 2 1 0.02 0.02 0.04 0.04 0.03
striker radius on the 3 2 2 0.03 0.04 0.03 0.04 0.04
variation is apparent. 3 2 3 0.12 0.15 0.03 0.11 0.II
Occasionally one laboratory 3 2 4 0.03 0.03 0.03 0.04 0.03
had a high or low CV, with 3 8 1 0.03 0.03 0.03 0.04 0.03
respect to the other three
3 8 2 0,04 0.04 0.03 0.02 0.03
laboratories. But typically
3 8 3 0.03 0.05 0.03 0.03 0,04
the high CVs reported for
the level 1 specimen groups 3 8 4 0.05 0.05 0.04 0.05 0.05 i
were due to one or two
statistical outliers in the
data. Generally, the variation for the level 1 specimens was low, and all of the pooled
CVs are 0.04 or less.
At energy level 2, the interlaboratory CVs for the various specimen groups are even
more consistent than the results for level 1. Typically, three laboratories had the same~CV
for a specimen groups, and one would differ slightly. All of the pooled CVs are 0.04 or
less; many are 0.02, Again, there was no apparent effect of striker radius on the variation.
At energy level 3, the interlaboratory CV results are slightly less consistent than the
level 2 results, and differentiate the results for several specimen groups. The 2 mm tests
for specimen group 3 have a significantly higher CV than the other level 3 groups (pooled
CV equal to 0.11), and the 8 mm tests for specimen group 4 had interlaboratory CVs
slightly higher than 0.04 (pooled CV equal to 0.05). Both of these specimen groups,
however, had acceptable CVs for tests made using the striker radius that they were
developed for: The CVs for the 8 mm tests on the group 3 specimens and the 2 mm tests
on the group 4 specimens were 0.04 and 0.03 respectively. So, for the more ductile high-
energy specimens, striker radius apparently had an effect on the variation in absorbed
energy.
In general, the interlaboratory CV for the various specimens groups were in good
agreement and appear to reflect differences in the specimens, not the impact machines.
However, laboratory 1 had the lowest variation most often, and this likely indicates some
influence of the impact machine on the variation assigned to the specimens.
Absorbed Energy
The absorbed energy results for the specimen groups are shown in Figures 1- 6 and
the data are given in the Appendix. The plots show the average absorbed energy for each
specimen group, with error bars representing + 1 standard deviation. The results for each
energy level are shown separately for the 2 and 8 mm tests. For comparison purposes, a
fifth average and error bar were added to the data for each specimen group. This bar
shows the average absorbed energy of the four laboratories for each group (the grand
average) and the ASTM E23 and EN 10045 limits for verifying the performance of
reference quality impact machines (see Table 1). These error bars offer convenient scales
for comparisons of these data, but the limits have no real significance for the horizontal
comparisons we make here. The data for each laboratory, within each specimen group,
are plotted from left to fight as laboratory 1, 2, 3, and 4.
Energy Level 1- As shown in Figures 1 and 2, laboratory 4 always had the lowest
absorbed energy for the specimen groups tested at this energy level. The highest values
were always from laboratory 2 or 3,' and laboratory 1 always had the intermediate values.
The relative relationships between these results were found consistently for both the
2 and 8 mm test results, but the magnitudes of the differences between the results were
dependent on the particular specimen group that was tested. The apparent differences in
the results for the laboratories were greatest for the tests made with specimen groups 1
and 3, slightly smaller for specimen group 2, and smallest for specimen group 4 (• J).
The average results for the tests were all within • 2 J of the grand average
determined for the energy level 1 specimen groups, and most were within + 1.4 J of the
grand averages. So these results are within the limits deemed acceptable for reference-
quality impact machines. The range in the results, however, pushes these limits. This is
partly due to differences in the designs of the C and U-type impact machines represented
here. The impact machine used by laboratory 4, for example, may be the most rigid
machine in this study (large capacity C-type machine) and the lower values produced by
this machine may be very good values (less energy absorbed by the machine in the test).
35 i i i I 35 I I I I
30 30
.•=25 @ASTN
~EH
.•25
r 0,)
g:
LU 20 LU 20
15 15
I I I I I I I I
10 10
o 1 2 3 4 o 1 2 3 4
Specimen Code Specimen Code
Figure 1- Energy level 1, 2 mm striker. Figure 2 - Energy level ], 8 mm striker.
130 i i i i 130 1 1 I I
110 110
--j
100 ~0-)100
b_
~)
~ 90 ~ 90
80 80
,o
I I I I
60 60
o 1 2 3 4 1 2 3 4
Figure 3- Energy level 2, 2 mm striker. Figure 4- Energy level 2, 8 mm striker.
300 I I I I 300 [ i i i
250 250
---j
>..
~200 200
C
UJ LU
150 12o
I I I I
100 100
o 1 2 4 o 1 2 3 4
Figure 5- Energy level 3, 2 mm striker. Figure 6- Energy level 3, 8 mm striker.
McCOWAN ET AL ON IMPACT VERIFICATION PROGRAMS 81
It must also be kept in mind here that the average values for two of the laboratories would
likely be lower if the C-type machines that are also used in their verification programs
were included in their averages. This would decrease the range in the results.
The 2 and 8 mm test results were equivalent, considering the standard deviations
associated with them. But there is a trend for the 2 mm results to have slightly higher
absorbed energies. The average differences between the 8 and 2 mm results for each
specimen group were as follows: (G1) -0.01 J, (G2) +0.13 J, (G3) +0.44 J, and (G4)
+0.06 J. The pooled standard deviations associated with the average energies for these
specimen groups were in the range of 0.5 to 1.0 J.
E n e r g y L e v e l 2 - The results of the laboratories for the level 2 specimen groups
(Figures 3 and 4) are well within 4- 5% of the grand averages determined for the groups
(none exceed • 3%). The results do not show significant differences between the
laboratories, and the effect of striker radius is small. In general the 2 mm test results have
slightly higher energy than the 8 mm results: the average differences between the 2 and 8
mm tests for the laboratories were 1.6, 1.1, 1.1, and 1.5 J for the specimen groups 1-4
respectively. In most cases, the differences are less than the standard deviations
associated with the average energies of the specimen groups being compared.
E n e r g y L e v e l 3 - The tests at energy level 3, Figures 5 and 6, show a significant
difference in the results for the 2 and 8 mm striker radius. The differences between the 2
and 8 mm test results are apparently related to the type of steel used for the specimens as
much as they are to the energy level of the specimens. For the specimens made from
4340-type steel (groups 2 and 4), the differences in the 2 and 8 mm tests were the
smallest. The group 2 results were mixed with respect to the influence of the striker
radius and the differences (average difference of-0.7 J) were not statistically significant.
The results for the group 4 specimens consistently show a higher energy for the 8 mm
tests, and the differences (average difference of-4.8 J) are on the order of one standard
deviation. For the specimens made from the XM32 steel (group 1), the 2 mm test results
were consistently higher than those for the 8 mm tests. The average difference for the
group 1 specimens was +10.0 J, which is statistically significant (standard deviations of the
average energies for group 1 were around 5 J). The tests with the specimens made from
the maraging steel (group 3) had the largest difference in 2 and 8 mm results (average
difference of-44.7 J), and this shows the results for the 8 mm tests to be higher than those
for the 2 mm tests. So the results for two steels (XM32 and maraging) show significant
differences for 2 and 8 mm test results, but the effect of the striker radius was quite
different (opposite).
The results also show that the group 3 maraging-steel specimens are better
verification specimens for the 8 mm test, for which they were developed. The variation in
the absorbed energies for the 2 mm tests were much higher than those for the 8 mm tests,
and the differences in the intedaboratory average energies were also greater for the 2 mm
tests. So it appears that the group 3 maraging-steel specimens perform adequately as 8
mm verification specimens, but are a poor choice for 2 mm verification tests. The group
1, 2, and 4 specimens all performed well as verification specimens for level 3, for both 2
and 8 mm tests ( all within the • limit, and most within the 4-5% limits).
Lateral Expansion 30o . . . . I , , , , i . . . . I . . . . I ' ' ' '
The lateral expansion for the specimens, " 240

Figure 7, increased with increasing absorbed >~
energies, as expected. The correlation is |
linear up to about 150 J. The 2 mm tests ~ 180
9
resulted in slightly increased lateral expansion
(and absorbed energy) for most of the ~ 120
specimen groups. The exception was <'~
9 Tup
specimen group 3 (level 3), where the 8 mm 60
o 8
tests had higher lateral expansion. This result
~ o2
is consistent with the higher absorbed energies
for the 8 mm tests on these specimens. 0.0 0.5 1.0 1.5 2.0 2.5
Lateral Expansion, mm
Hardness Figure 7- The grand averages for the

lateral expansion and absorbed energy
As shown in Figure 8, the absorbed for each specimen group are plotted
energy decreases with increasing hardness. here (averages of the 4 laboratories for
The relationship is approximately linear, if each group). Results for the 2 and 8 mm
only the type 4340 steel specimens are tests are shown.
considered. The XM32 steel fits the trend,
but the T200 maraging alloy does not. This result is not surprising, but the plot does
provide a good example of the relationship between hardness and absorbed energy for the
impact verification specimens. The variation in hardness measured for the specimen
groups was generally low. Reviewing this data it appears that a standard deviation in
hardness of between 0.1 and 0.2 HRC is routinely attained for samples at all energy levels.
This corresponds to a range in the hardness of 1 HRC for a specimen group (25
specimens). The specimens having the most
consistent hardness, however, did not
necessarily have the lowest variation in
absorbed energy.
There was a consistent offset in the
hardness values reported by the various
laboratories. Since hardness measurements
were not the focus of this study (not
mandatory), this result is not of direct concern
here. However, it is of general interest that
the average hardnesses for the specimen
groups measured by the four laboratories
were almost never within 1 HRC, and
sometimes differed by more than 2 HRC.
Discussion
Figure 8 - Hardness and absorbed
The specimens used for verifying Charpy energy of the steels used for vertification
testing.
impact machines around the world have similar variation in absorbed energy, which is the
principal factor used to evaluate the quality of these specimens. The coefficient of
variation for the specimens from all laboratories and energy levels typically ranged from
0.02 to 0.04. This result is interpreted as a benchmark for the quality of impact
verification specimens for the 1990's. Machine/specimen interactions (and other variables)
affect the variation assigned to a particular specimen group by a single machine, but in
general the differences in specimen variation were resolvable and appear to reflect the
homogeneity in the impact properties, rather than effects of machines, operators, and
sampling.
The relative outcome of the absorbed energy results for the laboratories was
sometimes dependent on the specimen group that was tested. This was particularly true
for the tests made using the lowest energy specimens (group 1), where the changes in
interlaboratory results due to the specimens altered the general interpretation of the
results. For example, considering only the tests made with specimen groups 1 and 2, one
might conclude that the results from laboratory 4 were different and lower than those of
the other laboratories, but considering only the results for specimen group 3, the results
from laboratory 2 appear to be different and higher than those of the other laboratories. If
only the results for the group 4 specimen were considered, we would conclude that the
laboratories all produce similar results for low-energy impact tests. These specimen
effects are secondary in many cases, but raise questions concerning what these apparent
differences in the results show us about the performance of the impact machines tested.
Because these specimen effects are not well understood for the most part and are not
clarified by these results, we refrain from speculating here. The data do; however, provide
a useful side-by-side comparison of these differences for our verification specimens a n d
will hopefully lead to discussions and testing that improves our understanding of
specimen/machine interactions.
The summary plot of the 2 and 8
mm tests results in Figure 9 shows that the
effect of striker radius is small (statistically
insignificant) for the specimens made from
4340 steel. The small differences in
energy for the 2 strikers are not consistent
over the 10 to 150 J range: they show the
tests results for the 2 mm striker to be
slightly higher for the specimens at energy
levels 1 and 2 (average difference of +0.16
and +1.33 J respectively), but lower at
energy level 3 (average difference of-2.75
J). The lateral expansion results indicate a Figure 9 - The 2 and 8 mm striker results
more consistent trend of increased ( grand averages) are shown. Thegrand
ductility (higher toughness) for the 2 mm average of the 2 and 8 mm strikers include
tests over the range of 10 to 150 J for the about 60 and 40 tests respectively. The
4340 specimens. This finding is error bars represent • standard deviation
consistent with several previous studies for the 8 mm averages.
that evaluated these effects for 4340 steels
[7-11].
For certifications, even small differences in the test results must be considered. So,
since our results do not show a predictable relationship for the striker radius effect, we
conclude that separate certifications for 2 and 8 mm tests are needed for verification
specimens made from 4340 steel.
The results for the XM32 and T200 steels were strongly effected by the striker
radius used for the test. For the very ductile T200 material, this greater sensitivity has
been explained by the fact that the larger radius 8 mm striker (and striker design) requires
more bending of the specimen during fracture and this requires more energy [7,8,10].
These same mechanical differences may also explain the better performance of the T200
specimens with the 8 mm striker: the increased bending results in more complete tearing
during fracture, which results in a more consistent fracture energy for the specimens. The
higher energy for the 2 mm tests on the XM32 material, however, has not yet been
convincingly explained. Studies with other materials, showing similar effects of striker
radius, have speculated that this increased energy for 2 mm tests may be due to deeper
penetration of the smaller striker combined with mechanical properties specific to the
material (strain hardening, strength, hardness, etc.) [10]. Here, the lateral expansion for
the 2 mm tests on the XM32 samples was lower than those for the 8 mm tests (and similar
to the 2 and 8 mm results for the 4340 steel), so considering the ductility associated with
the fracture does not provide an explanation.
Overall, the comparisons of our results are reassuring. Even though the different
specimen groups clearly had an effect on some test results, the trends for the test results
were generally independent of the specimens tested. The average absorbed energies
determined by the four laboratories for the various specimen groups, and the variation in
absorbed energies, are in good agreement. The laboratories differed mostly on results for
verification specimens of the lowest and highest energy (energy levels land 3). But the
results never differed from the average by more than 2 J (or 7.5%). In most cases the
ASTM 1236 requirement for qualifying reference machines of 1.4 J (or 5%) was met, and
if the verification systems could have been more directly comparable (by including C-type
machine results in the averages for laboratories 2 and 3), the range in absorbed energies
between the verification laboratories would likely decrease.
Our laboratories have long recognized that differences in the designs of impact
machines have an effect on impact test results, and our verification systems include this
factor (with the exception of Japan) when determining a certified energy for impact
verification specimens. Japan uses a single C-type machine for certifying verification
specimens, but the vast majority of the companies verifying impact machines in Japan also
have a C-type machine (so Japan does not necessarily have to consider this variable). In
Europe and the United States, however, many different types of impact machines are used
by our industries, so we try to include the differences due to machine design in our
certification procedures. As explained in the background section, the certified energy for
ASTM verification specimens is determined using the average energies from 3 different
impact machines (2 U-types from different manufactures and 1 C-type). The French
system uses one C-type and one U-type machine to determine certified values for
verification specimens. The verification system used by the European Commission (BCR)
uses the results from round-robin testing on master specimens (which includes many
different impact machine designs) to adjust the result of reference machines used to assign
certified values to verification specimens. This practical approach of averaging the
machine design effects has worked reasonably well, but the limits specified for verification
tests are quite stringent in some cases, and the machine design effects can be significant,
particularly when testing low-energy specimens (as shown by our results). However,
considering the ISO 148-2 requirements for verifying the performance of impact machines
(which allow a 4-4 J error for test at energies of 40 J or less, and a 10% error for tests at
energies over 40 J), the agreement found for our systems appears to be adequate for
purposes of international commerce. As impact machines of newer design and larger
capacity are produced we will have to reevaluate this approach (and our respective limits
on verification of test results).
As a f'mal point for comparison, we consider how well the certified energies for the
specimen groups tested in this study agree with the average energies determined by our
laboratories. As shown in Figure 10, the agreement is good. Again, it was difficult to
make a true comparison here, because the certified values for the Japanese specimens were
not available. However, the results for the Japanese impact machine in this study were
substituted for their certified values, and these values are expected to closely estimate the
certified values. For the 17 cases compared here, the average energy determined by the
laboratories in 12 cases differed from the certified value by less than 1.0%, 4 differed by
1.0 to 2.2 %, and 1 differed by 3.2%.
This is a comforting result, which 300
indicates that the certified values being
determined by the laboratories O
independently, are virtually identical to the 200

grand averages determined by the four
laboratories in this study. So, we believe
that industries using our specimens to
100
verify the performance of Charpy impact
Ul
machines to the requirements of ISO 148- |
2 (which is the machine verification

I I
document we all have in common) can be <
~ o
100 200 300
assessed in a fair and meaningful manner. Certified Reference Value, J
Figure 10- Average energy of the 4

Conclusions laboratories versus the certified energy of
the laboratory that provided the specimens.
1) The certified energies assigned to
verification specimens by the laboratories
which supplied them for this study agree closely with the average energies determined by
the four participating laboratories.
2) The impact verification specimens distributed by all of the laboratories are similar in
quality, based on the similar variations in energy found for the specimens tested in this
study. The variation in the energy of impact verification specimens is currently controlled
within a CV range of 0.02 to 0.04.
3) The specimens used to verify the performance of a Charpy impact machine can affect
the apparent performance of the machine.
86 PENDULUMIMPACT TESTING: A CENTURY OF PROGRESS
4) The absorbed energies measured with 2 and 8 mm strikers are very similar for 4340
steel, but can differ significantly for other materials used to produce verification
specimens.
Acknowledgments
C. N. McCowan would like to thank D. P. Vigliotti and T. A. Siewert from the NIST
Materials Reliability Division for their efforts in measuring and testing the specimens, and
in the preparation of this manuscript.
J. Pauwels wants to thank Mr. Gyppaz, responsible of the Quality Department of
Cogne Acciai Speeiali srl, Aosta (IT) for carrying out the IRMM/BCR measurements and
Dr. C. Ingelbrecht, responsible for physical reference materials within the IRMM
Reference Materials unit for his valuable assistance in the evaluation of the results.
References
[1] Pauwels, J., Gyppaz, D., Varma, R., and Ingelbrecht, C., European Certification
of Charpy Specimens: Reasoning and Observations, Pendulum Impact Testing: A Century
of Progress, ASTM STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American
Society for Testing Materials, West Conshohocken, PA, 1999.
[2] EN 10045-2; Metallic Materials - Charpy Impact Test - Part 2: Verification of
Pendulum Impact Testing Machines.
[3] ISO 148-3, Metallic Materials - Charpy Pendulum Impact Test - Part 3:
Preparation and Characterization of Charpy V Reference Test Pieces for Verification of
Test Machines, International Organization for Standards, Geneva, Switzerland.
[4] ISO 148-2, Metallic Materials - Charpy Pendulum Impact Test - Part 2:
Verification of Test Machines, Intemational Organization for Standards, Geneva,
Switzerland.
[5] JIS B 7740, Charpy Pendulum Impact Test - Reference Test Pieces for
Machines, Japanese Institute of Standards, Japan.
[6] JIS B 7722, Charpy Pendulum Impact Test - Verification of testing machines,
Japanese Institute of Standards, Japan.
[7] Ruth, E. A., Striker Geometry and its Effect on Absorbed Energy, Pendulum
Impact Machines: Procedures and Specimens for Verification, ASTM STP 1248, T. A.
Siewert and A. K. Schmider, Eds., American Society for Testing Materials, Philadelphia,
1995, p 101-110.
[8] Nanstad, R. K. and Sokolov, M. A., Charpy Impact Test Results on Five
Materials and NIST Verification Specimens Using Instrumented 2 and 8 mm Strikers,
Pendulum Impact Machines: Procedures and Specimens for Verification, ASTM STP
1248, T. A. Siewert and A. K. Schmider, Eds., American Society for Testing Materials,
Philadelphia, 1995, p 111-139.
[9] 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, J. M.
Holt, Eds., American Society for Testing Materials, Philadelphia, 1995, pp. 35- 53.
[10] Siewert, T. A. and Vigliotti, D. P., The effect of Charpy V-Notch Striker Radii
on the Absorbed Energy, Pendulum Impact Machines: Procedures and Specimens for
Verification, ASTM STP 1248, T. A. Siewert and A. K. Schmider, Eds., American
Society for Testing Materials, Philadelphia, 1995, pp. 140-152.
[11] Yanaka, M., et. al., Effects of the Striking Edge Radius and Asymmetrical
Strikes on Charpy Impact Test Results, Pendulum Impact Machines: Procedures and
Specimens for Verification, ASTM STP 1248, T. A. Siewert and A. K. Schmider, Eds.,
American Society for Testing Materials, Philadelphia, 1995, pp. 153-1701
Appendix
Data Summary
EMEA~ ESTD ~MEAN .ESTI tMEA~
LWv ',L SPECIME/~ LAB TUP
(J) (J) (mm) (HRC) (HRC)
1 1 1 2 26.60 0.56 0.02 46.32 0.18
1 1 2 2 27.71 1.03 0.32 0.04 45.92 0.20
1 1 3 2 27.88 1.13 0.22 0.02 45.69 0.14
1 1 4 2 25.35 1.14 0.27 0.04 45.78 0.10
1 1 1 8 26.66 0.55 0.12 0.03 46.35 0.13
1 1 2 8 27.62 0.63 0.22 0.02 45.00 0.00
1 1 3 8 28.31 1.12 0.19 0.03 45.69 0.10
1 1 4 8 24.99 0.34 0.22 0.02 45.90 0.11
1 2 1 2 25.95 0.61 0.16 0.01 44.03 0.09
1 2 2 2 26.65 0.90i 0.26 0.02 42.82 0.39
1 2 3 2 26.51 0.58 0.22 0.03 43.64 0.18
1 2 4 2 24.89 1.13 0.20 0.03 43.58 0.15
1 2 1 8 25.79 0.71 0.12 0.02 44.20 0.20
1 2 2 8 26.55 0.94 0.22 0.02 42.15 0.34
1 2 3 8 26.58 0.72l 0.20 0.02 43.68 0.23
1 2 4 8 24.58 1.37 43.48 0.17
1 3 1 2 16.52 0.25 0.04 0.01 47.75 0.21
1 3 2 2 18.75 1.26 0.19 0.05 46.33 0.99
1 3 3 2 16.83 0.72 0.08 0.02 46.57 0.36
1 3 4 2 15.83 0.38 0.11 0.01 47.10 0.34
1 3 1 8 15.87 0.78 0.02 0.01 47.70 0.20
1 3 2 8 17.81 0.50 0.10 0.02 46.60 1.70
1 3 3 8 16.85 0.58 0.07 0.02 46.54 0.25
1 3 4 8 15.64 0.33 0.09 0.01 47.09 0.33
1 4 1 2 28.25 0.67 0.11 0.01 43.90 0.16
1 4 2 2 29.43 0.64 0.21 0.03 43.00 0.00
1 4 3 2 29.19 0.68 0.13 0.01 42.91 0.10
1 4 4 2 28.11 0.74 0.18 0.02 43.29 0.12
1 4 1 8 28.41 0.66 0.09 0.01 43.85 0.17
1 4 2 8 29.08 0.98 0.18 0.02 41.89 0.22
1 4 3 8 29.00 1.03 0.11 0.02 42.92 0.10
~~[LESTD [HMEAN
VEL SPECIMEN LA] TUP !MEAN
(J) (J) I (mm) I(mm) I (HRC) (HRC)
1 4 4 8 28.25 0.37 0.17 0.02 43.22 0.15
2 1 1 2 112.13 4.59 1.31 0.05 34.88 0.16
2 1 2 2 113.88 4.00 1.47 0,04 33.88 0.30
2 1 3 2 114.92 5.69 1.46 0.06 34.08 0.21
2 1 4 2 111.23 3.50 1.41 0.05 34.03 0.18
2 1 1 8 112.66 4.66 1.28 0.06 34.92 0.18
2 1 2 8 111.12 2.86 1.39 0.04 32.65 0.67
2 1 3 8 112.84 2.84 1.41 0,05 33.98 0.17
2 1 4 8 109.19 3.41 1.41 0.05 33.89 0.25
2 2 1 2 66.84 1.51 0.73 0,04 38.97 0.09
2 2 2 2 70.23 1.68 0.90 0,09 37.63 0.35
2 2 3 2 68.49 1.08 0.74 0.06 38.42 0.17
2 2 4 2 67.96 1.00 0.78 0,04 38.15 0.11
2 2 1 8 65.91 1.21 0.65 0.03 38.98 0.14
2 2 2 8 66.92 0.96 0.73 0.04 37.10 0.21
2 2 3 8 67.89 1.02 0.73 0.04 38.45 0.10
2 2 4 8 67,32 1.36 38.03 0.15
2 3 1 2 96,28 1.78 1.10 0.04 36.17 0.24
2 3 2 2 99,31 2.50 1.26 0.05 33.97 0.61
2 3 3 2 98.87 2.64 1.21 0.04 35.11 0.20
2 3 4 2 99.15 2.05 1.25 0,04 34.92 0.30
2 3 1 8 96,11 2.14 1.11 0.04 36.03 0.34
2 3 2 8 96.72 2.20 1.19 0.06 33.65 0.53
2 3 3 8 97.74 2.65 1.14 0.04 34.95 0.37
2 3 4 8 98.47 2.21 1.22 0.03 35.03 0.18
2 4 1 2 104.39 2.50 1.07 0.04 34.98 0.06
2 4 2 2 107.72 2.04 1.20 0.04 34.00 0.33
2 4 3 2 104.94 3.11 1.15 0.04 33.88 0.15
2 4 4 2 106.34 2.50 1.15 0.05 34.24 0.14
2 4 1 ' 8 103.10 2.32 1.04 0.03 34.95 0.16
2 4 2 8 106.59 3.02 1.16 0.05 31.95 0.16
2 4 3 8 103.09 3.25 1.08 0.05 33.95 0.09
2 4 4 ' 8 104.77 3.22 1.15 0.04 34.29 0A3
3 1 1 2 153.61 3.74 1.79 0.11 29.47 0.09
3 1 2 2 161.46 3.92 1.94 0.12 28.67 0.36
3 1 3 12 157.56 5.59 1.92 0.08 28.29 0.23
3 1 4 2 158.27 6.19 1.83 0.06 28.18 0.18
3 1 1 88 147.74 3.97 1.72 0.08 29.10 0.17
3 1 2 149.98 4.38 1.85
1.82
0.09
0.07
26.85
40.97
0.34
3 1 3 8 146.08 4.26
3 1 4 18 146.92 6.08 1.77 0.26 28.43 0.18
3 2 1 2 113.93 3.81 1.41 0.06 30.78 0.16
3 2 2 2 114.98 4.99 1.55 0.05 28.03 1.94
LEVEL SPECIMEN rAB TIS

EMEAN ESgt LEMEAN LESTD HMEAI~ H S T D - ]
(J)
3 2 3 2 115.82 0.26
3 2 4 2 116.96 4.45 1.53 0.05 29.82 0.15
3 2 1 8 115.58 4.52 1.37 0.05 30.85 0.17
3 2 2 8 120.16 5.01 1.54 0.09 27.45 0.86
3 2 3 8 115.07 3.53 1.44 0.04 30.13 0.40
3 2 4 8 113.55 2.25 *
29.88 0.16
3 3 1 2 213.29 24.61 2.06 0.09 30.33 0.18
3 3 2 2 208.17 18.80 29.00 0.00
3 3 3 2 198.04 6.82 2.17 0.07 29.25 0.30
3 3 4 2 223.34 25.00 2.21 0.16 29.12 0.18
3 3 1 8 255.46 7.56 2.25 0.12 30.30 0.20
3 3 2 8 259.91 13.14 29.11 0.22
3 3 3 8 254.99 7.13 2.27 0.04 29.28 0.16
3 3 4 8 257.93 7.53 2.47 0.04 29.33 0.16
3 4 1 2 162.60 4.15 1.81 0.07 24.55 0.33
3 4 2 2 175.47 4.81 1.94 0.04 22.10 0.47
3 4 3 2 157.88 4.35 1.84 0.04 23.27 0.14
3 4 4 2 167.61 6.69 1.97 0.17 23.30 0.17
3 4 1 8 170.24 9.23 1.79 0.06 24.70 0.26
i
3 4 2 8 178.40i 8.80 1.94 0.06 22.03 0.61 i
3 4 3 8 162.31 6.37 1.79 0.03 23.22 0.15 1
3 4 4 8 17E69 8.20 1.89 0.04 23128 0.16 ]
Not fractured completely
J. Pauwels, l D. Gyppaz,2 R. Varma, 3 and C. Ingelbrecht I
European Certification of Charpy Specimens: Reasoning and Observations
Reference: Pauwels, J., Gyppaz, D., Varma, R., and Ingelbrecht, C , "European
Certification of Charpy Specimens: Reasoning and Observations," Pendulum
Impact Testing: A Century of Progress, STP 1380, T. A. Siewert and M. P.
Manahan, Sr., Eds., American Society for Testing and Materials, West
Abstract: The procedure for the production and certification of BCR (Bureau
Communautaire de Reference) Charpy specimens for machine verification according
to the European standard EN 10045-2 is described. A master batch is initially certified
by interlaboratory comparison and is then used as a reference to certify sub-batches of
similar impact toughness by comparative measurements on one dedicated machine.
Extensive testing over ten years has shown that, for SAE 4340 steel, a 2 mm striker
generally produces higher impact toughness than an 8 mm striker up to 120 J absorbed
energy. Another observation is that asymmetric fracture generally occurs in about 50%
of all cases, but that small series of tests may produce nearly all symmetric or nearly
all asymmetric failures, influencing the mean value of the measured impact toughness.
Keywords: Charpy V-notch, standardisation, asymmetric fracture, European

certification.
Introduction
The European Commission supports intemational standardisation by providing

BCR certified reference materials (CRMs), through the Measurements and Testing
Programme. The Institute for Reference Materials and Measurements (IRMM) of the
Joint Research Centre has the responsibility of maintaining and distributing the BCR
reference materials including V-notch Charpy specimens in the range 30 to 160 J
intended for the verification of testing machines according to the European standard
EN 10045-2:1992 (Metallic materials - Charpy impact test - Part 2: Verification of
the testing machine, pendulum impact). The fundamental feature of the European
1Reference Materials Unit, EC-JRC-IRMM, Retieseweg, B-2440 Geel, Belgium.

2Quality Department, Cogne Acciai Speciali, Via Paravera, 16, 11100 Aosta, Italy.
3National Physical Laboratory (NPL), Queens Road, Teddington, Middlesex,
TW11 0LW, United Kingdom.
90
PAUWELS ET AL. ON EUROPEAN CERTIFICATION OF CHARPY SPECIMENS 91
philosophy of certification of Charpys (and the main difference from the ASTM
system) is their traceability to an original 'master batch' of specimens at each energy
level, rather than to standard machines maintained by any one metrological authority.
The first certifications [1] carried out in the late eighties for batches of 30, 60, 80
and 120 J nominal energy concerned impact tests done in accordance with the ASTM
E#23-88 (Standard Test Methods for Notched Bar Impact Testing of Metallic
Materials) and ISO specifications (ISO 148-1983, Charpy impact test (V-notch), now
ISO 148-3:1998). However, today the EN standard is of general application in Europe,
and present certifications [2] are made according to EN 10045-2.
Certification Methodology
BCR CRMs are certified through collaborative studies involving, in general, 8 to

15 expert laboratories from various countries of the European Union. The
certifications are carried out in conformity with the BCR Guidelines for the
certification of reference materials [3], which are to a large extent based on the
relevant ISO-REMCO Guides [4]. However, as the size of heat treatment ovens
generally available to produce homogeneous batches of Charpy specimens is rather
limited and the demand for Charpy CRM samples quite large, it was considered to be
neither practical nor economical to certify each individual batch by interlaboratory
comparison. Therefore, the method which was adopted for the BCR Charpy CRMs
was to certify one batch per energy level by interlaboratory comparison, as a master
batch, and to certify subsequent CRM batches for general use by direct comparison
with a master batch of the same nominal energy (see Fig. 1).
Fabrication of the Charpy Specimens
Samples of 30 to 120 J are produced from SAE 4340 steel, and 160 J samples are
produced from ASTM 565 steel. Starting materials are homogeneous batches of
several tons of steel, forged into 14 mm square bars. The base materials were verified
for chemical and impact toughness homogeneity. At various places along the length,
specimens were machined, heat treated to produce a specific impact toughness at 20~
and measured. The acceptance criterion was that the relative standard deviation
determined on 25 to 30 samples is < 3%.
Charpy specimens are produced in individual batches of 750 for SAE 4340 or 450
for ASTM 565 steel. All samples from one batch are cut from the same section of the
billet in continuous succession along the length the bar. Before a batch of 750 (or 450)
specimens is heat treated, a test is done with a complete load of specimens of which
725 (or 425) are dummies and 25 are real specimens. After complete heat treatment,
the 25 test samples undergo the final machining and are broken. This is to verify that
the heat treatment conditions produce the desired energy level and to verify that all
working conditions are appropriate to produce a homogeneous batch. After this
preliminary trial, the real batch is heat treated as soon as possible, after which the
specimens are milled to final size and the notches are ground. Special attention is paid
to the surface finish of the notches in accordance with the test recommendations.
Figure 1 - Principle of the certification of BCR Charpy CRMs
Homogeneity Testing
The homogeneity of each batch is verified after the preparation is completed on 30

specimens taken at random and broken on the same machine. The observed relative
standard deviation are in general well below the acceptance criterion for homogeneity
fixed at + 3%.
Stability Testing
It was not expected that impact toughness would change with time, but in line with
normal procedure for CRM certifications, stability testing was carried out as part of
the 1990 certifications of the first BCR Charpy CRMs [1]. Two hundred samples
taken from one single 120 J batch of verified homogeneity were exposed for various
time spans to temperatures in the range 4~ to 90~ and broken on a pendulum of 300
J capacity. The experimental values (means of 5 measurements) are summarised in
Table 1. They demonstrate that there i s n o sign of any change of the measured energy
with time or temperature.
Table 1 - Stability tests carried out on 120 J V-notch specimens [I]. Values in Joules
Time of Temperature
storage
Months 4~ 25 ~ 60 ~ 75 ~ 90 ~
2 122.5 123.7 123.2 123.8 123.1

4 123.1 124.5
8 124.1 122.9
12 123.3 123.8 123.0 123.9
18 123.5 123.8 123.0 122.4
Certification o f a Master Batch
The certification of a master batch begins with a preliminary feasibility study,

consisting of an interlaboratory comparison involving usually 8 to 15 laboratories of
proven expertise from industrial and metrological organisations. The results of this
intercomparison are discussed with the laboratory managers and possible reasons for
error are identified. A batch of proven, good homogeneity is subsequently chosen as
candidate master batch and measured by a set of 6 to 12 laboratories selected on the
basis of their capabilities demonstrated in the feasibility study.
Following normal BCR procedure, the unweighted mean of the means of the p
accepted sets of results produced by the participating laboratories is taken as the
certified value (M) of the master batch, and the 95% confidence interval of these
means is taken as its uncertainty.
This confidence interval, given by:
S
+ te-1 a/-fi (1)
p = number of accepted sets of results

s = standard deviation of the distribution of the mean values
t~, = student factor forp-1 degrees of freedom
includes effects due to the between-laboratory differences, repeatability and

inhomogeneity. The repeatability and inhomogeneity parts of this uncertainty are,
however, of no practical use as the above uncertainty is based on large numbers of
measurements (generally 60 to 120) while, in practice, the users will break only five
specimens of the CRM. The uncertainty component UB due to between-laboratory
differences is important, however, as it is used subsequently in the certification of sub-
batches. It can be derived from the experimental results as:
SB
U B = t,,_l ~ (2)
where s. = the between-laboratory standard deviation.

It is understood that this uncertainty component UB will always be used in
conjunction with an inhomogeneity component calculated for the appropriate number
of specimens.
Certification of CRM Batches by Comparison with its Master Batch
Actual CRM Charpy specimens for sale are prepared from sub-batches of an
impact strength comparable to that of the master batch. They are certified by
comparison with the original master batch in a single laboratory using one machine
dedicated to BCR certification work.
The measurement therefore consists of determining the (small) difference between
the mean value of the batch to be certified (C) and that of the master batch (M).
A= C -M (3)
To obtain a value of A with sufficient confidence the number of samples of the

batch to be certified is fixed at 30 and the number of master batch samples is at least
25. Several sub-batches may be certified at one time.
The statistical tool used to certify all sub-batches is the statistical tolerance interval,
developed for assigning uncertainties when heterogeneity is not negligible. Its
meaning is that for a certain proportion of the complete batch (e.g. 95%), all the
individual sample values are contained in an interval with a probability of e.g. 95%. In
the case of Charpy testing, however, the common practice is to use the mean value of
five successive measurements. A method was therefore developed [5] to determine
statistical tolerance intervals for means of observations. The concept behind this work
is that the mean impact toughness value of five randomly selected specimens has a
larger probability than that of a single specimen to be close to the mean value of the
batch.
The certified uncertainty U o f B C R Charpy CRMs for the mean value of any set of
five samples of the particular batch is composed of:
9 the uncertainty U8 of the mean value of the master batch
9 the uncertainty component due to the inhomogeneity (UM)
9 the uncertainty ofA (U~).
The uncertainty on A, i.e. the confidence interval for the difference of the mean
values of the master batch and the sub-batch is given by:
/s~(n'-l)+s-2Z(---n2-1) ( 1 1/ (4)
-+' ..... '4 " 7,
where s and n are, respectively, the standard deviation and number of samples in the
two series (master batch or sub-batch) and the subscripts refer to the two series, and
tl-o~2 corresponds to ni + n2 -2 degrees of freedom.
The global uncertainty for a set of five certified samples is then:
v = +vl (5)
Observations Related to BCR Certification
One of the major advantages of the BCR certification philosophy is the "European
dimension" given by the participation of expert laboratories of several European
countries. The results of the impact toughness test are strongly dependent on a large
number of factors related either to the machine or to the testing procedure. The
complexity of some of the metrological controls that would be necessary to ensure
whether a machine is within tolerances explains, however, the usefulness of the
reference materials, which are the only quick and reliable means of making it possible
for the user to verify if his machine is in good order. Therefore the invariance given by
the traceability to the original master batch is certainly an advantage, even if a small
extra uncertainty (UA) has to be taken into account, and the procedure allows the close
monitoring over many years of the performance of the single laboratory and its
machine carrying out the CRM certification by the values determined for master batch
specimens.
Absorbed Energy Differences Between Symmetric and Asymmetric Fractures
When specimen breakage occurs such that the shear lips from both edges of the
fracture are found on the same broken half of the specimen the failure is said to be
"symmetric". When the shear lips are on different halves of the specimen the failure is
"asymmetric". Extensive data collection during large numbers of CRM certification
campaigns and parallel studies showed that a variable and often quite significant
proportion of the samples undergo asymmetric fracture with, at certain energy levels, a
statistically significant difference in measured energy value. Table 2 shows that, in the
range 30 to 120 J (SAE 4340 steel), the experimental value of
(A - S) goes from slightly negative at the low energy side to significantly positive at
the high energy side. The 160 J specimens (ASTM 565 steel) however show again a
lower value, indicating that the type of steel is significant. It should moreover be noted
that, despite the fact that all figures are based on series of at least 15 strikes,
sometimes very different values were observed. The same applies to the proportion of
asymmetric ruptures (see Fig. 2), which is generally close to 50%, but may in some
cases be really extreme without any obvious reason. For 120 J (where the effect is
largest and a large number of series is available) a valid comparison between results
obtained using the 2 mm and the 8 mm striker could be made, but no significant
difference was observed.
20,
9 30J
16 -- ~ !
[3 60J
~"~12
------8 -- .80Ji
120J
" @ 160J
J!
4~
o ,. .
I~~lJl~ljl
, . . ~ .A . . . .. . . BB, :
I
5 15 25 35 45 55 65 75 85 95
% of asymmetric rupture
Figure 2 - Frequency of asymmetric fractures observed on series of as least 15 tests
T a b l e 2 - Experimental difference in measured impact toughness between asymmetric

and symmetric fracture
( A - S ) in Joules
Nominal Energy Average Stand. dev. No. o f m e a s u r e m e n t

series
S A E 4340 steel
30 J - 0.2 0.6 32
60 J 0.0 0.4 8
80 J + 1.4 0.5 8
120 J * + 4.4 1.1 48
9 2 m m striker + 4.3 1.3 28
9 8 m m striker + 4.4 0.8 20
A S T M 565 steel
160 Joule +2.6 1.5 8
This effect may influence the A -value determined according to equation (3),
especially at 120 J, and possibly lead to an underestimation of Ua. On the other hand,
it increases the apparent heterogeneity of the CRM batch. Finally, one has to be aware
that when carrying out a 5-specimen test, as is normally done by users of the CRM, it
may not be excluded the fractures may all be symmetric or all asymmetric (Fig. 3), in
which case this effect may cause measurement errors of the same magnitude as the
certified uncertainty.
100
80
>,
60
1130J
B 120J
40
IJ.
20
0 L
0 1 2 3 4
No of asymmetric ruptures per 5 tests
Figure 3 - Frequency of asymmetric fractures observed on series of 5 tests
Absorbed Energy Differences Between 2 mm and 8 mm Striker
The present BCR certification is limited to EN 10045-2, and thus does not cover
the ASTM E#23-96. Originally, BCR certification covered both measurements using
the 2 mm (EN certification) and the 8 mm (ASTM certification) striker, with
detectable, but often quite insignificant, differences in measurement values, but
ASTM values were withdrawn from the certificates as "the reference values have not
been established on the three reference machines owned maintained and operated by
NIST in Boulder, CO" as is required by ASTM E#23-96. There is presently, therefore,
a lack of worldwide international comparability.
In the original BCR master batch certification studies [I], it was initially found, on
the basis of the interlaboratory comparisons carried out using 2 mm and 8 mm strikers,
that the 2 mm striker systematically produced higher results in the range 30 to 120 J
(see Table 3, 2nd column). However, the uncertainty 6n the observed differences was
so high that doubts existed about the reliability of these estimates. Moreover, neither
the machines nor the laboratories participating in the two certification exercises were
identical. Therefore, three laboratories were selected to carry out fiLrther tests on
batches of both 80 and 120 J nominal energy (see Table 3, 3rd column for results), and
this exercise again indicated a positive difference for the 80 J samples, but gave a
value close to zero for the 120 J samples. In 1991, further laboratory intercomparisons
were carried out [6], which led to the determination of the differences given in the 4th
column of Table 3. These differences, although still quite imprecise, confirmed the
belief that, at energy levels at least up to 100 J, the radius of the striking edge would
have a small effect on the impact toughness (the 2 mm striker giving higher values)
and the difference would increase slightly with energy.
Table 3 - Experimental differences in measured impact toughness using 2 mm (M2mm)

and 8 mm (Msm~)strikers (uncertainties are estimates of 95% confidence intervals)
( M2mm - M8mm ) in Joules
Nominal Original BCR Study by 3 labs BCR results Systematic

Energy observation study at 120 J
1990[1] 1990[1] 1991 [6] 1997
30 Joule 0.2 4- 2.2 -- 0.2 4- 0.7 --
60 Joule 0.9 + 2.4 -- 0.5 + 0.9 --
80 Joule 2.6 + 3.3 1.6 :k 1.6 0.7 + 1.2 --
120 Joule 1.0 + 7.4 - 0.8 + 5.4 - 3.3 + 2.1 0.7 + 0.2
Then, in 1997, a large series of measurements (600 samples) was carried out using
the BCR reference machine, which was characterised by an unusually high
repeatability (see Table 3, column 5, and Table 4). On this occasion, the difference
between the 2 and the 8 mm striker was found to be systematically positive.
Considering the above information, together with ' data from the international
intercomparison organised by NIST (this symposium), it appears that the difference in
measured energy between 2 mm and 8 mm strikers is approximately proportional to
energy level for SAE 4340 steel. Two limited series of 160 J BCR samples (ASTM
565 steel) were also tested in the NIST intercomparison and gave larger positive
differences (+6J and +3J), so different steels probably have a different M2m,n - Msmm
dependence, possibly also linear. It should be noted, however, that the differences in
energies measured using the different strikers are not large compared with the
precision required by the existing standards.
It is, therefore, our present conviction that EN 10045-2 and ASTM E#23-96 testing
procedures produce, at least up to 160 J, small but detectable differences in measured
energy values. The EN values are found to be systematically higher with the difference
increasing with energy. Worldwide international trade would, however, benefit from
future harmonisation between standards. Therefore, further investigations of
experimental differences between the EN and ASTM test methods, including a study
of different steel types, may be of importance.
Table 4 - Precision measurements of differences in impact toughness (J) between

2 mm (m2mnO and 8 mm (Mgmm)strikers at 120 Jnominal energy (95% confidence
intervals on individual series of 30 measurements are of the order of l J)
Nominal energy, M2mm Msmm Difference

batch No and date
120U 5/97 118.69 118.18 0.52

120 U 6/97 118.42 117.88 0.54
120 V 5/97 118.36 117.60 0.76
120 V 6/97 118.07 116.97 1.10
120 Z 5/97 112.85 112.28 0.57
120 Z 6/97 112.92 111.98 0.94
120 AA 5/97 113.30 112.78 0.52
120 AA 6/97 113.20 112.70 0.50
120 AB 5/97 114.03 113.50 0.53
120 AB 6/97 114.23 112.89 1.34
MEAN DIFF. 0.73 + 0.21
Acknowledgment
The authors wish to acknowledge the contribution of H. Marchandise to the

development of the BCR Charpy programme and the analysis of recent data.
References
[1] Marchandise, H., Perez-Sainz, A., Colinet, E., "Certification of the of V-notch
Charpy Specimens", EUR 12598 EN (1990).
[2] Varma, R.K., "The certification of two new master batches of V-notch Charpy
impact toughness specimens in accordance with EN 10045-2:1992", EUR
report to be published.
[3] "Guidelines for the Production and Certification of BCR Reference Materials",
BCR/48/93 (1994).
[4] "The role of reference materials in achieving quality in analytical chemistry",

REMCO (ISO Committee on reference materials), ISO (1997).
[5] Steyaert, H., "Statistical tolerance intervals", EUR-11639 EN (1988).
[6] Marchandise, H., Perez-Sainz, A., Colinet, E., "Certification of the of V-notch
Charpy specimens", BCR Internal document XII/323/91-EN.
Maria Sundqvist I and Guocai Chai 2
Stability of a C-type Impact Machine Between Calibrations
Reference: Sundqvist, M. and Chai, G., "Stability of a C-type Impact

Machine Between Calibrations," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American
Society for Testing and Materials, West Conshohocken, PA, 2000.
Abstract: In order to investigate the stability of an impact testing machine

between ordinary calibrations, impact tests have been performed on a regular
basis using specimens from an isolated batch of material. A nickel based alloy,
W-Nr. 1.4563, was chosen due to its ageing stability. The material was heat
treated and a microstructure analysis was carried out ensuring a homogenous
structure in all specimens. The specimens were then cut out in the longitudinal
direction.
The machine is a 300 J C-type pendulum and usually approximately 30 specimens
are tested every day. Commonly the materials are very tough, generating energy
values above 80% of the machine capacity.
For the stability analysis, a set of control specimens was tested on the first day of
every week and the results were recorded. Notched type A specimens (Charpy V)
were used and tests were performed with an ISO striker. On the other days ordinary
specimens were tested. The total number of tested specimens and especially the
number of specimens generating energy values above 240 J were documented.
The deviation of energy values obtained by the control specimens between the
calibrations was evaluated. By using a stability diagram the correlation between
energy divergence, time and number of dally tested specimens was studied.
The results could be used as complementary guidelines, apart from the regular
radius measurement of the anvils, for determining optimal intervals between
calibrations for similar machines used in the high energy area.
Keywords: impact testing, calibration, stability
1 Manager, Mechanical Metallurgy, AB Sandvik Steel, 811 81 Sandviken, Sweden.

2 Research scientist, Mechanical Metallurgy, AB Sandvik Steel, 811 81 Sandviken,
Sweden.
lOO
SUNDQVIST AND CHAI ON C-TYPE IMPACT MACHINE 101
Introduction
Since Russel [1] reported a new machine with a swinging pendulum for impact
testing one hundred years ago, the impact test by a pendulum machine has been well
developed and now is used widely to measure the energy absorbed during fracture of
a test specimen for a wide range of materials and conditions.
The impact test has become a useful material evaluation technique owing to the
standardisation of the test and its fast performance. As known, the results obtained
from an impact test are strongly dependent upon the specimen geometry, the
geometry of the anvils and striker, and the machine conditions. Furthermore, the
scatter in the test results between individual machines also can be large due to the
different causes [2]. Without standardisation, the impact test has little meaning [3].
Both in ASTM E 23, Standard Test Methods for Notched Bar Impact Testing of
Metallic Materials (S 1), and EN 10045-2, Metallic Materials - Charpy impact test -
Part 2: Verification of the testing machine (pendulum impact) (S 2), verification tests
are required in order to reduce the scatter of the measured data. The time interval
between two verifications, however, is difficult to determine since it depends on the
maintenance standard and number of times the machine is used (S 2). The purpose of
the present study was to better understand how tough specimens affect the precision
of impact testing and to optimise the time interval beween the verifications through
regular stability control tests.
Material and Specimens
The material used for the stability control test was an austenitic stainless steel
with a nominal chemical composition as shown in Table 1.
Table 1 - Nominal chemical composition of the steel (UNSN08028) (wt%)
C Si Mn P S Cr Ni Mo Cu
0.020 0.6 2.2 0.025 0.015 27 31 3.5 1.0
This grade steel has a stable structure. The as-received material was a hot extruded
bar with a diameter of 120 mm. The bar was cut into different blocks with
thicknesses of approximately 60 rnm. These blocks then were heat treated in order to
obtain a homogeneous microstructure of the material. Figure 1 shows the
microstructure of the 3rd block. The microstructures from both longitudinal and
transversal directions are comparable but the grain sizes are quite large. The average
grain size is about 130 ~tm (equivalent to ASTM 2.5). The microstructures from other
blocks are also similar.
The specimens for impact testing were taken from the blocks according to
SS 11 01 51 Metallic Materials - Sampling for impact testing (S 3) in the
longitudinal direction and from half the radius of the block.V-notched test pieces
with reduced section (5 mm) according to EN 10045-1, Metallic materials - Charpy
impact test - Part 1: Test method (S 4), were used.
Figure 1 - Microstructure of the steel usedfor the test. (a) from longitudinal
section, (b)from transversal section. Magnification 50X.
Experimental
The test machine is a 300 J C-type of pendulum with a daily test of approximate
30 samples. Commonly these samples are very tough and generate energy values
above 80% of the machine capacity. Because of the velocity change of the pendulum
during these high energy tests, results are of course always expressed as approximate
according to valid standard (S 1).
The stability control test was started by a verification test with standard reference
specimens according to (S 2) and followed by 10 specimens from the stability
control batch. Then the specimens for the stability control test were regularly tested
on the first day of every week. Five specimens were tested for each control test. The
total number of daily tests and especially the number generating energy values above
240 J between two control tests were documented. The high energy specimens are
expected to affect the wear more than the less tough ones.
Since the material used for control test is very tough at room temperature, and
also to ensure minimal temperature variation, the control test was performed at -
196~
Results and Discussion
Impact Energy of the Material and its Effect on the Deviation
Figure 2 shows the impact energy of the specimens tested for stability control as a
function of time. The average impact energy calculated for all control specimens of
the material is 120 J with a standard deviation, c, of 5 J. Most of the impact data are
within the average value + 2~.
140
2 n d anvil 3rd anvil
13s . . . . . . . . . . . . . -"- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
average + 2~
, "~
130 t- ~--e-
|125 . . ". . ="

. . . :~:i . . . x:. ....
. . " . . "=. . ". . ". . . x--*
. . . = .... 1 9
= . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x 9
~ . 120 ~ 9 ;: 9 = = 9 7 I ,:C I *
. . . . . . . = . . . . . . . . . . . . . . . . . . . _x . . . . . . . . =---
i 9 It
average - 2~
10s . . . . . . i - - - _r_ . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . .
1$t b l o c k ; 2nd b l o c k 3rd blOck i 3rd r

100 I I I ~ b 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
Date
Figure 2 - Summary of the impact test results from the control batch at -196%.
During the entire testing period, three anvils and material from three blocks were
used. The results can be divided into three regions according to the different anvils.
The blocks have to be considered as well. The average impact energy of the material
from the individual blocks is summarised in Table 2. The difference in the impact
energy between individual blocks is a maximum 6 J, which is of the same magnitude
as the standard deviation of all samples tested. This implies that the analysis of the
stability should be limited to the same control block and anvil.
Table 2 - Impact energy of an austenitic stainless steel (UNS NO8028)from

individual block at -196~
Block 1 Block 2 Block 3

Impact energy, J G, J Impact energy, J o, J Impact energy, J (3, J
117 4 123 4 118 4
Although the test was performed at -196~ the material still shows high impact
energy values. The fracture investigation showed that the fractures of the specimens
were ductile with small inclusions in the dimples ( Figure 3). Unfortunately, scatter
of impact energies due to material properties is usually larger in these energy levels
than for more brittle material. Grain sizes can affect the scatter as well. This certainly
does not make the analysis of changes due to wear of the anvils easy.
Figure 3 - Fractures from impact test of an austenitic stainless steel (UNS

N08028) at -196~ (a) from block 2, (b) from block 3.
Deviation Resulted from the Damage of the Anvils
Wear of the Anvils - The anvil is the component in a pendulum machine that has
to be changed most often due to its damage. Figure 4 shows the part of an anvil
which is worn out (anvil no. 3). Small cracks occur on the edge of radius which has
become blunt. When measuring the radius in the dull area with a radius gage, it is
found to be somewhere between 1 and 1.25 mm.
The anvil can be worn out in two main ways. The wear either can proceed
successively, where the impact testing causes a small local plastic deformation and
further on formation of cracks leading to a blunt edge with a bigger radius as a
consequence. The damage also can be caused by instant failure where a piece of
material suddenly is peeled off. A combination of the two failure mechanisms is also
common (Figure 5).
Figure 4 - Damage part of an anvil due to impact

testing. Magnification 3X.
Figure 5 - Damaged part of an anvil with a piece of material peeled off
The wear of anvils of course is accelerated when testing very tough materials and
for good reasons is believed to have an effect on impact testing results. As known,
the impact energy indicated is strongly dependent upon the distance between the
anvils. If a deformation takes place, the distance may be affected with changed
energy values as a result. Friction may be influenced as well due to the wear of the
anvil and will contribute thereby to the absorbed energy indicated.
Effect of Total Number of Impact Testings and Toughness of the Tested Material
on Deviation - The evaluation is concentrated on anvils no 2 and 3. The first anvils
are not included in the analysis due to instant failure after only a short time i n
operation.
Figures 6 and 7 show the impact energy of control test specimens with the
accumulated number of daily samples tested.
145 / 2rid anvil, 2rid block
140 .I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135 . . . . . . . . ~,~, ~ . . . . . . . . . . . . . . . . . -~.~-~'-~ ; ..........
~ 125-
~ 120 - -
_E
11, . . . . . ; . . . . ---', . . . . . - . . . . . . . . . . . : - - - - - - - : - - - ....
I \ .~ . = f
110 . . . . . . . . . . . . . . . . . . . . . . -"~ ; ,,~-'- . . . . . . . . . . . . . . . . . . .
~Average
~ l l - Average - 2s
105 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .~A- A v e r a g e §
100 I q I I I I I I I
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Total n u m b e r of specimens
145
2nd anvil, 2nd block
140 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135 ......... ;~: ............. - >-*- - -,.-:- ..........
_ 13o . . . . . 7-'- . . . . . . -~-- =-'~ - - - . p ~ ' - . . . . . . . . . . . . . . . -~-< - - -
_s . " " ' d

115 .... .,"~ - . . . . . . . . . . . . . -~ . . . . . . . . . . ; ~'--- :---'---- ....
,,o ::::::::::::::::::::::::::::::::
. _ - -_.. ~. : _ _ _
105 .--A- Average + 2s

b)
100 I I I I I I I
100 200 300 400 500 600 700 800
N u m b e r of s p e c i m e n s > 2 4 0 J
F i g u r e 6 - Variation of impact energy of control test specimens with times of testing

for the second anvil, a) from number of totally tested specimens and b) from tested
specimens above 240 J..
145 -
3rd anvil, 3rd block
140 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
N13o .................. ~'- ..... ~-x~- ...................
,25 ................ !; ......... ?-',, ................

= =-~-'--"-~2---'--"-~ . . . . . ~-~7~,~---
11o '~ . . . . . . . . . . . /
105 . . . . . .
9
~
...
. . . . . . . . . . . . . . . . . . . . . .
--,.ra,
, , , , , 4- Average - 2s
;;
' ~ & " Average + 2s
100 I I P I I I
0 500 1000 1500 2000 2500 3000 3500
Total n u m b e r o f specimens
145
3rd anvil, 3rd block
140 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135 A,
~A
130
N I "\
i 125 ............. -/ ......... ~ ........... ,-A. . . . . . . .
. . . . . . .
E 115- P" ~ .........
110 ! ,%. . . . . . . . . . ~___-~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
,.11 . - 4 - Average - 2$
105 ---'%-~-~11-'-'~--" . . . . . . . . . . . . . . . . . . . . . ,~i,- Averaf/e+2s ---
100 I I I I I
0 200 400 600 800 1000
Number of specimens > 2 4 0 J
Figure 7 - Variation of impact energy of control test specimens with times of testing
for the third anvil, a) from number of totally tested specimens and b) from tested
specimens above 240 J.
The average value changes with the number of samples tested, but no tendency of
increase or decrease of this value can be observed. The differences are small
compared to the scatter of the material itself. In the "b) part" of both figures the
average energy is plotted against the number of tough specimens tested. As expected
these curves show almost the same shape as those referring to the total number of
specimens tested but give an idea of the amount of tough specimens usually tested in
this machine. The result shows that stability control of the machine with control
specimens cannot be used as a determination when a recalibration is needed, at least
not in this energy level of control specimens. Use of a more brittle material with less
material scatter possibly could show more obvious tendencies of energy changes as
the anvils are gradually worn.
However, there is one interesting thing to be mentioned in this connection. For
both anvils approximately the same number of tough and totally tested specimens
were performed before the anvils were considered worn out by visual inspection. The
numbers of specimens (total vs tough) were 3063 and 1048 respectively for anvils no.
2. The corresponding number of samples for the third pair of anvils was 3378 and
1021 respectively. This could give a slight indication that approximately the same
number of, at least, tough specimens can be tested until the anvils have to be changed
and the pendulum recalibrated.
Another conclusion that could be drawn from the results is that the interval of
changing anvils with corresponding calibration of the machine is short enough in this
case as the mean of control specimens does not change during the period of testing.
This is valid at least for energy levels that are not below 120 J. Further control in this
energy level in combination with a lower one should be made in order to understand
the influence of machine wear on the impact test results.
Concluding Remarks
1) Stability control testing with specimens in this energy level will not be influenced
by acceptable wear of anvils according to visual inspection - neither the average
values nor the standard deviation.
2) The verification interval of a pendulum machine seems to depend on the number
of high energy specimens tested, especially the specimens with energies above 240 J.
3) Further investigation should be done with control specimens of a lower energy
value.
References
[1] Russell S B, "Experiments with a new machine for testing materials by impact",
Transactions ASCE, Vol. 39, June, 1898, p.237.
[2] Fahey N. H., "Effect of variables in Charpy impact testing", Material Research
standards, Vol. 1, No. 11, Nov. 1961.
[3] Manahan M. P, McVowan C. N., Siewert T A, and Holt J M., "Notched bar
impact testing standards have yielded widespread benefits for industry",
ASTM Standardization news, February, 1999, p. 30.
G6rard Galban, 1Gilbert Revise, 1Denis Mougin, 2 St6phane Laporte, 1and St6phane
Lefran~ois~
Indirect Verification of Pendulum Impact Testing Machines:

The French Subsidiary from Its Origins to the Present, Changes in Indirect
Verification Methods, Effects on Dispersion, and Perspectives
Reference: Galban, G., Revise, G., Mougin, D., Laporte, S., and Lefranqois, S.,
"Indirect Verification of Pendulum Impact Testing Machines: The French
Subsidiary from Its Origins to the Present, Changes in Indirect Verification
Methods, Effects on Dispersion, and Perspectives," Pendulum Impact Testing: A
Century of Progress, STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds.,
American Society for Testing and Materials, West Conshohocken, PA, 2000.
Abstract: A straightforward method of checking pendulum impact testing machines was

developed at the end of the 1970s at the request of French industrialists. This method
was based on the use of reference test pieces to compare the energy provided by a
reference machine with that of test (or industrial) machines. These principles are outlined
in French standard NFA 03-508 and France had a national source of unnotched,
reference test pieces. The manufacture of reference test pieces, characterization of
calibrated batches and the indirect verification of industrial machinery therefore have
been practised for twelve years or so.
Various European studies and changes in standardisation protocols have led to the
replacement ofunnotched test pieces by Charpy notched test pieces. France developed a
source of reference test pieces of this type in the early 1990s. The French system, which
is based on a partnership comprising the F6d6ration Fran~aise de l'Acier (French Steel
Federation), steelmakers Aubert and Duval, and national laboratories (CTA and LNE),
has been in operation since European standard EN 10 045-2 was first applied in 1993.
Experience together with discussions between partners and users has led to considerable
advances in the processes used to manufacture these test pieces, thus improving the
dispersion obtained: The quality situation, especially accreditation procedures
implemented by national organisations such as COFRAC (Comit6 Fran~ais
d'Accr6ditation - French Accreditation Comity), has also contributed to advances in
perfecting methods and assessing approximations.
1National Testing Laboratory (LNE), Zone Industrielle Trappes-Elancourt,

29 Avenue Roger Hennequin, 78197 Trappes Cedex.
2 Arcueil Technical Centre (CTA), 16bis Avenue Prieur de la C6te d'Or, 94114 Arcueil
Cedex.
109
Keywords: Indirect verification, reference test piece, error, repeatability, uncertainty.

Impact testing machines, unnotched test pieces.
Introduction
The metrological requirements of laboratories with impact testing machines used to

characterise the mechanical features of metallic materials demanded proof of high-quality
performance. The only feasible method comprised verification of the geometric and
physical parameters of the machine - a method that was taken up and supplemented in
the form of the direct method advocated in current standards. The length of time taken to
carry out this procedure proved disadvantageous. Moreover, it was impossible to make
any connection with physical values in terms of test results. The so-called "indirect"
technique was therefore developed, providing a rapid, straightforward method for
checking machine performance, a link with national standards and a source of laboratory
inter-comparison.
The development of a national source of reference test pieces nevertheless called for
assurances as regards production and operating techniques in order to reach a satisfactory
level of homogeneity to correct the effect of the test piece on the one hand, and on the
other hand, to improve machine performance, thus providing a means of reference
characterisation.
Indirect Verification Using Unnotched Test Pieces (from 1970 to 1993)
The Choice of Unnotched Test Piece
As outlined above, the indirect verification process is intended to provide

industrialists with a straightforward, rapid and inexpensive method for checking the
performance of their impact testing machines in terms of accuracy and repeatability and to
enable them to assess any potential drift. In view of the context, since this involved
comparison of the energy output of reference and test machines, unnotched test pieces
were chosen as these have a better dispersion potential given by the absence of the notch.
They also facilitate production and are less expensive.
Standardizing References
Standardisation requirements relating to verification methods for pendulum impact

testing machines have led to the compilation of national standards. French standardisation
bodies under the aegis of the Association Fran~aise de Normalisation (AFNOR) (French
Association for Standardisation) in conjunction with the Bureau de Normalisation de la
GALBAN ET AL. ON IMPACT TEST MACHINES 111
Sid6rurgie (BNS) (Organisation for Standardisation in the Iron and Steel Industry)
formed working parties which resulted in the publication o f standard Verification o f
Impact Testing Machines (NFA 03-508 in June 1967). The inspiration for this standard
came from ISO Recommendation Verification of Pendulum Impact Testing Machines for
Testing Metallic Materials (R 442 o f July 1965). Apart from the implementation o f a
direct verification method, Recommendation R 442 already introduced a technique based
on the comparison o f energy values between a reference machine and an industrial
machine using urmotched test pieces.
Standard N F A 03-508 was revised thoroughly in 1985 to introduce the features o f
the reference machines and to define the indirect verification method. This version applied
up to 1990.
The various types of reference test pieces are defined in standard N F A 03-508. The
method for collecting these test pieces is also stipulated. Thus in one bar, 5N cuts are
made with N ranging from 6 to 10. Test pieces 1, N+I, 2N+l, 3N+l and 4N+l are bent
on the reference machine. The others are used to check the test machines. The test
temperature is set at 20~ + 2~
Differences between the 1967 and 1985 versions of the standard in terms o f the
principal criteria are shown in the following Table 1:
Table 1 - Evolution of Standard NFA 03-508
N F A 03-508 Ed. 1967 NFA 03-508 Ed. 1985

Batch uniformity criterion T5-T1 < 5xT1/100 T5-T1 _< 1 J where T < 40 J
T5-T1 _<2.5% T where T >_40 J
Number of verification levels 1 3
(10% Ep, 20to 30% Ep, 50% Ep)
Number of test pieces per level 5 5
Acceptance criteria for test Tm-T_< 2 J where T < 40 J
machines: . Error Tm-T < 5% T Tm-T<5%T whereT>40J
9Repeatability T5-T1 < 10% T1 T5-T1 < 2 J where T < 40 J

T5-T1 < 5% T where T > 40 J
where:
Ep = potential energy.
T = reference energy value.
TI and T5 = energy values measured.
Tm = mean value measured.
The French System
This revolves around two laboratories which are very involved in national metrology:
the LNE - a public institution with industrial and commercial leanings, and the CTA -
another public institution.
Both laboratories have reference machines and benefit from experience in the field of
impact deflection tests and in test machine verification. The manufacture of reference test
pieces also was managed independently by the two laboratories despite the fact that they
shared the same base metal supplier. Each laboratory characterises its own test pieces and
therefore issues its own calibration certificates. The provision of test pieces for checking
test machines is also managed independently by each laboratory.
Unnotched Test Pieces: L N E Subsidiary
Reference Machine - The reference machine used is a Tinius Olsen 74 Impact

pendulum impact testing machine, the characteristics of which comply with ASTM Test
Methods For Notched Bar Impact Testing of Metallic Materials (ASTM E 23). It has a
U-type hammer and an energy potential of 358 J (Figure 1).
Figure 1 - The Tinius Olsen machine
The machine is fixed on a foundation weighing 1313 kg, which is more than 40 times
that of the pendulum.
Manufacture o f Unnotched Test Pieces - Unnotched test pieces are manufactured

strictly in compliance with standard NFA 03-508.
The technique involves the use of one steelmaker, who produces the steel according to
the same manufacturing process, and one machine-tool operator, who makes the
unnotched test pieces according to the same working method. The steelmaker provides
steel bars (XC10 grade). The bars are cylindrical with a diameter of 14 mm and are
produced by hot-rolling. They are 3.20 m in length. Each batch is identified in
alphabetical order. After 1990, the bar section became rectangular in shape near the final
section, thus reducing the number of machining processes required.
GALBAN ET AL. ON IMPACT TEST MACHINES 1 13
The steelmaker checks the mechanical characteristics of the steel in order to guarantee
homogeneity.
The test pieces are machined with the following dimensions:
- length 55 mm,
- width 10 mm,
- variable thickness to reach different energy levels: 2 . 5 , 3 , 4 , 5 , 7 mm,
- 90 ~ interfacial angles.
These sides are designed with a tolerance of 0.01 mm in terms of length and 0.1 ~ for
the angles. The finish is achieved by grinding.
Each test piece is individually marked with its own specific number on one of the
surfaces bearing the bar number
Example: A8 82 means the 82 nd test piece in bar A8.
Bending on the reference pendulum impact testing machine then is carried out
according to a highly specific identification procedure. The test pieces are identified in the
same bar one after the other as follow:
X, LNE1, Y, Z, LNE2, T, X, LNE1, Y, Z, LNE2, T, X, LNE1, ...
Codes X,Y,Z and T represent test pieces building up different client sets (5 test
pieces for each set), LNE1 the characterised set on the reference machine and LNE2 a
replacement set.
The characteristics of the sets of test pieces provided by LNE are shown below
(Table 2):
Table 2
Machine energy Thickness of the test pieces Mean bending energy Uncertainty
5.5 mm 100 J 0.3%
150 J 4.0 mm 50 J 0.6%
2.5 mm 20 J 1.9%
7.0 mm 180 J 0.2%
300 J 5.0 mm 80 J 0.4%
3.0 mm 30 J 1.2%
As described before, five test pieces are bent on the reference machine for each
energy level.
The test pieces intended to check industrial machinery are conditioned with
protective graphite and packed in plastic boxes with foam protection ready for delivery.
BNMAccreditation - The laboratory of the Mechanical and Equipment-Testing

Department of the LNE received CEA accreditation (Centre d'Evaluation Agr66 -
Accredited Evaluation Centre) by the BNM (Bureau National de M6trologie - National
Bureau of Metrology) in 1985 under number 80.505. This accreditation bestows
national recognition in this calibration activity. The scope of the accreditation applied
to:
the parameter measured: bending energy
the range of measurement: 25 to 287 J
uncertainty: _+ 1.5%
- material: steel
- method: in accordance with the following standards: N F A 03-508, ISO R 442.
Calculation o f Uncertainties - This is based on determination o f the bending energy

measured (E), the formula for the calculation o f which is given below:
E = M cos (13 + 1/2 (2 15 **2 x 10 **-5 + 50 15x 10 **-5)) - M cos ~ - PL - Efd (1)
where:
M = pendulum moment
o~= angle of fall
13= angle of rise
PI = energy loss due to the expanding device
Efd = energy loss during arm descent
Since values PI, Efd are low, so is their degree o f approximation. This can be
overlooked compared with the other values. The same applies to values in 10"*-5.
Since the uncertainties expression (I) includes physical parameters, its calculation
is directly based on measuring techniques used to determine these parameters.
The following values were used by LNE:
Over the 25 J to 287 J range: I = 1.5% E
The CTA Subsidiary
Reference Machine - The reference machine used is a 300 J Otto Wolpert pendulum
impact testing machine fitted with a C-type hammer. Figure 2.
Figure 2 - The machine

Manufacture o f Curved Test Pieces - The test pieces are manufactured strictly in
accordance with standard NFA 03-508. The material is supplied by one steel maker - the
Aubert and Duval Company - as a 10 NC 6 grade bar. Machining and preparation are
carried out by the CTA. The key stages are similar to those of LNE with the following
exceptions:
- annealing at 825~ under low vacuum before grinding,
- grinding of the finished sides,
- stress-free annealing at 200~ for 2 hours.
B N M Accreditation - The CTA laboratory received BNM accreditation on 7 June

1985 under number 75-502. BNM accreditation of the CTA laboratory was awarded on
the same basis as that of LNE. The scope of the accreditation applied to:
- the parameter measured: bending energy,
- the range of measurement: 15 to 200 J,
- uncertainty: + 1.5%,
- material: steel,
- method: in accordance with standard NFA 03-508.
Uncertainties- The calculation of uncertainties includes determination:

- of the pendulum moment which can be broken down into:
9 the value of the pendulum force,
9 the distance 12 value;
energy losses;
-
angle values for pendulum rise and fall;

-
geometric parameters relating to:

-
9 the distance between supports

9 centring of the test piece.
The uncertainty value (I) was determined taking the various energy levels into
account. Thus:
- for the range 15 J < E < 50 J: I = + 0.2 + 5.4. 10 ** -4 E with an ann of 150 J,
- for the range 50 J < E < 150 J: I = + 0.3 + 5.4. 10 ** -4 E with an arm of 300 J.
lntercomparisons
Table 3 shows the position of the two laboratories as regards crossover tests.
Table 3
Mean values (in Joules)

1987 1991
LNE CTA LNE CTA
20,5 20,4 22,5 22,4
26,9 26,5 24,3 24,8
63,7 62,7 68,9 67,6
83,2 81,9 72,8 72,4
131,9 129,4 134,8 132,8
180,6 177,6 161,8 159,9
Overall, these results indicate:

9 good reproducibility of results for both laboratories and between the two
intercomparison periods,
9 a slight shift between the two machines in terms of absolute value, increasing with
high energy values (1.7% deviation).
Performance Assessment
Average distribution ofunnotched test pieces comprised 150 to 250 sets annually
between 1981 and 1992, i.e., approximately 33000 test pieces.
For the CTA: on average 50 sets per year, i.e., around 9000 test pieces.
Changes In Standardisation Protocols, the Use of Notched Test Pieces
BCR Studies and the Supply of BCR Test Pieces
Initial intercomparison with Charpy V test pieces took place in 1983 following an
enquiry phase carried out by Euratom d'Ispra. The following laboratories were involved:
NPL, BAM and LNE. The results revealed substantial dispersion and prompted LNE to
investigate the effect of geometrical parameters (study submitted to ASTM in 1990) [1].
BCR launched a programme for the manufacture of Charpy V reference test pieces but
difficulties in finding a steel maker meant that the batches were not available until 1986.
10 000 test pieces were produced at 30, 60, 80 and 120 joules.
France was faced with the dilemma of which reference test pieces to choose:
unnotched or Charpy test pieces.
lntercomparisons
Numerous intercomparison programmes were conducted over this period both in

Europe (led by the BCR) and France.
Thus in April 1988, a comparative study between LNE and CTA was carried out
within the scope of investigations by the ECISS and the ISO using unnotched and Charpy
test pieces. The same material produced by Aubert and Duval (alloyed steel with a Rm
1300 MPa) was used in this comparison. The semi-finished product was a bar, 18 mm in
diameter. The test pieces were worked as follows by LNE:
- 40 slugs, 55 mm long,
- repeated individual identification,
- No. pairs = V-notched test pieces machined in accordance with standard Impact
Test on Steel Materials (V-notched specimens) (NFA 03 161) + surface grinding,
- No. impairs = unnotched test pieces machined in accordance with standard
NFA 03 508.
The thickness of the unnotched test pieces was determined in advance in order to
obtain energy values similar to those of the Charpy test pieces (i.e. - 40 J). Two batches
were produced and characterized on the two reference machines. The results are given in
the following Table 4.
Table 4
Mean Standarddeviation Dispersion(Em~-Emin)
Unnotched CTA 41.2J 0.5J 1.7J

LNE 40.3J 0.3J 0.9J
V-notched CTA 39.2J 3.0J 6.3J
LNE 35.7J 1.9J 4.5J
The results confirm the considerable dispersion with notched test pieces leading
France to continue using unnotched test pieces.
The French Situation in the European Context
In the European context, France was isolated quickly in its defense of the
unnotched test piece because of two conflicting principles: on the one hand, good
repeatability o f results with the unnotched test piece and comparison of the intrinsic
performance of two machines during indirect verification and, on the other hand, the use
of the same test piece as that employed in the impact test.
It therefore became difficult to pursue the French method although in Europe, France
was one o f the few countries to use a national source of reference materials for these
verification procedures.
The various standardisation meetings conducted within the ECISS led to ratification
of the adoption of the Charpy test piece.
Changes in European Standardisation Protocols
The indirect verification method was introduced into European Standard Metallic
Materials - Charpy Impact Test - Part 2: Verification of Pendulum Impact Testing
Machines (EN 10045-2) as from December 1992. It is based on the use ofCharpy V
reference test pieces. Industrial machines are checked on at least two levels within the
scope o f application of pendulum impact testing machines. The test is conducted using
test pieces allowed to stand at a temperature of 20 + 2~ Error and repeatability are
determined as in the previous method. The acceptance criteria are listed in the following
Table 5.
Table 5 -Acceptance Criteriafor Testing Machines
Energy level Repeatability Error

<40J <6J <4J
>40J < 15%E < 10% E
Appendix B of the standard also stipulates the criteria for preparing and
characterising reference test pieces. The geometric parameters o f a reference machine are
also specified. Moreover, the reference pendulum impact testing machine must be used
only for the characterisation of reference test pieces. The deviation values for the
validation of these test pieces are listed in the following table 6.
Table 6 - Standard Deviation for Batches of Reference Test Pieces
Energy level Standard deviation

<40J <2J
>__40J <5%E
The French Choice
Since the European standard acts as a substitute for the French standard, France was
obliged to take this standard into account. Furthermore, experience gained in the
manufacture and characterisation ofunnotched reference test pieces and the need for
French industrialists to continue with this indirect method led to the creation of a French
source o f Charpy reference test pieces.
The French Subsidiary for Charpy Test Pieces
The Partnership
At the same time as the European standard was introduced, armed with its
experience in the indirect verification of pendulum impact testing machines and the
problems associated with the production of reference materials, France decided to retain
this expertise and to consolidate the organisation by creating a partnership bringing
together the FFA (Frdrration Fran~aise de l'Acier - French Steel Federation) representing
industrialists and standardisation authorities, steel makers Aubert and Duval, and the two
Cofrac-accredited national laboratories to characterise impact test pieces.
Qualification Studies
The methodology was published within the scope of the 1995 symposium [2]. The
broad outlines are given below:
- Selection of materials and processing: Steel covering a broad energy range (15 to
150 J) for the same grade was chosen, acting at the tempering temperature. The grade
selected was NC 40 MW alloyed steel (ASTM equivalent A540 grade B 23).
Qualification tests: The tests were carried out within the scope of a French
intercomparison programme. The batch size for KV France test pieces was cut to 120 test
pieces per level (20 J, 70 J, 120 J). Fifteen test pieces (5 unnotched, 5 BCR, and 5 KV
France) were used for three energy levels (20-30 J, 70-85 J and 120-180 J) in the
intercomparison. 14 machines were selected from tests conducted with unnotched and
BCR test pieces.
The results of the qualification tests revealed dispersion values similar to those
recorded with BCR test pieces. This was deemed satisfactory to launch the
industrialisation phase.
The Industrialisation Phase
Selection of Semi-finished Products and Processing - The semi-finished product

should guarantee a dispersion of energy values in accordance with the requirements of
standard EN 10045-2 while permitting "batch sizes consistent with industrial
manufacturing at acceptable cost prices. Batch size was set at 1000 test pieces. To this
end, a form optimisation test was carried out to ensure that the form selected was
conducive to heat treatment which was as uniform as possible. The following processes
were carried out:
- quenching which promotes uniform characteristics in the test piece sampling areas
- tempering to obtain the different energy levels: 15 to 20 J; 70 to 75 J; 110 to 120 J.
- control o f the homogeneity of the semi-finished product.
Machining of test pieces - The L N E is responsible for the manufacture o f reference

test pieces. The L N E and the C T A have drawn up specifications for the machining and
control o f test pieces. Some tolerances have therefore been reduced compared with the
requirements o f standard E N 10045-2. The dimensions are listed in the following Table 7.
Table 7 - Dimensions of Reference Test Pieces
Designation Dimensions French reference E N 10045-2

test pieces
Length of test piece 55 mm +0 +0
- 0,25 mm - 0,25 mm
Height of test piece 10 nun + 0,04 mm + 0,06 mm
Width of test piece 10 mm + 0,04 nun _+0,06 mm
Angle of notch 45 ~ _+ 1~ + 1~
Ligament length 8 nun + 0,04 mm + 0,08 mm
Radius of curvature of base of notch 0,25 mm _+0,025 mm • 0,025 mm
Distance between the symmetry plane of 27,5 nun _+0,10 mm +_0,10 mm
the notch and one of the endfaces of the
test piece
Angle between the symmetry plane of the 90 ~ + 2~ + 2~
notch and the longitudinal axis of the test
piece
Angle between adjacent faces 90 ~ + 0,10 ~ _+0,10 ~
T w o machine-tool sub-contractors were qualified.

Thus, one o f them used conventional cutting methods, i.e.,:
- machining by milling with finished sides o f + 0.1 to 0.2 mm,
- surface grinding and notch creation by wheel grinding.
A photograph o f the notch is shown in Figure 3.
Magnification x 25
Figure 3 - Level 70 J
The second machine-tool operator cut the test pieces by means o f electroerosion.
Thus, apart from creating the notch by diamond cut grinding, the general shape is
obtained with virtually finished dimensions (a few hundredths) (Figure 4).
Magnification x 25
Figure 4 - Level 70 J
Characterisation - Fifty test pieces are collected at random for each batch supplied:
25 having been to be characterised by LNE and 25 by CTA. The dimensions of each test
piece are checked prior to rupture on the reference pendulum impact testing machine.
The test pieces are conditioned in the laboratory at 20~ + 2~ The test pieces are
broken at the same temperature in accordance with the instructions given in standard
Metallic Materials - Charpy Impact Test - Part 1: Test Method (EN 10045/1).
The two laboratories calculate the mean breakage energy value and the experimental
standard deviation o f each batch of 25 test pieces. Batch acceptability (according to
criteria specified in standard EN 10045-2) and reference energies can be determined from
these results. Within the scope of their accreditation by COFRAC/Calibration, both
laboratories established optimum uncertainty from the methods used to calibrate the test
pieces. The calculation principle is described in the following paragraphs.
Results - The following examples are used to rank the results obtained with batches
o f the same characteristics on the two reference machines (Table 8).
Table 8 - Examples Using Different Batches (Units in J)
Batch E lne E etca S Ine S etca S std 1 Eref Sref 2

1D 23.11 22.84 0.78 1.22 2 23.1 1.8
3E 68.74 65.73 0.86 0.87 3.4 67.2 3.5
5F 113.96 109.56 2.92 3.15 5.6 111.8 6.7
1 S std = Standard deviation allowed by the standard.
2 Sref incorporates the experimental uncertainty and that of the method. It is expressed with
two c.
Uncertainties ( L N E example) - Uncertainty was calculated with reference to the

Guide to the Expression of Uncertainty in Measurement (ISO T A G 4).
Uncertainties are calculated on the basis o f the principle used to determine the
amount o f energy absorbed during breakage, as explained before.
Thus:
E a = M (cos {3 - cos ~ ~ (2)
After correction for energy losses:
E a = M (cos (13 + h (13)) - cos ct ~ - P + P'max / 2 (3)

where:
P = total energy loss,

P'max = maximum loss due to air and rolling,
h (13) = losses during rise,
E a = M (cos (13 + h (13)) - cos 13~+ 1/2 cos f~l - 1/2 cos ctl) (4)
where:
131 --the rise angle for a angle of fall ctl.

13~--the rise angle for a angle of fall ct~
Since the measuring system is an optical coder, energy Ea is expressed as follows:

Ea = M {cos ( 3 6 0 / I c max (Icl3 + 1/2 f(Icl3)) - cos ( ( 3 6 0 / I c max) x Icl3~
+ 1/2 cos ( (360/ Ic max) x Ic131)- 1/2 cos ((360/ Ic max) x Iccd)} (5)
where:
Ic max = 10000 digits

or in numerical terms:
E a = 240.225 [cos (3.31 10"*-2 + 3.5982 10"*-2 (Ic13)
+ 1.61 10"*-8 (Ic13"'2)) + 0.49462] (6)
Uncertainty in the Ea calculation takes account o f uncertainty in the following

variables:
M , Ic m a x , Ic13, Ic13~ , Icl~l, Iccd.
Non-correlation o f these variables was assumed.
The uncertainty factors are issued from the measuring techniques. Uncertainty o f the
E a calculation is obtained by the quadratic addition of the elements o f uncertainty to give
the following expression:
~]U 2 M ( E a ) + U 2 I c m a x ( E a ) + U 2Ic13 (Ea) + U 2Ic13 ~ (Ea) +
U (Ea) = U 2 Icl31 (Ea) + U s Ic~l (Ea). (7)
The uncertainty associated with geometrical parameters is added to this uncertainty.

The following factors are taken into consideration:
- the distance between anvils, and its uncertainty Uan (Ea),
centering o f the test piece, and its uncertainty Uc (Ea).
-
The other parameters are overlooked since these are of secondary importance:
Utot (Ea) = x/U 2 (Ea) + U2an (Ea) + U2c (Ea) (8)
Global uncertainty calculated in this way therefore is expressed as follows:

Utot (Ea) = 0.314 + 8.43E-4 Ea (in Joules) (9)
In order to ensure national uniformity between the two accredited laboratories; the
uncertainty taken into account is expressed as follows:
Utot (Ea) = + (1 + 3 10"*-3 E) (in Joules) (10)
Determination o f National Reference Energies - Reference energy is determined

from the results obtained with the two French reference machines. Thus a statistical
calculation principle has been devised in order to take machine-induced dispersion into
account9 In fact, as shown earlier, the two machines are of different design but are
nevertheless typical o f the range of test machines available. Moreover, as can be seen
from the various intercomparison results, they yield high-energy results that deviate in
terms of absolute value. Consequently, by taking an average of the values obtained with
both machines, no preference is shown for one group of industrial machines over the
other.
Reference energy values then are calculated for the sets of 5 test pieces that will be
used for the indirect verification of industrial machines. So uncertainty is determined by
applying Student's law:
9 $5 = t ( 0 . 9 7 5 / 4 8 ) s/x/5,
. $5 =2.011 s/x/5.
Energy-related uncertainty takes an expansion factor of 2 , i.e., 2 S 5.
Table 9 gives examples of different batches characterised.
Table 9
Batch SS(E)
1 D (20 J) 1.84 J
3 E (70 J) 3.46 J
5 F (120 J) 6.72 J
The situation regarding accreditation
The accreditation o f cafibration laboratories - During the creation of French

sources of reference test pieces for the indirect verification of pendulum impact testing
machines, the need to attach this activity to the national calibration programme in order
to lend it recognition was presented to the BNM. This activity consequently was
incorporated in the reference materials subsidiary and both laboratories received BNM
accreditation.
Changes in metrology in the early 1990s reflected on this activity in two ways: on the
one hand, it affected its position in the calibration chain and, on the other hand, it affected
its accreditation principles. Thus the two laboratories were assimilated with the services
de m&rologie habilitrs (SMH) (expert metrology services). Accreditation focuses not
only on the technical aptitude of the laboratory but also on the quality structure
introduced by the latter9
The regulations for accreditation applied by COFRAC/Calibration are based on
compliance with ISO General Requirements for the Competence of Calibration and
Testing Laboratories (ISO/CEI n ~ 25). A quality structure complying with the
requirements stipulated in standard General Criteria for the Operation &Testing
Laboratories (EN 45001 - December 1989) and ISO/CEI n ~ 25 has been implemented for
this activity.
Agreement - Laboratory accreditation leads to the drawing up of an agreement with

COFRAC who defines:
9the laboratory's commitments,
9 a list o f signatories of calibration certifications,
9the scope of the accreditation,
9 optimum uncertainty proposed by the laboratory.
lntercomparisons - Involvement in intercomparison studies is required within the

scope of accreditation. This is manifested by regular comparisons with BCR test pieces
and naturally between the two laboratories.
The following intercomparison studies have been carried out :
RNE intercomparison in 1990 (Table 10):

Table 10 - Results Obtained with Charpy Test Pieces
Level B C R reference v a l u e L N E value C T A value

m s m s m s
20 J 25.5 J 0.9 J 26.06 J 1.5 0J 24.58 J 1.47 J
70 J 74.1 J 1.1 J 73.22 J 1.36 J 70.00 J 1.48 J
120 J 123.3 J 2.65 J 123.42 J 2.70 J 113.84 J 5.46 J
9BCR intercomparison in 1997 (Table 11):

Table 11 - 20 J, 120 J a n d 160 J epr Results
Level B C R reference value LNE value

nl $ Ill $
30J 30.0 J 1.4 J 30.54 J 1.34 J

120 J 116.6 J 4.9J 115.97 J 5.48 J
CTA/LNE comparisons:
These comparisons are regularly carried out during characterisation of the reference
batches (see graphs in pages 20 and 21).
Accreditation o f the verification organisations - The organisations responsible for

the verification of industrial pendulum impact testing machines conducted with reference
test pieces in accordance with standard EN 10 045-2 are subject also to accreditation
regulations aimed at bestowing recognition for quality performance and satisfying
prescriber requirements in terms of quality assurance standards.
Cofrac/Tests are responsible for accreditation based on document 1002 [3] which
supplements the requirements of standard EN 45 001.
Comparative Dispersion Analysis
Evaluation o f Unnoteched Test Pieces
The following graphs 1 to 6 illustrate changes in the mean values of characterisations

carried out by LNE in the form of energy value and dispersions.
Is
!o
IS
! I I I I I t I
0
Nay-80
Graph 1 - Level 25 J
0.0
0.7
Ql
~ 0.5
~1
~ @,4
~ 0,3
['~ 0.2
0,1
I I l I I , I I i
0 ....
NOV-G0 iv~-DI .ka~e! s4p-91 04o-4;I ~r-9~ ~ OcMl~
71.8
71.6
71,4
71.2
71
~" 70.8
-~ 70.6
~,~ 70,4
70.2
7o
69,8 I I I I I I I
Nov-90 Mat-91 Jun-91 Sep-91 Der Ar Jul-92 0ct-92 Jan-93
I.O
1o4
.m !
E
ti:
o.a
r
0.4
0.2'
0 I I I I I I I I
M~.gl .k~lil Sep-91 I)~.~1 4~r4)2 .hi.e2 0r Jen-1~3
153
152,5
152
151.5
150,5
149.5
! I I
149
NOV-90 Mar-91 Jun-91 Sep-91 Der A~r-92 Jul-92 Or Jan-93

2,5
.=_ 2
J
E
[~ 1.5
0,5
0 i i I i ! ! i !
Nov-90 Mar-91 Jun-91 Sep-91 Dec-91 Apr-92 Jul-92 O~-92 Jan-93
The results confirm the excellent reproducibility of the method: very low dispersion of
each batch characterised.
Evaluation of Charpy V Test Pieces
The following graphs 7 to 12 highlight changes in the various batches characterised.

9 : LNE. [] : CTA.
30
6
25
9
o
8 a
20
0 I I I ! i
0 1 2 3 4 5
2,5
~. 2
._o
=
> 1.5
D
O
0,5
O I ) ! !
1 2 3 4
Graphs 7 and 8 - Level 25 J
128 P E N D U L U M I M P A C T T E S T I N G : A C E N T U R Y OF P R O G R E S S
73
72
71
70
~ 89 o
~ 68
67
o o o
65 i i i i i
1 2 3 4 5
2.5
o~. 1.5
0
I=1
~ 0.5
0 i i i i i i
0 1 2 3 4 5 6
G r a p h 10 - Level 70 J
124
122
120
118
116
o
114
~J
112 o o
(3
110
108 I I I I I I
1 2 3 4 5 6
G r a p h 11 - Level 120 J
4,5
4 0 9
0
~3,5
0
9= 3 ID
II 9
2.5 o
~1,5
0,5
I I I I I
0 1 2 3 4 5 6
These graphs show:

- raised dispersion levels associated with the type of test piece,
- discrepancies between the two laboratories in terms of energy values at raised
levels.
Comparison
The following graphs 13 and 14 reflect the position of some french laboratories with
regard to reference energies (batches 1D (20 J) and 5D (120 J)).
EnerlD' (,D
29,11
I
23.11 Ii
I 84 !
I I [
17,11
il 'lJllllill~jiilli~J~l
l~ i~s iIIll fill
Graph 13
~zaeqD'O)
i6J i
Graph 14
Changes
lmproved Machining Processes
Evaluation of the various manufacturing processes used to produce V-notched

reference test pieces has shown that, if expertise were guaranteed overall, dispersion
criteria tolerated by standard EN 10045-2 being met, equipment mastery would be
ensured and efforts could be concentrated on the quality and, above all, regularity of
machining.
The Choice of Overall Expertise - LNE is a contractor specialising in the

manufacture of reference test pieces and called on sub-contractors for the machining of
these test pieces. As discussed before, these sub-contractors used different techniques9
Changes in the economic climate on the one hand, with risks affecting the financial
position of these companies, and the element of competition on the other hand, with the
wider distribution of BCR test pieces, have led LNE to review its manufacturing policy
for test pieces. Although the choice of steelmaker was inevitably given the quality of the
semi-finished products supplied, a decision was made by LNE to introduce in-house
machining. The aim was three-fold:
9to overcome problems associated with external dependence,
9to cut manufacturing costs,
9to improve notch machining quality.
The choice of Cutting by Electroerosion - Studies to determine the optimum process

led LNE towards cutting by electroerosion. The aim was to achieve the complete
machining of the test piece using this technique in order to avoid long and costly grinding
GALBAN ET AL. O N IMPACT TEST MACHINES 131
stages. Among the various technologies under consideration, the immersion process was
selected for its cutting performance and absence of corrosion.
The Results
Controls carried out during process qualification focused on:

9 surface condition,
9 notch profile,
9 decarburisation,
9 geometric tolerances (surface flatness, perpendicularity, etc.).
Surface Condition Was Checked f o r Roughness - The mean value of Ra (average

deviation from the mean line) obtained by this process was about 0.56, with a maximum
of 0.60. This conforms the LNE specification (Ra < 0.8).
Surface Flatness Was Checked as Follows - Example of flatness evolution on one

side (Figure 5).
Measurement points
Figure 5
Results - Notch profile, Figure 6.
Figure 6
Perpendicularity was checked using a profile projector (magnified 20 x). The

appearance of the notch was also controlled.
Extension to New Energy Levels
Production of a very high energy level (around 200 J) currently is under

investigation. This level will complete the current range and extend the verification
potential for industrial machines. Initial tests were carried out in 1997 leading to highly
satisfactory dispersions but with cases of non-breakage posing test validation problems in
terms of standard EN 10 045-1 (Table 12).
Table 12 - Characterisation Results
Machine Mean (J) Standard deviation (J) Dispersion (Emax-Emin) (J)

LNE 215.1 5.75 24.5
CTA 211.5 4.06 21.0
A&D 221.2 7.1 24.6
Studies are continuing with other grades of steel in an attempt to overcome this
problem.
Low Energy Machine Requirements
The problem of checking impact deflection machines, as presented during the 1970s,
nowadays applies to low-energy machines used to characterise plastic materials and
composites. These are essentially machines of a few joules (< 20 J), the functioning
principles of which are identical to those of other machines.
Studies should result in:
- the definition of an adapted reference machine,
- selection of a material and reference test piece.
Conclusions
France has had an indirect method for the verification of pendulum impact testing
machines for over twenty years.
The initial choice ofunnotched test pieces yielded very low dispersion levels
justifying the merits of this type of test piece. Improvements in production techniques and
machining are nowadays leading to dispersion levels with Charpy test pieces on a par with
previous levels. Financial aspects have, however, been considerably affected. The French
source is at the present time one of the four sources of certified test pieces in the world.
REFERENCES
[11 Revise, G., "Influence of a Dimensional Parameter of a Test Machine on the Results
of a Test", Charpy Impact Test - Factors and Variables: ASTM STP 1072 [1990].
[2] Galban, G., Le Muet, I., Mougin, D., Revise, G., Roche R., and Roesch, L.:
Presentation of the French Supply of Charpy V Reference Test Pieces - Pendulum
lmpact Machines, Procedures and Specimens for Verification. ASTM STP 1248
[1995].
[31 Doc. Reg. 1002 (F6vrier 1995) : Exigences/l satisfaire pour les laboratoires d'essais
ou d'6talonnage d'analyses accr6dit6s ou candidats ~ une accr6ditation et modalit6s
d'application.
Daniel P. Vigtiotti, ~Tom A. Siewert, 1and Chris N. McCowan ~
Maintaining the Accuracy of Charpy Impact Machines*
Reference: Vigliotti, D. R, Siewert, T. A., and McCowan, C. N., "Maintaining the

Accuracy of Charpy Impact Machines," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. R Manahan, St., Eds., American Society
for Testing and Materials, West Conshohocken, PA, 2000.
Abstract: The quality of the data developed by impact machines tends to degrade over
time, due to the effects of wear and vibration that are inherent in the test. This is the
reason that impact standards specify periodic direct and indirect verification tests. Each
year, we provide reference specimens for indirect verification of over 800 machines
around the world. From evaluation of the absorbed energies and the fractured specimens,
we are able to deduce the origin of energies that are outside the allowed ranges, and
report these observations back to the machine owners. This report summarizes the basis
for these observations and will allow, it is hoped, machines to be maintained at higher
levels of accuracy.
Keywords: absorbed energy, Charpy V-notch, impact test, machine repair,

misalignment, verification testing, worn anvils
Introduction
The low cost and simple configuration of the Charpy impact test have made it a
common requirement in codes for critical structures such as pressure vessels and bridges.
However, accurate results can be obtained only from impact machines that remain in
good working condition, such as within the tolerances specified by ASTM Standard Test
Methods for Notched Bar Impact Testing of Metallic Materials (E 23). We find that
many of the critical tolerances can be monitored by post-fracture examination of the
NIST verification specimens that we distribute to companies all around the world.
Our examination of over 2000 sets of these specimens each year allows us to identify
problems that are often not recognized during routine measurement of machine dimensions
or routine check procedures. We have learned to recognize what marks on the broken
1 Technician, Group Leader, and Metallurgist, respectively, Materials Reliability Division,

National Institute of Standards and Technology, Boulder, Colorado 80303
* Contribution of NIST; not subject to copyright.
134
Copyright9 by ASTMInternational www.astm.org
VIGLIO'rTI ET AL. ON MAINTAINING ACCURACY 135
verification specimens indicate factors that could be affecting the results. We can then
advise our customers to recheck or replace the anvils or the striker, tighten bolts, check
bearings, check machine alignment or level, check cooling bath or thermometer, or review
testing procedures. This paper describes the most common problems that we detect, and
gives advice on how to avoid or correct most of them.
Direct Evaluation
A routine check consists of a free swing check and a friction and windage check.
The free swing is a quick and simple test to determine if the dial or readout is performing
accurately. A proper zero reading after one swing from the latched position is required
on a machine that is equipped with a compensated dial. Some machines are equipped
with a non-compensated dial. Such a dial is one on which the indicator cannot be
adjusted to read zero after one free swing. The user should understand the procedure for
dealing with a non-compensated dial. This information should be available from the
manufacturer.
The friction and windage test will give the user the condition of the bearings. We
suggest that any deviation of more than 5% is excessive and the bearings should be
inspected.
We suggest that the user develop a daily log or shift log to be kept with the
machine. The log can be used to track the zero and friction values. The log can also
include information such as number of tests, materials tested, and any other useful
comments. A sample log is attached as Appendix 1.
Machine Preparation
The Charpy test is a dynamic test. Therefore, bolts may loosen over time. The
tightness should be checked on the anvil bolts, the striker bolts, and the baseplate bolts.
The manufacturer can supply the torque values for the anvil and striker bolts. The base-
plate bolts should be torqued to the recommended torque values for the grade and size of
the nuts and bolts. We recommend the use o f " J " or "T" bolts only. (See Appendix 2.)
We do not recommend lag type bolts. These bolts are made to withstand only static
loads. We believe that over time, the insert portion of lag bolts will loosen in the
concrete. As lag bolts are continually tightened, they can pull out of the concrete and be
tightened against the base of the machine, giving the impression of a properly mounted
machine. This condition is very difficult to detect. A machine with this problem will
cause high energy values at the low-energy level. The procedure used to mount the
master reference machines is attached as Appendix 2.
The anvil and striker radii should be carefully inspected for proper dimensions
and for damage. Damage can be detected easily with a visual inspection and a check for
smoothness by nmning a finger over the radii. We find that radius gages are usually
inadequate to measure the critical radii. We recommend making molds of the radii and
measuring the molds on an optical comparator. Occasionally even a new set of anvils
and striker may have incorrect radii. We recommend that new anvils and strikers be
inspected before being installed in the machine. Since the radii can be considered
consistent before use, they can be measured directly on an optical comparator or other
optical measurement system.
We recommend centering tongs such as those described in ASTM Standard E 23.
The tongs should be inspected for wear or damage. A proper set of tongs is critical for
the accurate placement of the specimen. Some machines are equipped with a centering
device. The device should be inspected for wear and proper operation. We do not
recommend the use of centering devices for low temperature testing because it can delay
the time between removal from the bath and fracture, and so may exceed the allowed
five-second interval.
The temperature indicator should be calibrated immediately before testing.
Since ice water and dry ice have constant temperatures, they make quick and easy
calibration media.
Post-Fracture Examination
Many machine problems can be monitored by post-fracture examination of the

NIST standardized verification specimens. Following are the most common of these
problems. In many cases, suggestions on how to correct or avoid them in the future are
included.
Worn Anvils - Most of the wear of an impact test machine occurs on the anvils and
striker. We evaluate this wear by examining the gouge marks that are formed on the
sides of high-energy specimens when they are forced through the anvils. Anvils that are
within the required tolerance of the standard will make a thin, even gouge mark all the
way across both pieces of the broken specimen. As the anvils wear, they will make a
wider, smeared mark across the specimen halves. Figure 1 shows the change in the
gouge marks. When the wider smeared marks are observed on a customer's specimens,
we recommend that the anvils be changed, because the reduction in energy needed to
push the specimens through wom anvils eventually drops the machine below the lower
tolerance in the energy range. You can monitor the wear on your machine by retaining
some specimens that are tested with new anvils and comparing them to specimens of
Figure 1
VIGLIO-I-I-I ET AL. ON MAINTAINING ACCURACY 137
similar composition and hardness that are tested as the anvils wear. For specimens at a
similar absorbed energy, the gouge marks will grow wider and smoother as the anvils
wear.
Off-Center Specimen - An off-center specimen strike occurs when a specimen is
not centered against the anvils, so the striker contacts the specimen to the side o f the
notch. The low-energy specimen best indicates when an off-center strike occurs. We
identify this condition on the specimens by finding that the gouge marks caused by the
anvils are not equidistant from the machined notch edges, and the striker gouge mark is
offset the same amount from the notch (Figure 2). Also, as seen in Figure 2, the fracture
surface of a correctly tested low energy specimen is flat and both halves are even.
However, the fracture surfaces of a specimen that has been tested off-center are on an
angle. The more off-center the strike, the steeper the angle will be. This problem
increases the energy needed to fracture a specimen. The most common causes for this
slipping are worn or damaged centering tongs, a worn or misaligned machine centering
device, careless test procedures, or the use of a cooling fluid that is too viscous at the test
temperature, which causes the specimen to float on the specimen supports. Most machine
manufacturers should be able to provide new centering tongs. We have found that ethyl
alcohol is one of the best cooling media because it seems to evaporate quickly from the
bottom o f the specimen to prevent specimen floating.
Figure2
O f f - C e n t e r S t r i k e r - This differs from the off-center specimen in that the

specimen is centered against the anvils so the anvil gouge marks are equidistant from the
machined notch edges. However, the striker does not contact the specimen precisely
opposite the notch. Figure 3 shows this appearance. An off-center striker is usually
attributed to the pendulum shaft shifting off center. This shift can be the result of a loose
alignment ring on the shaft or a loose bearing block on the machine. This problem also
increases the energy needed to fracture specimens at all energy levels.
Figure3
U n e v e n A n v i l M a r k s - Frequent testing of subsize specimens can cause the anvils

to wear unevenly. Figure 4 shows an example of these uneven wear marks at each energy
level of our reference specimens. Since this wear is restricted to a small area that the full-
size reference specimen contacts, there is usually no effect on the energy required to
fracture the specimen. This anvil condition presents two problems. First, since subsize
wear is usually not indicated by a change in the energy required to break a reference
specimen, inspection of the broken specimen is required. This wear will cause the anvils
to be out of tolerance according to the requirements in the standard. This means that the
machine does not meet the direct verification requirements of the standard and is
therefore not eligible for the indirect verification process. The second, and more
important problem, is that the subsize specimens are being tested in an area of the anvil
that is worn. When the wear is substantial, this condition will produce artificially low
sub-size energy values. The anvils should be replaced on a machine with this condition.
VIGLIOTTI ET AL. ON MAINTAINING ACCURACY 139
l~igure 4
Chipped Anvils - Sometimes an anvil can be chipped. Figure 5 shows that this
condition can be detected easily on all three energy reference specimens. The low-energy
specimen is affected the least amount because it is the hardest specimen and therefore has
a more brittle fracture. The ductile high-energy specimen will produce higher than
normal energy results and the very ductile super-high-energy specimens are affected most
by a chipped anvil. This condition should be detected easily by a visual inspection
before using the machine. New anvils are required when an anvil is chipped.
Figure 5
Anvil Relief- Some Charpy machine manufacturers have designed a machined

relief at the bottom of the anvil (Figure 6). This anvil design does not meet the direct
verification requirements of ASTM Standard E 23. The relief has caused high-energy
results in our ductile high and super-high-energy specimens. It can also cause twisting of
the specimens, during fracture, that may also contribute to energy values higher than
normal at all energy levels. Since the relief is designed into the anvils and does not
appear to add an excessive amount of energy to the test, we at NIST continue to verify
these machines.
Figure 6
Damaged Anvils - Under some test conditions, usually for elevated temperature
testing, the anvils can wear to a rough finish that creates excessive friction (Figure 7).
This damaged condition is detected best on the high and super-high specimens. Damaged
anvils usually cause the gouge marks to become wider and push the specimen material to
form a ridge that can easily be detected with the fingernail. This damage usually causes
artificially high energy results at the high and super-high energy levels. Damaged anvils
must be replaced.
Figure 7
VIGLIO'I-I-I ET AL. ON MAINTAINING ACCURACY 141
B e n t P e n d u l u m - Figure 8 shows the gouge

marks created by a pendulum bent in the direction of
the swing. This gouge mark is usually deeper on the
top edge of the specimen as it sits in the machine.
As shown in Figure 9, the striker contacts the top
edge of the specimen first, causing excessive
tumbling and twisting. This excessive activity can
cause the specimen to interact with the striker or the
pendulum after fracture to create additional energy
loss. A bent pendulum can be detected by placing an
unbroken reference specimen in the machine and
placing a piece of carbon paper on the surface opposite
the notch. At this point, move the pendulum about Figure 8
one inch away from the specimen and release it so that it contacts the specimen only
once. This will make a mark on the specimen that can be inspected. If the pendulum is
not bent, the mark should appear the same width across the specimen. If the pendulum is
bent, the mark will be wider at one edge and become thinner or even not visible at the
other edge (Figure 8). We recommend that a new pendulum be installed on a machine
with this problem.
Figure 9
Summary
The condition and accuracy of Charpy machines cannot be checked only by

comparing results of NIST reference specimens to the Master Reference Machines
located at NIST, Boulder, CO. Some machine problems cause artificially low results
while other machine problems cause artificially high results. In addition, deviations in
procedures can cause similar results. These machine problems and procedural deviations
may go undetected for years without some sort of physical check. For this reason,
examination of the broken specimens is a critical part of the verification process. Many
machine problems can be avoided or corrected with the information presented in this
paper. Also, suggested changes in procedure can help can help to insure a successful test.
To obtain verification specimens or to clarify procedures for verification testing, you may
use the following information:
Verification specimens can be ordered from the NIST Standard Reference

Materials Program. Phone: (303) 497-6776 Fax: (303) 948-3730 email:
[email protected]
Questions on verification procedures can be answered by the Charpy Program

Coordinator. Phone: (303) 497-3351, fax: (303) 497-5939, or email:
[email protected]
VIGLIO-I-FI ET AL. ON MAINTAINING ACCURACY 143
APPENDIX 1
EXAMPLE LOG
DATE FREE SWING FRICTION VALUE COMMENTS

VALUE
IDIGITAL (J] DIAL (ft-lbf): DIGITAL (J)
I
DIAL (ft-lbf) !
I I
I I
I I
I I
I I
I I
J I
I I
I !
I I
I I
I !
I I
I I
I I
I I
I I
APPENDIX 2
MOUNTING PROCEDURE FOR REFERENCE CVN MACHINES
This is a detailed procedure developed by NIST to mount the three Master Charpy
Reference Machines. This procedure is not intended to be substituted for any installation
procedure provided by the manufacturer of the machine.
The foundation o f the impact machine is critical to insure accurate results. Energy losses
through the foundation must be kept to a minimum. We recommend making a
foundation of 7000 pound mix concrete that measures 152.4 cm (60 in.) long by 91.4 cm
(36 in.) wide by 45.7 cm (18in.) thick. Usually you will need to cut a hole in the floor to
accommodate the new foundation. If other equipment in the area could affect the
machine operation, you should isolate it from the floor with expansion-joint material.
Hold-down bolts used to secure the machine to the foundation should be o f the inverted
"T" or "J" type. The bolts, nuts, and washers should have a strength o f grade 8 or higher.
We recommend using bolts with a diameter of 22 mm (7/8 in.). At NIST we used 22 mm
(7/8 in.) grade 8 threaded rod, cut into pieces 61 cm (24 in.) long. We then welded 22
mm (7/8 in.) pieces o f the same threaded rod, six inches long, to the end of the 61 cm (24
in.) pieces to make inverted T bolts.
We then positioned the machine over the center o f the foundation hole. The machine was
held approximately 10.2 cm (4 in.) above the floor using spacers suitable to hold the
weight of the machine. The T bolts were positioned in the machine-base mounting holes
with a nut below and above the base o f the machine. The nuts were tightened to keep the
T bolts straight while the concrete was poured. The ends o f the T bolts were positioned
approximately 2.5 cm (1 in.) from the bottom of the hole. The machine was then leveled
on the spacers. Leveling did not need to be as accurate as the final leveling.
Reinforcement bars were attached to the top of the horizontal rod previously welded to
the bottom of the T bolts. The reinforcement bars were attached in the form of a box
connecting the four bolts. Another box formation o f reinforcement rods was attached to
the T bolts 25 cm (10 in.) above these rods. The concrete was then poured under the
machine. The concrete was finished as level as possible at this time. Before the concrete
fully hardened, we removed concrete from around each T bolt to create a cavity o f
approximately 2.5 cm (1 in.). This cavity would enable a nut to be threaded below the
surface of the concrete. The machine was left in this position for 72 hours.
After 72 hours, the nuts on top of the base plate were removed and the machine was lifted
offthe T bolts. The bottom nuts were then threaded down into the cavities created before
the concrete hardened. The nuts were left high enough on the T bolts to enable the use of
an open-end wrench to adjust them after the machine was positioned on them. At this
point, the base o f the machine was coated with a light oil to keep grout from adhering to
it. The machine was then lifted onto the T bolts and was positioned on the adjustment
nuts. The machine was now ready to level. A machinist's level was used to insure
VIGLIO-I-FI ET AL. ON MAINTAINING ACCURACY 145
meeting the tolerance of 3:1000 in: The critical leveling procedure was done using the
four nuts under the machine. After the machine was leveled, we wrapped the outside of
the nuts with duct-seal putty to facilitate their removal from the T bars later in the
process.
At this point, the base of the machine was ready to grout. Heavy cardboard forms were
placed around the base of the machine to keep the grout under the machine. It was
necessary for the grout to flow completely under the machine, making sure the base of the
machine was in total contact with the grout. The grout was installed under the machine.
The machine was left in this position for 72 hours.
After 72 hours, the machine was lifted offthe T bolts. The grout was inspected for
cavities and for surface contact with the bottom of the machine. The putty was removed
form around the nuts. The nuts were removed from the T bolts. After removing all
debris from the grout, the machine was lifted over the T bolts and rested on the grout.
Washers and nuts were installed and tightened. The level was checked at this point. The
T bolts were cut offto approximately 12.7 mm (1/2 in.) above the nuts. The nuts were
torqued to 380 ft-lb. The final level was checked at this point.
NOTE:
Special non-shrinking grout is recommended. This grout is available at most industrial
hardware stores.
If you have any questions concerning this procedure, please contact Daniel Vigliotti by
phone at (303) 497-3351, by fax at (303) 497-5939, by email at
[email protected], or by mail at NIST, Division 853,325 Broadway, Boulder, CO
80303-3328.
Enrico Lucon, 1 Rachid Chaouadi, 1 Albert Fabry, 1 Jean-Louis Puzzolante I and Eric Van
Walle I
Characterizing Material Properties by the Use of Full-Size and Subsize Charpy

Tests: An Overview of Different Correlation Procedures
Reference: Lucon, E., Chaouadi, R., Fabry, A., Puzzolante, J.-L, and Van Walle, E.,
"Characterizing Material Properties by the Use of Full-Size and Subsize Charpy
Tests: An Overview of Different Correlation Procedures," Pendulum Impact Testing: A
Century of Progress, STP 1380, T. A. Siewert and M. E Manahan, Sr., Eds., American
Abstract: A total of 565 instrumented impact tests (232 performed on full-size and 333
on subsize Charpy-V specimens) have been analyzed in order to derive meaningful
assumptions on the correlations existing between test results obtained on specimens of
different size. Nine materials (pressure vessel steels) have been considered, in both the as-
received and irradiated state, for a total of 19 conditions examined. For the analysis of
data, conventional as well as novel approaches have been investigated; the former ones,
based on a review of the existing literature, include predictions of upper-shelf energy
(USE) values by the use of normalization factors (NF), shifts of index temperatures
related to energy/lateral expansion/shear fracture levels, and a combination of both
approaches (scaling+shifting of energy curves). More original and recent proposals have
also been verified, available in the literature but also proposed by SCKoCEN in the frame
of enhanced surveillance of nuclear reactor pressure vessels. Conclusions are drawn
regarding the applicability and reliability of these methodologies.
Keywords: Charpy V-notch, instrumented impact testing, subsize specimens,

correlations between different specimen sizes.
Current reactor pressure vessel surveillance practice relies heavily on the Charpy V-
notch (CVN) impact test, to which actual fracture toughness bounds are indexed. A
common feature of such surveillance tests is the unavailability of large quantities of
irradiated material, which may derive from the necessity to operate on broken halves of
previously tested specimens or from the limited space normally available for Charpy V-
notch specimens in irradiation facilities; the same background problem holds true in the
case of integrity assessments and residual life predictions of service-exposed plant
1 SCKoCEN, Boeretang 200, B-2400 Mol, Belgium.
146
LUCON ET AL. ON FULL-SIZE AND SUBSIZE CHARPY TESTS 147
components [1], which have to be maintained in operation and therefore cannot be

destructively sampled.
All this makes the use of subsize impact Charpy V-notch specimens (MCVN) very
appealing, in view of the valuable information which can be extracted from very limited
amounts of material, provided effective test procedures and reliable correlations with full-
size specimen data are available.
Instrumented impact tests have been carried out in the past years in SCKeCEN, at
temperatures spanning the range from lower shelf to upper shelf behavior, in order to
characterise the impact properties of 9 different pressure vessel steels, using both full-size
and subsize Charpy V-notch specimens. Some of the steels have been characterised in the
as-received (unirradiated) state and under different conditions of irradiation, for a total of
19 conditions examined.
The subsize specimen geometry currently used in SCKeCEN is the Kleinstprobe
mentioned by the German Standard DIN 50 115 "P~fung Metallischer Werkstoffe -
Kerbschlagbieversuch" (Figure 1). This is the reference geometry for the standardisation
efforts currently under way with the ESIS Technical Subcommittee 5 "Dynamic Testing
at Intermediate Strain Rates" [2, 3].
60o_+2 ~
V ]~ 1 _+0.1
/ 4-+0.1
R=0.1 _+0.025
27_+0.6 3__.0.1
1 - Geometry of the subsize Charpy V-notch specimen (dimensions m mm) used for
Figure
SCKoCEN tests.
Adjustment of Test Temperature for the Subsize Specimen Tests
The problem of temperature control in performing impact tests on subsize

specimens has always been a major concern for experimentalists, due to the minimal size
of the test piece which can cause considerable thermal loss in the few seconds that elapse
between removal from the conditioning medium and impact by the striker. The limit of
five seconds which ASTM Standard Test Methods for Notched Bar Impact Testing of
Metallic Materials (E 23-96) requires for standard full-size specimens is evidently
inadequate in this case.
The availability of two subsize specimen data sets, tested partly by VTT (Espoo,
Finland) and partly by SCKeCEN for two of the materials examined, has allowed
analysis of this problem in order to find an expression for the proposed temperature
adjustment; the analysis was based on the fact that VTT uses a pneumatic transfer system
of the subsize specimens which allows testing in less than 1 second after removal from
the cooling/heating medium. Therefore, test temperatures reported by VTT can be
considered, from a practical point of view, as "true" values as compared with the nominal
SCK "furnace temperatures" (which are subsequently altered by the 2 + 3 seconds
required for specimen transfer).
The following analytical expression for the temperature adjustment was obtained by
fitting temperature shifts read at the same USE percentage on VTT and SCK curves, in
the range between lower and upper shelf:
T,,~e =Tj, .... -0.1119-(T~,,~ac ~ -14.79) (1)
All test temperatures related to instrumented tests on subsize specimens performed

by SCK have therefore been preliminarly adjusted to their "true" value before carrying
out the analyses subsequently performed.
Overview of the Analyses Performed
Different approaches to the problem of correlating subsize with full-size specimen

data have been considered, with the aim of establishing a reliable procedure to be used for
extracting significant information from instrumented tests on non-standardized specimen
geometries. Correlation methodologies found in literature can be grouped in the following
two categories:
1. Methods based on the normalization of absorbed energies: these have been proposed
mainly with the aim of estimating the USE (Upper Shelf Energy) of full-size (fs)
specimens from subsize (ss) data, using a relationship having the form:
USE.f~ = N F xUSE~.~ (2)
where NF is a normalization factor, which may be based on the fracture area [4, 5], the
nominal fracture volume [4-7] or some other expression containing elastic [8] and/or
plastic [9] stress concentration factors.
Another USE normalization procedure, based on iterative calculations, has been
proposed by VTT [10] and has also beeen checked against experimental data.
2. Methods based on the shifting of index temperatures: the values of transition
temperatures or DBTT (ductile-to-brittle transition temperatures), as determined on
subsize specimens, are empirically correlated with corresponding temperatures for
full-size test pieces by the use of simple relationships having the form:
DBTT,:,. = DBTT,~, + M (3)
where the temperature shift M is determined empirically from the analysis of

nmnerous experimental data [11-14], and DBTT (index temperatures) can have the
following expressions: T41J (corresponding to T1.9J for subsize specimens), T68J (T3.1J),
T28J (T3.15J), T0.89mm(T0.3mm)and FATTs0 (50% shear fracture appearance, for both
specimen types). Such equivalencies are based on the ratio of upper shelf values.
Moreover, ORNL has recently proposed [14] a general procedure normalizing

subsize to full-size specimen data, in terms of absorbed energy and temperature; this
method includes both steps outlined above:
a. normalization of energy values, using a NF which is proportional (through the
measured SFA value) to the amount of ductile fracture in the test;
b. shifting of test temperatures based on the value of M empirically determined.
3. Analysis of the instrumented trace (load diagram approach): the aim of this, as such,
is not to look at the conventional impact parameters, but:
9 to convert forces into dynamic strength values for the two test configurations;
9 to compare the shifts measured on the two specimen geometries for the
temperatures NDT (Nil Ductility Temperature), TI and To, the latter two
respectively corresponding to the onset of ductile behavior (SFA>0) and the onset
of fully ductile behavior (SFA=100%);
9 to compare SFA values measured on the fracture surface with those calculated
using characteristic forces from the instrumented force/displacement diagram.
Correlation Methods Based on the Sealing of Absorbed Energies
Prediction of Full-Size USE by the Use of Normalization Factors
The correlation method most widely used in Europe consists in establishing an

empirical ratio of the full-size USE (USEfs) to the subsize USE (USEss) based on a large
number of tests. Another approach, quite common in North America and Japan, tries to
correlate USE values with the ratio of various geometrical parameters; this ratio can be
expressed in terms of a normalization factor (NF).
The most commonly used expressions for NF which can be found in the literature,
are the following:
(Bb) f~
NF~ - (4)
(~b) ~
(Based on the ratio of fracture areas, with B = specimen thickness and b = specimen
ligament depth - Corwin et al. [4, 5].)
[(Bb)3/2]t,'
NF2 = [(Bb)3/2],.~. (5)
(Based on the ratio of nominal fracture volumes - Corwin et al. [4,5].)

( Bb 2)f;
NF3 - (Bb 2),,~,. (6)
(Based on a different expression for the ratio of nominal fracture volumes - Lucas et al.
[6, 7].)
Bb2 I
~ , )f,, (7)
(With L = span and Kt = elastic stress concentration factor - Louden et al. [8].)
Considering the ratio of geometrical parameters between standard 10x10x55 and
subsize DIN 50 115 Charpy-V specimens, one gets the following numerical values for
NF:
NF 1 = 8.9
NF 2 = 23.7
NF 3 = 26.5
NF 4 = 13
In addition, Sokolov and Alexander [14] have determined the following empirical
value for NF by averaging the values of USEfs/USEss obtained on 10 different materials
(mostly pressure vessel steels with different heat treatments):
NF5 = 21.3
Considering the nine materials (and 15 conditions) examined in the present work,
the average value of the normalization factor is:
NF 6 = 21.56
with standard deviation ~ = 2.64. The average calculated value ('NF6) is in reasonable
agreement with NF2 (nominal fracture volume, eq. 5) and very good agreement with NF 5
(ORNL), but the scatter of individual data points is nevertheless very pronounced.
Actual predictions of USErs obtained using the six different normalization factors,
are plotted in (Figure 2), with reference to the ideal prediction line (predicted USErs =
measured USErs) and +20% tolerance bounds.
The analyses performed have shown that:
1. the approaches which use normalization factors NF1, NF3 and NF 4 yield unsatisfactory
predictions for the materials considered;
2. the use of normalization factors NF2, NF5 and NF 6 provides reasonably small average
(absolute) deviations from measured values, in the range 10% + 15%, the smallest
value being given by NF6;
3. nevertheless, the maximurn absolute deviation for any of the different normalization
factors is always higher than 30%; therefore, irrespective of the formulation chosen,
the presence of "outlier" materials cannot be avoided.
250
[] USEfs,NF1
r ss ~
zx USEfs,NF2
....6 o/6 ~
A 200 o USEfs,NF3
+ USEfs,NF4 A s"ts A AO..~'"'""
[u
t~ • USEfs,NF5
o USEfs,NF6
.~ 150 +20% .--" A,.-" X~ ..A""
o :oJ" ,,I"~ ............ X~
o oo -zk'" A.," ..-
o zx ~ - " ~ -'~" .........
"0 100 " ;-~.-
x...6 ~ - " - 2......
0%-" ++++ + + ++
o". ......
,~-' s "
.~"...-" : ..... +
§ § 12 H H
9 ... []
a. 50 9"" ~-7 ..-l-"" ++ + [] ~ []
- " J " ..-" + [] O
- - ; " .--"+o o oo
9 i i i i I i i i i I i i , , [ , , i ,
0 50 100 150 200 250

Measured Full-Size USE (J)
Figure 2 - Comparison between predicted and measured full-size USE values, obtained
by the use of different normalization factors
Based on the calculations performed, the simple approach of using a normalization

factor for predicting USErs is therefore not recommended, due to the risk of seriously
under- or over-predicting the actual values; using a simple fitting function seems, at the
present state, a safer and more convenient solution.
Prediction of Full-Size USE Using the VTT J-lntegral Approach
A procedure to evaluate upper shelf energies for standard Charpy-V specimens

from subsize specimen data has been proposed by VTT [10,15]. This is based on the
combination of the defining formula for the J-iutegral and the power-law fit J = ( A a ) m
which describes the stable crack growth behavior; the following expression has been
suggested:
10 ( 8 ~ l+m
user,. =usE,,,.-~.-~) (8)
where factor m depends on specimen size, notch dimensions, and material. An

appropriate function was fitted to the measured USE values of different materials in order
to define the unknown factor m in eq. 8 as follows:
m:I l (9)
where the fitting parameters Z and n were found to have the estimated values of 184 J and
0.34, respectively [10].
Using the USE measured on subsize specimens, hence, the predicted value for
standard Charpy-V can be calculated by solving iteratively eqs. 8 and 9.
The results of the application of the VTT iterative procedure to the materials
considered in this paper is given in (Figure 3).
250
.'" sS SS
o" SS
o" Ss
A
200 -" ~/( ....'~ "
u,,I o-
.N 150 .o *''" SS S .." "~ ~176

." -"
jSSSSs~( ''~
~ ....-'* ~ ~(
~" S S ."
" 100
.oo ~
~9 50 VTT method
o'S ~176
,~.'4~":" ....... + 2 0%
i i i r I L i i i I i i r r i i i I I I i i t
0 50 100 150 200 250

Measured Full-Size USE (J)
Figure 3 - Comparison between predicted and measured full-size USE values, obtained
by the use of VTT J-integral approach
In agreement with the findings reported by VTT in [10], all but two of the materials
considered fall within the • tolerance band (some of them very close to the 1:1 line).
Moreover, the average absolute deviation of the predicted values is less than 10%, thus
supporting the use of this approach as an alternative to methods based on empirical
normalization factors.
Correlation Methods Based on the Shifting of Characteristic Index Temperatures
Various definitions of index temperatures (or transition temperatures) have been

introduced for conventional Charpy-V impact tests to be used in regulatory codes or
specification criteria. Each one of them is related to a predefined level of absorbed energy
(41 J or 68 J), lateral expansion (0.89 mm) or shear fracture appearance (50%, FATTs0 =
Fracture Appearance Transition Temperature).
For the DIN 50 115 mm subsize specimen, the criteria based on the energy and
lateral expansio n levels above mentioned have been redefined as:
41J--+ 1.9 J
68 J --~ 3.1J
0.89 mm + 0.3 mm
This correspondence was based on the mean ratio of upper shelf values [16,17]. The
conventional SFA level (50%), being a relative measurement, does not need redefinition.
The mean difference in the values of index temperatures between subsize and full-
size specimens has been established on the basis of numerous correlation tests [10,16,17]:
Tj., =Z,.~ +65~ (10)
where: Tes = T41J, T68J, To.89mm, FATTs0(fs) and Tss = T1.9J , T3.1J , To.3mm, FATTs0(ss). The
standard deviation is 15 ~ giving a 95% confidence interval (+2~y) o f + 30 ~
This simple and widely popular approach has been checked using the transition
curves determined on the 19 materials here considered for standard and subsize
specimens. Figure 4 shows calculated values of index temperatures for full-size samples
with respect to the corresponding equivalent temperatures for subsize specimens.
The mean correlation line calculated from the analysed data, imposing unity for the
slope, is the following:
T~. = T,.,. + 59.6~ (11)
with a 95% confidence interval of+43.3 ~ (standard deviation ~y= 21.7 ~

If the data are fitted with an ordinary straight line (i.e. without imposing parallelism
with the 1:1 line), the result is the following:
T~. =1.17. T~:~.+57.8 (12)
which gives a lower value for the mean quadratic residual (360.1) than eq. 11 (462.1).
288
D 41J/1.9J ..... " ~ f.,,'"
o,.. 150 +8=J,=lJ I T==T"+598"cl § ..-" .... o

5. \ ..... .....
E
x 180 o FATT50
"0
C ..... ....-
< 58
14.
O
0 .'"" " rn ..,""
E
E
~"
O
-50
T-
- 1O0
- 158 ~ i i i,'i' , i i i , , i , i , , , , i i i , , i i , i i i i , i i i i i , i
-1 O0 -50 8 50 100 150 200 258 380

41JI68JlO.9mmlSO%SFA i n d e x t e m p e r a t u r e s (~
Figure 4 - Comparison between index temperatures measured on full-size specimens and

on subsize specimens, with fitting lines
Although the average shift (59.6 ~ is reasonably close to the value reported in the
literature (65 ~ index temperatures related to energy levels (41 and 68 J) show a
remarkable amount of scatter; on the other hand, the standard deviation measured for
AT0.89/0.3 and AFATTs0 is reasonably consistent with the values reported in the literature
(15 ~ The use of index temperatures related to energy levels, therefore, is not
recommended for correlation purposes.
It has to be noted, moreover, that for some irradiated materials, some of the
predefined energy/lateral expansion levels cannot even be reached, due to severe
toughness degradation.
Normalization of the Entire Energy Transition Curve: Scaling + Shifting
ORNL has recently proposed [14] a normalization procedure for the curves of
absorbed energy versus temperature, which involves both steps that have been outlined in
the previous chapters:
(a) normalization of impact energies obtained from subsize Charpy-V specimens by
means of the following equation:
E# rNFh,': 1O0- SFA%

L SFA% 7 (13)
"• 1O0 § NFa~c"~- ] ~
where NFbrittl e is the fracture area (Bb) normalization factor (eq. 4), which has been
proven to work satisfactorily in conditions of prevailing brittle fracture (lower shelf),
and NFductil e is an empirical normalization factor, corresponding to the ratio of upper
shelf energies, which for the DIN 50 115 subsize specimen is equal to 21.3;
(b) test temperatures are then shifted forward by an amount which has been empirically
determined for different geometries, the value for DIN 50 115 being equal to 38 ~
A "personalized" version of this procedure has also been applied to subsize specimen
data analysed in this paper, by attributing the following values to the normalization
factors and the temperature shift in eq. 13:
NFbrittle = 8.9
(as in ORNL procedure),
NFductile = 22.56
(average value ofNFexp) , and
AT = 61.8 ~
(average index temperature shift considering only temperatures related to energy levels).
The results obtained vary strongly from material to material, since the
representation of the full-size transitional/upper shelf behavior is obviously affected by
how close the individual correlation parameters (DBTT shift and USE ratio) are to the
average values computed on the 19 material conditions. The general concept introduced
by the ORNL approach (developing an equivalent full-size curve from subsize data
through scaling and shifting) remains certainly valid: this approach could be used to
derive sufficiently reliable representations of full-size energy curves, based on subsize
specimen data.
Analysis of the Instrumented Test Trace (Load Diagram Approach)
Determination of Characteristic Temperatures
The load diagram, which yields characteristic force values (force at general yield,
maximum force, force at brittle fracture and crack arrest) as a function of test
temperature, has been demonstrated to be one of the most fundamental features of the
Charpy V-notch impact test; it is moreover directly correlated to the appearance of the
fracture surface (SFA).
In upper shelf conditions, the most significant fraction of absorbed energy and
lateral expansion derives from plastic deformation, associated with ductile crack growth
under conditions which do not represent effectively the constraint and stress-strain field
experienced by the tip of a sharp crack in a real structure, such as a pressure vessel steel.
Consequently, it has been extensively demonstrated [19-21] that the temperature at

which a fixed level of energy is absorbed during the test (such as 41 J or 68 J) does not in
all cases accurately trace with acceptable accuracy the effects of service exposure (e.g.
irradiation) on the ductile-to-brittle transition temperature and the cleavage fracture
toughness. It has been therefore contended that this can be achieved more reliably by
using other characteristic temperatures obtained from the analysis of the load diagram.
This latter statement has been verified within the present work, from the point of
view of the correlations between standard and subsize impact specimens; in other words,
the aim has been establishing whether the shifts between full-size and subsize specimens
for these characteristic temperatures were more consistent than in the case of
conventional index temperatures previously considered (T41J/1.9J , T68J/3.1J , Zo.89mm/o.3mm ,
T28J/3.15J, FATTs0)-
The characteristic temperatures which have been determined by analysing the
instrumented data of all the materials under examination are the following:
1. T1 (ductile crack initiation temperature): it marks the onset of actual ductility in the
test, corresponding to the temperature at which the shear fracture appearance begins to
exceed 0%, or at which force at general yield and maximum force start to divert.
2. T o (upper shelf onset temperature): it marks the onset of CVN upper shelf,
corresponding to the temperature at which shear fracture appearance reaches the 100%
value, or at which no more brittle propagation can be spotted on the test diagram.
3. NDT (nil ductility temperature): originally proposed by Pellini in 1953, it is defined as
the highest temperature at which a crack, initiated by impact of a test piece containing
a small flaw, propagates under purely elastic conditions. The Drop Weight (Pellini)
test was developed and standardized (ASTM E208) for allowing tests on relatively
small-scale specimens. An alternative procedure for obtaining NDT estimates from
instrumented impact tests (crack arrest loads, Fa) has been proposed [21], based on the
definition of NDT as the temperature below which the arrest load threshold is zero.
Experimental NDT values have been verified to correspond to temperatures at median
arrest load of 3 kN in case of full-size specimens [21]; the equivalent value for subsize
specimens was calculated using a scaling factor of 14.14 [22], thus setting the value Fa
(subsize) = 0.21 kN.
Arrest loads, for all materials considered, have been fitted using a "master curve"
formulation [21], expressed by the equation
F.(T) = Q. e '4r-Nm') (14)
where Q is the threshold value (3 for full-size and 0.21 for mini) and
ot = O.026for T > N D T
ot = O.020for T < N D T
NDT values were therefore calculated by minimising the sum of square residuals
between the fitting curve (14) and the experimental data, excluding points where Fa =
0 (master curve is too flat and confidence interval too large) and which showed
excessive deviation from the master curve shape (outliers).
As far as these "alternative" characteristic temperatures are concerned, the results of
the analyses have not been particularly favourable, due to the following reasons:
a) The quality (i.e. the intrinsic uncertainty) of the instrumented data, namely the
characteristic values of force, displacement and energy, is absolutely crucial for the
reliability and consistency in the determination of characteristic temperatures,
especially TI and T 0.
b) As a consequence, highly scattered data sets, as often is the case for subsize
specimens, can seriously hinder the determination of these temperatures.
c) In some other instances, the tests performed have not allowed a clear identification of
the behavior in the lower shelf region (thus affecting the determination of Tl and NDT)
or in the upper shelf region (To).
d) The choice of the specific function for fitting experimental values can also affect the
measured values; even changing the degree of a polynomial fit can produce a
significant difference in the resulting temperature (of the order of 10 ~ or more). An
effort has been made to use the same function when elaborating the same quantities
from different materials, but in some instances the scatter of results and/or the scarcity
of data has made the use of "simpler" equations unavoidable (e.g. a linear straight line
instead of a parabolic or third-order curve).
Comparison Between Calculated and Measured Values of Shear Fracture Appearance
It has been verified through magnetic emission detection of crack initiation [20]
that ductile crack initiation in an impact test occurs at a load F i such that
e=F (15)
where, generally, k ~ 0.5, i. e. the initiation load (Fi) is roughly equal to the average
between Fgy (load at general yield) and F m (maximum load).
The implication of eq. 15 is therefore that the shear fracture appearance of Charpy-
V impact specimens can be calculated from the instrumented trace using the following
relationship, where Flu is the load corresponding to brittle crack propagation:
F,-Fo ] •
SFA%= 1 F., +k(F,.-F~) (16)
Similar equations have been proposed in the literature and even in draft standards
[24, 25], but they possibly lack the physical background that led to the derivation of eq.
16; basically the same equations are currently being considered in other draft standards
relevant to instrumented testing of miniaturised specimens [2, 26].
The reliability of eq. 16 has been verified for all materials considered, by direct
comparison of the calculated SFA values with values directly measured on the specimens'
fracture surfaces. The results are given in Figure 5 for full-size specimens and Figure 6
for subsize specimens, with straight lines referring to two accuracy levels (4-20% and
4-30%); the first level (4-20%) is the order of approximation currently reported by the
various draft standards [2, 24-26] for a l l the equations proposed.
100 ," ," o o,'

,/ 9 o o/
gO .... *20% '" "g o /o

,/ j ~ o o o o~o
...... +30% , ' oo o oo/,,""
80 .~ , , o o o~ o ~
,' , oOZe o ,"
/ 9 o o
70 ,-" ,, Io ,
,/" 99 O0 ~ 0 9 9 /
.,. , oOo~oo ,, ..
60 ,' , " a~'o o ,, ..
/ ," o o ~b o o ," /,'
50 / ,,o oo o ~ oo 0% /.
/ , ~o o 9 /'
40 .... ,'o o o %/0 " .."/

/, 9 OoOf o o 9 ,
~1 , , o 9 /"
30 o ~ ~ ' :
, OO%o~oOe o , ..,
20 ~176o ~/o~ ,'' ."""
10 ~b= /
~@ o ,, ,' . , , , ."
o o ~o ~9 ,,/
0 ~ '= : ~ r ~ , ' ' r , J i , , J i , ,
0 10 20 30 40 50 60 70 80 90 100
Calculated SFA (%)
Figure 5 - Calculated and measured values of SFA for full-size specimens9
100
/ , %~
90 .... -+20% /. ,9 [] ~,~ -
,,/" , 9 [z:] D o ~ Ei3 13
80 ...... :P.30% .. , = ~ D
9 ' / a agl"
7O , / ,6 a ~ %m[] ," /
,/'" 9' oh Qo 9 .,
,~ 60 / a 9 o / 9149 ,/
IJ. ,/' O0 D/ 0 ~ /*
t~ /' ~_ p9 ./
"o 50 /' J D / da []
/, 9 / o j 99 .B"
e 40 9i , ~ ,/o 99 o/ , , /
30 D Q El Q .I3
9 0/0 0 [] 9149 /'
20 o o ~ ' o~
9c~ //
10 .. o
0
0 10 20 30 40 50 60 70 80 90 100
Calculated SFA (%)
Figure 6 - Calculated and measured values of SFA for subsize specimens.

As far as full-size specimen tests are considered, only a negligible percentage of the
data points (1.4% - 3 tests out of 214) lie outside the 4-20% approximation band;
moreover, all 214 data are within the :530% band.
In the case of subsize specimens, a higher percentage of data lie outside the 4-20%
band (6.1% - 14 tests out of 228); if the level of approximation is relaxed to 4-30%, this
percentage drops however to 2.2% (5 in 228). This lower degree of accuracy could
somewhat be expected, since on one hand subsize specimens test diagrams normally
show a higher degree of oscillations, and are therefore more difficult to analyze; on the
other hand, SFA measurements on such a tiny fracture surface (9 mm9 are associated to a
higher degree of uncertainty.
It is worth noting that subsize specimen data points falling outside the 4-20%
accuracy limit are, with just one exception, resulting from calculated values over-
estimating the measured values.
All considered, the effectiveness of eq. 16 in evaluating shear fracture appearance
has been demonstrated, although the accuracy is better for full-size than for subsize
specimens.
Conclusions
1. A temperature adjustment is needed to account for the subsize specimens transfer time;
differences between nominal "furnace" temperatures and actual test temperatures, due
to thermal losses, can be as high as 13 ~ at -100 ~ and 21 ~ at 200 ~ The
alternative would be using a very fast robotic transfer system or using in situ
heating/cooling techniques.
2. The normalization factor (NF) experimentally determined from the ratio of upper shelf
energies (USE) of full-size and subsize specimens is in agreement with some of the
formulations given in the literature; nevertheless, for the prediction of full-size USE,
better results have been obtained by the use of an exponential fitting function and a
novel approach proposed by VTT.
3. The average shift calculated for conventional index temperatures (related to predefined
energy/lateral expansion/SFA levels) between full-size and subsize specimens is in
reasonable agreement with the well-known literature value of 65 ~ However, as far
as the scatter is concerned, temperatures related to energy levels (41/1.9 J, 68/3.1 J) are
associated to a more pronounced standard deviation than T0.89/0.3mmand FATTs0.
4. A normalization procedure for the subsize absorbed energy transition curve, recently
introduced by ORNL, can yield a reasonable approximation of the full-size impact
behavior, which can be improved by applying "personalized" coefficients based on the
analyses performed.
5. The application of the load diagram approach, with the calculation of alternative
characteristic temperatures (T~, T0, NDT), has not given favourable results, mainly due
to the high degree of uncertainty associated to characteristic force, energy and
displacement values measured from the instrumented data traces. This is particularly
relevant to subsize specimen tests and is, in turn, caused by several factors, such as:
highly scattered or limited data sets, differences induced by the use of alternative
fitting functions and the lack of a consolidated procedure for interpreting the tests.
6. Shear fracture appearance (SFA) values, calculated from characteristic forces, are in
excellent agreement with values measured (or estimated) from the specimens' fracture
surface; the approximation is better for full-size instrumented tests, which are usually
more straightforward to analyze with respect to subsize data.
7. The use of subsize specimens of the 3x4 type, such as those considered in this work,
for regulatory purposes might always be subject to the absolute uncertainty that exists
for the data. The scatter for the data points on which the correlations are based is the
limiting factor. Indeed, a lower-bound principle will invoke the addition of 2cy margins
on the correlations. This leads to penalizing values for the DBTT/USE values to be
used in safety cases. On the other hand, subsize specimens might provide reliable
information on annealing recovery, irradiation shifts and toughness of thin-walled
structures. Here, more research needs to be performed.
References
[1] Lucon, E., Bicego, V., D'Angelo, D., and Fossati, C., "Evaluating a Service-
Exposed Component's Mechanical Properties by Means of Subsized and Miniature
Specimens," Small Specimen Test Techniques Applied to Nuclear Reactor Pressure
Vessel Thermal Annealing and Plant Life Extension, ASTM STP 1204, W. R.
Corwin, F. M. Haggag and W. L. Server, Eds., American Society for Testing and
Materials, Philadelphia, 1993, pp. 311-323.
[2] "Proposed Standard Method for Instrumented Impact Testing of Subsize Charpy V-
Notch Specimens of Steels," Draft 7:12 June 1997, prepared by the Working Party
European Standards on Instrumented Charpy V-Notch Testing of Subsize
Specimens of the European Structural Integrity Society [ESIS] TC5.
[3] Lucon, E., "European Activity on Instrumented Impact Testing of Subsize Charpy
V-Notch Specimens (ESIS TC5)," Pendulum Impact Testing: A Century of
Progress, ASTMSTP 1380, T. Siewert and M.P. Manahan Sr., Eds., American
[4] Corwin, W. R. and Houghland, 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 Specimensfor Testing Irradiated Material, ASTM STP 888, W. R.
Corwin and G. E. Lucas, Eds., American Society for Testing and Materials,
Philadelphia, 1986, pp. 325-338.
[51 Corwin, W. R., Klueh, R. L., and Vitek, J. M., "Effect of Specimen Size and Nickel
Content on the Impact Properties of 12Cr-1MoVW Ferritic Steel," Journal of
Nuclear Materials, 122-123, 1984, pp. 343-348.
[6] Lucas, G. E., Odette, G. R., and Scheckerd, J. W., "Subsized Bend and Charpy V-
Notch Specimens for Irradiated Testing," The Use of Small-Scale Specimensfor
Testing Irradiated Material, ASTM STP 888, W. R. Corwin and G. E. Lucas, Eds.
American Society for Testing and Materials, Philadelphia, 1986, pp. 304-324.
[7] Lucas, G. E., Odette, G. R., Scheckerd, J. W., and Krishnadev, M. R., "Recent
Progress in Subsized Charpy Impact Specimen Testing for Fusion Reactor
Materials Development," Fusion Technology, Vol. 10, 1986, pp. 728-733.
[8] Louden, B. S., Kumar, A. S., Garner, F. A., Hamilton, M. L., and Hu, W. L., "The
Influence of Specimen Size on Charpy Impact Testing of Unirradiated HT-9,"
Journal of Nuclear Materials, Vol. 155-157, 1988, pp. 662-667.
[91 Kayano, H., Kurishita, H., Kimura, A., Narui, M., Yamazaki, M., Kano, Y., and
Shibahara, I., "Effects of V-Notch Dimensions on Charpy Impact Test Results for
Differently Sized Miniature Specimens of Ferritic Steel," Materials Transactions
JIM34(ll), 1993, pp. 1042-1052.
[101 Planman, T., Wallin, K., Valo, M., Ahlstrand, R., and Kohop~ig,J., "Comparison of
Some Impact Test Results on ISO-V and KLST type Charpy Specimens," 5th
Hungarian Seminar on Fracture Mechanics: Miskole, Hungary, April 6, 1995.
[111 Amayev, A. D., Badanin, V. I., Kryukov, A. M., Nikolayev, V. A., Rogov, M. F.,
and Sokolov, M. A., "Use of Subsize Specimens for Determination of Radiation
Embrittlement of Operating Reactor Pressure Vessels," Small Specimen Test
Techniques Applied to Nuclear Reactor Pressure Vessel Thermal Annealing and
Plant Life Extension, ASTMSTP 1204, W. R. Corwin, F. M. Haggag and W. L.
Server, Eds., American Society for Testing and Materials, Philadelphia, 1993, pp.
424-439.
[12] Klausnitzer, E., Kristof, H., and Leistner, R., "Assessment of Toughness Behavior
of Low-Alloy Steels by Subsize Impact Specimens," Transactions of the 8th
International Conference on Structural Materials in Reactor Technology, Brussels,
August 1985, Vol. G, International Association for Structural Mechanics in Reactor
Technology, 1986.
[131 Alexander, D. J. and Klueh, R. L., "Specimen Size Effects in Charpy Impact
Testing," Charpy Impact Test: Factors and Variables, ASTM STP 1072, J. Holt,
Ed., American Society for Testing and Materials, Philadelphia, 1990, pp. 179-191.
[14] Sokolov, M. A., and Alexander, D. J., "An Improved Correlation Procedure for
Subsize and Full-Size Charpy Impact Specimen Data," NUREG/CR-6379, ORNL-
6888, 1997.
[15] Wallin, K., "Mini-ja Normaalikokoisten Charpy-V-koesauvojen Tulosten V~ilinen

Korrelaatio," VTT-MET B-207, 1992 (in Finnish).
[16] Klausnitzer, E., "Micro-specimens for Mechanical Testing," Materialpr~fung 33,

1991, pp. 132-134.
[17] Ahlstrand, R., Klausnitzer, E., Lange, D., Leitz, C., Pastor, D., and Valo, M.,
"Evaluation of the Recovery Annealing of the Reactor Pressure Vessel of NPP
Nord (Greifswald) Unit I by Means of Subsize Impact Specimens," IAEA
Specialists' Meeting on Radiation Embrittlement of Nuclear Reactor Pressure
Vessel Steels, Balatonftired, Hungary, Sep 26-28, 1990.
[181 Wallin, K., "Murtumissitkeyskorrelaatiot," VTT Research Report 428, 1986 (in
Finnish).
[19] Fabry, A., van Walle, E., Chaouadi, R., Wannijn, J.-P., Verstrepen, A., Puzzolante,
J.-L., Van Ransbeek, Th., and Van de Velde, J., "RPV Steel Embrittlement:
Damage Modeling and Micromechanics in an Engineering Perspective," SCK, CEN
Report BLG-649, 1993.
[20] Fabry, A., van Walle, E., Van de Velde, J., Chaouadi, R., Puzzolante, J.-L., Van
Ransbeek, Th., Verstrepen, A., "On the Use of the Instrumented Charpy-V Impact
Signal for Assessment of RPVS Embrittlement," Evaluating Material Properties by
Dynamic Testing, ESIS Publication 20, E. van Walle, Ed., Mechanical Engineering
Publications Limited, London, 1996, pp. 59-78.
[21] Fabry, A., "Nuclear Reactor Pressure Vessel Integrity Insurance by Crack
Arrestability Evaluation Using Loads from Instrumented CVN Tests," Reactor
Pressure Vessel Thermal Annealing Demonstration Conference - EPRI, Santa Fe,
New Mexico, USA, Mar 31-Apr 1, 1998.
[22] Fabry, A., private communication, 1998.
[231 Server, W. L. and Ireland, D. R., "General Yielding of Notched Three-Point Bend
Specimens," Dynatup Technical Report TR 72-19, Effects Technology, Inc., Santa
Barbara, California, 1972.
[24] "Proposed Standard Method for the Instrumented Charpy-V Impact Test on
Metallic Materials," Draft 10, prepared by the Working Party European Standards
on Instrumented Charpy Testing of the European Structural Integrity Society
(ESIS) TC5.
[25] ISO/DIS 14556, "Steel -- Charpy V-Notch Impact Test -- Instrumented Test
Method," Edition: 1 (1998).
[26] "Proposed ASTM Standard Method for Instrumented Charpy V-Notch and
Miniaturised Charpy V-Notch Impact Tests on Metallic Materials," Draft 6: July
1998, prepared by ASTM Sub-Committee E28.07.08 (Miniaturised

Charpy/Instrumented Testing).
Y. Yamaguchi, 1 S. Takagi, 1 and H. Nakano I
Effects of Anvil Configurations on Absorbed Energy
Reference: Yamaguchi, Y., Takagi, S., and Nakano, H., "Effects of Anvil
Configurations on Absorbed Energy," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. P. Manahan, St., Eds., American Society
Abstract: The effects of anvil configurations on the absorbed energy were quantified
using Japanese V-notch reference test specimens at three energy levels (30 J, 100 J,
and 160 J). All tests were performed on a 500 J C-type machine with a 2 mm radius
striking bit. The test temperature was 0~C. The effects of five critical parameters of
anvils and supports were analyzed with an orthogonal array L18. These effects were
finally expressed as a formula of the uncertainty propagation. Furthermore, the
difference of absorbed energy with different tapers of anvils was studied.
Keywords: anvil configurations, tapers of anvils, influence on absorbed energy,

reference test pieces
Introduction
Charpy V-notch impact testing is used by hundreds of organizations in Japan.

The impact machines are verified by using reference test specimens, but some
dispersion in the test results has been found among the organizations. The sources of
the dispersion can be categorized into two items; one is the test procedure including
the machine itself, and the other is the material of the reference test specimens. There
have been several studies to understand the effects of the former sources. The
dimensional parameters of an impact machine have been investigated, whereas
aMaterials Measurement Section, National Research Laboratory of Metrology,

1-1-4 Umezono, Tsukuba-shi, Ibaraki, Japan 305-8563.
164
YAMAGUCHI ET AL. ON ANVIL CONFIGURATIONS 165
bending specimens with 3, 5 and 7 mm thickness were used instead of Charpy

reference specimens [1]. The effect on the surface finish of anvils and striking bits has
also been described [2]. In this paper, the effects of the dimensional parameters and
alignment in critical parts of a Charpy impact test machine are quantified by using V-
notch Japanese reference test specimens, which give the practical results and the direct
interpretation. An orthogonal array Lls is adopted to minimize accidental errors and
improve the reliability of the experiments.
In addition, the influence of the taper of the anvils is investigated. Although 10~
taper of anvils based on ASTM are widely used in Japan, ISO Metallic materials-
Impact testing-Preparation and characterization of Charpy V reference test pieces for
verification of pendulum impact testing machines (DIS 12736) requires the taper to
have 11~ ~ Recently the specification in JIS Charpy pendulum impact test -
Verification of testing machines (B 7722) has been revised to meet the ISO
requirement. To make clear the effect of this alteration, the absorbed energy with
different tapers ranging from 9 ~ to 12~ is obtained.
Experiments
Specimens
All specimens were V-notch Japanese verification specimens. The specimen

dimensions were satisfied with the requirements of ISO/DIS 12736. The material
specification and preparation are published in Japanese Industrial Standard JIS
Standard Specimens for Charpy Impact Testing Machine (B7740-1990). Three energy
levels of 30 J, 100 J and 160 J were commercially available, and designated as L, H
and SH, respectively. Each specimen has a unique identification number, and its
hardness is measured. The typical CV values (deviation/mean) of absorbed energy are
about 3 % at all energy levels [3].
Machine
A 500 J C-type machine with a 2 mm striker was used. The machine is fixed on
a 1500•215 mm steel block foundation, which is embedded in a concrete floor
to ensure a rigid mounting of the machine. All tests were performed at the temperature
of 0 ~ The specimens were conditioned for at least 15 minutes in a cooling bath
filled with alcohol liquid, and tested within 5 seconds.
Determination of the Effect of Anvil Geometry and Alignment
Five factors of anvil geometry and alignment were examined to estimate their
influence on the absorbed energy. Three experimental levels were selected for each
factor. Figure 1 shows a schematic drawing of the five factors which were A, the
distance between the planes containing anvil surfaces (0, 0.2, 0.4 mm); B, the distance
between anvils (40, 40.4, 40.8 mm); C, the radius of curvature of anvils (1, 1.5, 2.0
mm); D, the distance between the center of the striker and the center of the anvil gap (0,
0.5, 1.0 mm); E, the distance between the planes containing support surfaces (0, 0.2,
0.4 mm). Table 1 summarizes these factors with the three experimental levels and the
related tolerances for reference machines in ISO 12736. The three levels cover a larger
range than the tolerances so that obvious effects are obtained.
The effects of these factors were studied through 18 runs of the experiment
which were planned according to an orthogonal array L18. The test matrix I.,18with data
is shown in Table 2. The numbers 1, 2 and 3 in Table 2 indicate the three experimental
levels. The L18 array permits up to eight factors, but in this experiment only five factors
are assigned, and thus three columns indicated by 'e' are used to analyze errors. Several
special anvils and supports which had different thicknesses and radii of curvature were
machined to meet each experimental condition presented in Table 2. Figure 2 shows a
typical experimental set-up, which corresponds to experiment No. 15 in Table 2. As
shown in Figure 2, all five factors were varied simultaneously with the experimental
condition. Each effect was individually deduced with the orthogonal array after the 18
runs were finished, thereby the accidental errors were averaged and the reliability of
the experiments was improved.
Five specimens were tested in each run at every energy level. The total number
of specimens was 90 at one energy level. The radius of curvature of anvils was
measured with a profile measuring machine. The distance between anvils was
measured with an inside micrometer.
Determination of the Effect of Tapers of the Anvils
Four different anvils having the tapers of 9 ~ 10 ~ 11 ~ and 12 ~ were examined.

Ten specimens were tested with each anvil in the first run, and then an additional
fifteen specimens were tested. The taper shapes were measured with the profile
measuring machine.
Anvils
Anvils
Striker ~ '
S u p p ~ ~
Specimen
,4. Distance between the Planes B. Distance between Anvils.
Containing Anvil Surfaces.
Anvils
C. Radius of Curvature of Anvils.

Center of the Anvil Gap
~ t e r of the Striker
D. Distance between the Center of the Striker E. Distance between the Planes
and the Center of the Anvil Gap. Containing Support Surfaces.
Figure 1 - Selected Error Factors.
Table 1 - Factors and Levels.
Content of level Specification

Symbol Error factor
Level 1 Level 2 Level 3 in ISO 1
A Distance between the planes containing 0.0 mm 0.2 mm 0.4 mm < 0.1 mm
anvil surfaces
B Distance between anvils 40.0 mm 40.4 mm 40.8 mm 40 +0"1 mm
C Radius of curvatures of anvils 1.0ram 1.5ram 2:0mm 1.0+o~ mm
D Distance between the the center of the 0.0ram 0.5mm 1.0mm < 0.25 mm
striker and the center of the anvil gap
E Distance between the planes containing 0.0 mm 0.2 mm 0.4 mm < 0.1 mm
support surfaces
1 ISO/DIS 12736 specification for reference machines.

O~
oa
Table 2 - Assignment and Data

"D
Ill
B e CD E e Energy level of specimen z
No. e A E3
L H SH c
r-
E
1 11311 11 11 28.9 29.3 29.3 29.7 29.7 103.2 104.1 102.7 103.2 101.9 168.5 175.6 168.5 163.2 174.6
2 11 22 22 22 29.3 28.2 28.2 28.9 28.9 88.8 94.4 91.9 89.7 92.7 157.6 172.7 150.1 152.4 156.2
3 11 33 33 33 28.6 28.9 29.3 27.5 29.3 94.9 94.4 91.4 94.9 94.0 165.6 160.9 154.3 153.8 154.8
0
4 12 11 22 33 30.4 29.3 28.6 29.7 28.2 94.0 93.1 95.3 97.0 96.2 163.2 159.0 167.5 153.3 169.9 .--t
5 12 22 33 11 28.2 28.6 30.0 30.0 28.9 93.1 93.6 94.4 96.2 93.6 169.4 162.8 158.0 161.3 158.5 .--t
m
6 12 33 11 22 29.7 30.4 30.7 29.3 29.7 103.2 98.8 102.7 100.5 98.8 172.2 170.8 172.2 173.7 185.1 -t
7 13 12 13 23 30.7 31.8 30.0 30.0 30.7 96.6 101.4 105.8 104.1 102.3 181.8 178.0 173.7 184.2 175.1
8 13 23 21 31 28.9 28.9 28.9 28.9 28.6 94.9 93.6 89.7 90.6 92.3 153.3 151.0 154.8 155.2 151.0
9 13 31 32 12 27.9 28.6 28.6 27.5 29.3 92.3 91.4 90.1 92.3 87.1 155.7 153.3 150.1 151.0 157.6 0
m
10 21 13 32 21 28.6 28.6 29.3 29.7 28.9 92.7 93.2 90.1 94.0 91.9 150.5 157.6 162.3 159.0 152.4 z
11 21 21 13 32 29.7 30.4 30.4 30.4 30.7 98.8 99.7 100.5 99.7 102.3 175.6 177.5 180.4 181.3 185.6 c
12 21 32 21 13 29.3 29.3 28.9 28.9 29.7 89.3 94.4 91.4 94.4 94.4 158.5 152.9 152.4 154.3 144.9
0
13 2 2 12 31 32 28.9 29.7 28.9 28.9 27.5 91.0 91.0 88.9 93.2 93.2 150.5 158.5 154.8 156.6 155.2 71
"u
14 2 2 23 12 13 30.0 27.9 30.7 30.7 29.7 97.9 98.4 101.9 104.1 99.2 179.4 166.1 165.1 176.1 173.2 21J
15 2 2 2 31 23 21 29.3 30.7 29.7 31.1 29.7 94.4 96.2 97.1 89.7 90.6 160.9 163.7 156.2 159.5 159.5 0
D
2o
16 23 13 23 12 30.4 29.3 28.9 30.4 29.7 94.9 94.9 99.2 98.4 97.1 166.1 168.9 166.1 173.7 174.2 m
17 23 21 31 23 28.9 30.0 28.6 29.3 28.9 89.7 88.9 91.4 92.7 91.9 153.3 155.7 151.5 156.6 159.5 co
18 23 32 12 31 30.4 29.3 29.3 31.1 28.9 94.9 96.6 95.3 101.0 100.1 162.3 164.2 179.4 166.6 172.3
2 Experimental setup of No. 15 is shown in Figure 2.

3 Experimental levels.
Center of the Anvil Gap
R = 1.5 rnm
/
i! ~.~" ./~., ] " ~ Centerof the Striker
~ Anvils
Striker Supports Specimen
t",l
d
Figure 2 - An Exampleof the Setupfor ExperimentNumber15.
Results
Determination of the Effect of Anvil Geometry and Alignment
The analysis of variance is shown in Table 3. The first-order error el represents

the variation between 18 experiments, and the second-order error e: is variation
between repetitions. As shown in Table 3, el was not significant compared with e2 at
all energy levels, which implies that the experimental conditions were controlled well,
so both errors were pooled into e in Table 3. The significant factors are noted by the
symbol ** affixed in Table 3 when the variance ratio is greater than the 1% value in
the F-table. The factor A is not significant at L, H and SH. The factor B is significant
at H and SH. In contrast, factor C and D are significant at all levels. The factor E is not
significant at all levels. Figures 3 through 12 and Tables 4 through 8 show the
experimental results. The trends of each factor in three energy levels are almost same.
For comparison between the influence of the present factors and that of the
variation of the specimens, the obtained effects were expressed as a formula of the
uncertainty propagation in the approximation of non-correlated independent variables.
It was assumed for convenience that the absorbed energy E was the function of these
five quantities of factors. The expression for the combined variance is given by
,(OF4Ox,)2u2(x,) (1)
where
u (E) = standard uncertainty of the absorbed energy,

OE / Oxi = sensitivity coefficient,
u(xl) = a standard uncertainty of the estimate x,.
To estimate the sensitivity coefficient in a simple manner, the relation between
the absorbed energy and each quantity was assumed to be linear. The sensitivity
coefficients are shown in Table 9. It is also assumed that these quantities are tested
with limit gages, which are used to confirm whether the quantities are within the
tolerances or not. Thus it can only be said that it is equally probable for the quantity to
lie anywhere within the tolerance, i.e. a rectangular distribution. The variance of the
expectation value for the rectangular distribution is given by
U 2 (X i ) = a 2/12 (2)
where
a = the whole range of the tolerance.

Table 9 summarizes the calculated results of the uncertainty. The radius of anvils
is important for H level specimens, while the distance between the center of the striker
and the center of the anvil gap is more important for the SH level. However, if the
uncertainty is compared with the variation of the specimens, for instance CV value,
the influence of the present sources may be limited.
Table 3 -Analysis of Variance Tables.
(a) L Level
Source Degree of freedom Variation Variance Variance ratio

f S V Fo
A 2 1.20 0.61 1.3
B 2 0.29 0.15 0.3
C 2 18.56 9.28 20.2 **
D 2 6.86 3.43 7.5 **
E 2 1.98 0.99 2.2
e 79 36.23 0.46
(el) (5) (3.55) (0.72)
(e2) (74) (32.68) (0.44)
Total 89 65.13
(b) H Level

f S V Fo
A 2 11.66 5.83 1.2
B 2 73.94 36.97 7.6**
C 2 1212.41 606.21 125.4 **
D 2 78.53 39.26 8.1 **
E 2 27.16 13.58 2.8
e 79 381.81 4.83
(el) (5) (53.02) (10.60)
(e2) (74) (328.79) (4.44)
Total 89 1785.52
(c) SH Level

f S V Fo
A 2 58.31 29.15 1.0
B 2 255.73 127.87 4.5 **
C 2 5264.61 2632.31 93.0 **
D 2 1027.89 513.94 18.2 **
E 2 10.92 5.46 0.2
e 79 2236.18 28.31
(el) (5) (419.33) (83.87)
(e2) (74) (1816.85) (24.55)
Total 89 8853.64
Table 4 - Influence of the Distance between the Planes Containing Anvil Surfaces (Factor A).
Distance between the planes containing anvil surfaces

Energy 0 mm 0.2 mm 0.4 nun
level Average Average Difference Difference Average Difference Difference
[J] [J] [J] [%] [J] [J] [%]
L 29.2 29.5 0.3 0.9 29.4 0.2 0.7
H 95.6 95.9 0.3 0.3 95.0 -0.6 -0.6
SH 162.5 164.4 2.0 1.2 163.2 0.7 0.5
110 180
32
105 170
~o
30 loo g -
------ I ][ m 1602
95] ~ ~
28
<
26
012 014 85 |
0.2 014 140 0 012 014
Distance [ram] Distance [ram] Distance [ram]
(a) L Level (b) H Level (c) SH Level
Figure 3 - Factorial Effect of Factor A at Different Energy Levels.

The error bars indicate confidence limit with 95% reliability.)
15
H . . . . -0-....
I0 SH m---
o,
-5
"100 011 012 013 014 0.5

Distance [mm]
Figure 4 - Relative Effect of Factor A.

Table 5 - Influence of the Distance between Anvils (Factor B).
Distance between anvils

Energy 40 m m 40.4 m m 40.8 m m
[J] [J] [J] [%] [J] [J] [%]
L 29.5 29.3 -0.1 -0.4 29.4 -0.1 -0.4
H 96.8 94.9 -1.9 -2.0 94.9 -1.9 -2.0
SH 165.4 163.4 -2.0 -1.2 161.3 -4.1 -2.5
110 180
E lO5
170
30 100 ~ '
~ ~ - ~ 160'
~ 28
o ~ 150
< .~
< 90 <
2
' ' 85 ~ ~ 140 ~
40.0 40.4 40.8 40.0 40.4 40.8 40.0 40.4 40.8
Distance [mm] Distance [mm] Distance [mm]
Figure 5 - Factorial Effect of Factor B at Different Energy Levels.

(The error bars indicate confidence limit with 95% reliability.)
15
L
H .... -o- . . . .
I0
SH m---
5
8
On
t5
-10 i i J i
40.0 40.2 40.4 40.6 40.8 41.0
Distance [ram]
Figure 6 - Relative Effect of Factor B.

Table 6 - Influence of the Radius of Curvatures of Anvils (Factor C).
Radius of curvature of anvils

Energy 1.0 m m 1.5 m m 2.0 m m
[J] [J] [J] [%] [J] [J] [%]
L 30.0 29.3 -0.7 -2.2 28.9 -1.1 -3.7
H 100.7 93.7 -7.0 -6.9 92.2 -8.4 -8.4
SH 174.1 159.3 -14.8 -8.5 156.7 -17.4 -10.0
i10 180
32 q
105
170"
o 30i -
o 100:
160
95
28
.~ 90 .~ 150
< < <
26 ,
' 85 140
1.0 1.5 2.0 1.0 1.5 2.0 1.0 115 2.0
Radius [mm] Radius [mm] Radius [ram]
Figure 7 - Factorial Effect of Factor C at Different Energy Levels.

10
H .... -o- . . . .
5 SH n-__
E 01
~. -5
-10 s--..~_._ ~
,
"15.0 1.5 2.0
Radius [mm]
Figure 8 - Relative Effect of Factor C.

Table 7 - Influence of the Distance between the Center of the Striker and the Center of the Anvil Gap
(Factor D).
Distance between the planes containing the center of the anvil gap
Energy 0 mm 0.5 mm 1.0 mm
[J] [J] [J] [%] [J] [J] [%]
L 29.2 29.2 -0.1 -0.2 29.8 0.5 1.9
H 95.2 94.6 -0.6 -0.6 96.8 1.6 1.7
SH 160.2 161.9 1.7 1.1 168.0 7.9 4.9
110 180
32
105 17o
30 100
~ .
'x:;t
160:
95:
O
.~ 90
150
< < <
26
85
015 1.0 0 015 1.0 1400 015 1.0
Distance [mm] Distance [ram] Distance [mm]
Figure 9 - Factorial Effect of Factor D at Different Energy Levels.

15
L 9
H .... ~m- . . . .
10 SH [] -
g 5 ,AI
8
Oi~ 1
-5
"I00 015 1.0

Distance [ram]
Figure 10 - Relative Effect of Factor D.

Table 8 - Influence of the Distance between the Planes Containing Support Surfaces (Factor E).
Distance between the planes containing support surfaces

Energy 0 mm 0.2 mm 0.4 mm
[J] [J] [J] [%1 [J] [J] [%]
L 29.3 29.6 0.3 1.1 29.3 0.0 0.0
H 96.3 95.2 -1.1 -1.1 95.1 -1.2 -1.3
SH 163.3 163.8 0.5 0,3 162.9 -0.4 -0.2
110 180
32
~" 105 "~ 170
~o
30
~ m 1602 3_ -~
95] ~
28
o o 150
,~
< 90 ,~
<
26
012 014 85 012 014 1400 012 014
Distance [mini Distance [mm] Distance [mm]
Figure 11 - Factorial Effect of Factor E at Different Energy Levels.

15
L 9
H .... -O-. . . .
10 SH t~---
-5
-100 011 012 013 014 0.5

Distance [mm]
Figure 12 - Relative Effect of Factor E.

Table 9 - Uncertainty Propagation in the Evaluation of Effects of Anvil Geometry and Alignment.
i Sensitivity coefficient at Contribution to u2(A) at different energy levels
2 different energy levels
Tolerance u 2 (x/) = ~ 2 3A u2(A)=f gA] 2u2(xi) tJ2]
Source xi [J/mm]
ai [mm] [mm2 ] ,~xi
L H SH L H SH
A Distance between the planes containing 0.1 0.8333 x 10 -3 0.51 -1.48 1.84 0.21 x 10 -3 1.83 x l 0 -3 2.83 x 10 -3 G)
C
anvil surfaces 0
T
B Distance between anvils 0.1 0.8333 x 10 -3 -0.13 -2.40 -5.16 0.01 x 10 -3 4.78 x l 0 -3 22.20 x 10 -3
m
C Radius of curvatures of anvils 0.1 0.8333 x 10 -3 -1.10 -8.41 -17.36 1.02 x 10 -3 58.89 x10 -3 251.22 x 10 -3
D Distance between the center of the 0.25 5.2083 x 10 -3 0.55 1.61 7.86 1.56 x 10 -3 13.51x 10 -3 322.10 x ! 0 -3 F
striker and the center of the anvil gap 0
z
E Distance between the planes containing 0.1 0.8333 x 10 -3 0.03 -3.06 -0.97 0.00 x 10 -3 7.80 x 10 -3 0.78 x 10 -3
z
support surfaces
t-
Total UC z2 ~ U 2 [J2] 2 . 8 1 x 10 -3 8 6 . 8 1 x 10 -3 599.14 x l 0 -3 O
0
z
Standard uncertainty Uc [J] 0.05 0.29 0.77 "11
Expanded uncertainty U = k 'ur [J] 0.11 0.59 1.55 c
0
z
"4
Determination o f the Effect o f Tapers o f the Anvil
Table 10 shows the experimental results. Since there were not significant
differences between the first and second run, the error due to setting of the anvils could
be negligible. The mean values and standard deviations are also presented in Table 10.
Figure 13 shows the absorbed energy with different tapers at each energy level. The
hypothesis that the mean absorbed energy at 10 ~ taper was equal to that at 11 ~ taper
was tested by using Student's t distribution with the 95% reliability. The difference at
L level was within the 95% reliability, and the hypothesis was not rejected. In contrast,
the hypothesis was rejected at H and SH levels, where the mean energy at the 10 ~ taper
was 3% higher than that at the 11 ~ taper. These discrepancies might not be important in
verification of industrial machines where the acceptance error may be 10%, but they
must be considered when reference machines are verified.
110[ 180
32 1 0 5 ~ . ~ 170,
30 ~100~
~6o
O
28
.~ 95t
-~
< 90
.~ 150
<
< 26
, 85 I 140
10 111 12 9 10 111 12 9 10 1'1 12
Taper [degree] Taper [degree] Taper [degree]
Figure 13 - Effect of Taper of Anvils at Different Energy Levels.
Table 10 - Experimental Result of Evaluation of the Taper of Anvils.
Energy level L H SH
Taper of anvils 9* 10" 11" 12" 9* 10" 11" 12" 9* 10" 11" 12"
30.6 30.2 30.9 29.5 112.3 101.7 97.8 100.4 170.2 162.6 157.0 169.2
30.6 31.3 31.7 33.5 106.5 105.2 99.9 102.1 164.5 163.1 162.1 156.0
30.6 31.3 29.2 31.7 102.6 100.4 96.5 103.0 175.0 165.9 154.6 156.0
30.6 30.9 31.3 31,3 108.8 105.7 95.6 94.7 167.4 169.2 158.4 153.7
33.5 31,7 30.2 31.7 99.1 102.6 103.5 102.1 178.8 175.4 162.1 177.8
30.2 29.9 30.9 32.0 106.5 100.4 99.9 102.1 165.5 165.9 156.5 156.0
31.3 30.2 29.9 31.3 109.2 106.5 97.8 106.1 163.6 173.1 153.2 157.0 >
30,6 32.0 30.9 32.0 104.8 100.4 101.3 98.2 167.4 166.9 172.1 158.4 fi3
t-
30.2 30.9 29.9 30.6 106.5 96.9 96.9 101.7 174.0 174.5 165.5 166.4 O
"1"
32.4 30.6 31.3 30.2 107.9 107.0 102.1 98.2 178.8 175.9 155.1 166.4
107.0 104.8 102.1 101.3 169.2 145.7 175.0 148.5 m
30.9 30.6 32.0 30.2 -4
30.9 30.6 31.3 29.9 105.2 104.3 98.6 96.5 166.4 165.0 149.9 158.8 >
.r-
Absorbed energy ,4i [J] 30.2 31.7 30.6 30.6 104.8 99.5 98.2 101.3 164.5 167.4 153.2 153.7
30.6 29.9 32.4 30.6 111.4 101.3 102.6 99.9 163.6 159.8 151.3 162.1 O
z
30.9 30.6 30.6 30.6 107.0 102.6 98.2 98,6 180.2 166.4 162.6 166,9 )>
32.7 29.9 30.9 31.3 100.8 98.6 99.5 95.6 162,1 160.7 169.7 151.8 z
<
30.9 31.3 31.3 31.7 107.4 106.1 96.9 98.2 164.5 158.4 157.9 155.1 F
31.3 30.2 29.2 32.0 102.6 105.2 100.8 97.3 164.0 158.8 174.5 165.5 0
0
32.0 31.7 29.9 30.6 102.6 98.6 95.2 96.5 171.2 171.2 151.3 164.5 z
32.0 29.2 31.7 32.4 101.7 102.1 102.6 100.8 178.3 161.2 158.8 161.2
31,7 32.4 30.6 31.7 102.1 100.4 95.6 104.8 171.6 163.6 153.7 169.2 c
32.0 30.2 33.1 30.2 103.5 105.7 97.8 100.4 167.8 160.3 152.3 152.3
32.4 30.6 30.2 32.0 105.2 99.5 101.3 100.4 176.4 163.1 165.0 168.3
z
30.9 30.9 31.7 30.6 101.7 103.5 102.6 103.5 167.4 174.5 161.2 158.8
29.2 99.5 96.0 98.6 99.9 178.3 161.7 157.0 171.2
Average _~ [J] 31.3 30.8 30.8 31.2 105.1 102.2 99.3 100.1 170.0 165.2 159.6 161.0
Standard deviation s4 [J] 0.9 0.8 1.0 0.9 3.5 3.1 2.5 2.8 5.8 6.8 7.3 7.3
t.o
Conclusions
The influence of the critical parts of impact machines are investigated. The
relevant uncertainties are calculated providing that these parameters are independent.
When the present parameters are within the tolerances of reference machines in ISO
12736, the influence of the parameters is not significant compared with that of the
variation of the specimens. However, the radius of anvils is not disregarded for
industrial machines because the difference of absorbed energy between 1.0 and 1.5 mm
radius is more than 5% at SH level.
The relation between the absorbed energy and anvil tapers are obtained. The
absorbed energy decreases with increasing the taper angles. The difference of tapers
between ISO and ASTM should be considered when the reference machines are
verified. It should be mentioned that there are many other parameters which have the
influence on the absorbed energy, and complex interactions between the parameters
can be expected. Further investigations are needed to understand the uncertainty of
Charpy impact test.
References
[1] Revise, G., "Influence of Dimensional Parameter of an Impact Machine on the

Results of a Test," Charpy lmpact Test: Factors and Variables,ASTM STP 1072,
J. M. Holt, Ed., American Society for Testing and Materials, Philadelphia, 1990,
pp. 35-53.
[2] Ruth, E. A., Vigliotti, D. P., and Siewert, T. A., "Effect of Surface Finish Charpy
Anvils and Striking Bits on Absorbed Energy," Pendulum Impact Machines:
Procedures and Specimens for Verification,ASTM STP 1248, T. A. Siewert and
A. K. Schmieder, Ed., American Society for Testing and Materials, Philadelphia,
1995, pp. 91-100.
[3] Hida, N., "Production of Charpy Impact Verification Specimens and Verification of
Machine Performance," Pendulum Impact Machines: Procedures and Specimens
for Verification, ASTM STP 1248, T. A. Siewert and A. K. Schmieder, Ed.,
American Society for Testing and Materials, Philadelphia, 1995, pp. 171-181.
M. P. Manahan, Sr.1 and R. B. Stonesiferz
The Difference Between Total Absorbed Energy Measured Using An Instrumented

Striker and That Obtained Using an Optical Encoder
Reference: Manahan, M. R, Sr., and Stonesifer, R. B., "The Difference Between

Total Absorbed Energy Measured Using An Instrumented Striker and That
Obtained Using an Optical Encoder," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. R Manahan, Sr., Eds., American Society
Abstract: The application of optical encoders in Charpy impact energy measurement has
improved significantly the accuracy of "dial energy" determination. Instrumented
strikers offer an alternative method of energy measurement which is accurate and
reproducible for both conventional and miniature specimen testing while providing
additional useful information such as general yield load, peak load, brittle fracture load,
and brittle fracture arrest load. It has been observed that the total absorbed energy
measured using these two technologies, while generally in good agreement, sometimes
differs by a significant amount. The instrumented striker total absorbed energy has been
found to be higher or lower than the optical encoder energy depending on the ductility of
the test specimen and other factors. This paper examines and provides explanations for
these energy differences. A summary of mechanisms for pendulum energy loss, other
than in fracturing the test specimen, is provided along with estimates of the amount of
energy which may be associated with each mechanism.
Keywords: impact testing, instrumented striker, absorbed energy, optical encoder
Introduction
The energy measured by the dial in a Charpy test is basically the potential energy lost
by the pendulum as determined from the difference in height of the pendulum's mass
before and after its swing. This energy commonly is measured by either a mechanical
dial gage or an electrical encoder that essentially measures the pendulum angles at the
start and end of the swing. It is the energy that goes into dynamically deforming and
fracturing the test specimen that is intended to be measured by the Charpy test. Energy
that is lost through other mechanisms is not intended to be included in the measured
Charpy energy. For example, loss due to friction of the pendulum with the surrounding
1President, MPM Technologies, Inc., 2161 Sandy Drive, State College, PA 16803.
2 Owner, Computational Mechanics, Inc., 1430 Steele Hollow Road, Julian, PA 16844.
181
air and loss due to bearing friction are corrected routinely by measuring the energy loss
due to swinging the pendulum without a specimen being installed in the test machine.
Since it is the energy that goes into the specimen that is intended to be measured, a more
direct means for measuring this energy is to measure the work that is done on the
specimen by the striker during the impact. This work can be calculated if the force
exerted on the specimen and the displacement at the striker contact surface is known
throughout the period of contact between the striker and the specimen. It is possible to
build a load cell into a Charpy striker that provides impact force as a function of time.
The displacement history can be obtained by using the initial velocity at the time of first
contact and the striker deceleration history. The measured force is used to compute the
deceleration history of the pendulum using Newton's second law (e.g., F = ma). Two
integrations of the deceleration versus time function, combined with the initial velocity,
provide the required displacement function.
The impetus that led to the work reported in this paper was the observation that
significant differences can sometimes exist between Charpy energies measured via the
dial gage/encoder technology and those determined by calculating the work done on the
specimen via the use of a striker load cell. This paper describes several opportunities for
such energy differences to arise. Some of the mechanisms illustrate that the dial
gage/encoder approach can result in larger energies being reported than actually go into
the specimen. Other mechanisms illustrate that loads measured by striker load cells may
involve some error. As should be expected, neither approach to measuring the energy to
break the specimen (i.e., the Charpy energy) is perfect.
A combination of experimental and numerical studies was used in this work, The
bulk of the experimental work was done using 4340 steel specimens which were
provided by the National Institute of Standards and Technology (N/ST). Three
calibration specimen types were used, low energy (~,17 J), high energy (~100 J), and
super high energy (~225 J), so that the entire spectrum of specimen behavior (very brittle
to very ductile) could be studied. Comparing encoder-based energies to the integrated
striker energies from NIST's low, high, and super high specimens, for a large number of
tests, it was found:
9 low energy specimens resulted in encoder energies 10-17% greater than the
instrumented striker's integrated energy (2 to 3 J difference)
9 high energy specimens resulted in encoder energies 2-3% less than the instrumented
striker's integrated energy (3 to 4 J)
9 super high energy specimens resulted in encoder energies 1-2% les_.__ssthan
instrumented striker's integrated energy (3 to 4 J)
It can be seen that the energy differences for the low energy specimens are much larger
than for either the high or super high specimens when viewed on a percentage basis. On
an absolute energy difference basis, the low energy specimens have slightly less
difference than the high and super high specimens. The sign of the energy difference
changes as the ductility of the specimen increases.
It was also observed that the low energy specimen energy difference was dependent
on whether the specimen halves exited from the front or the back of the machine. When
tested at room temperature, the low energy specimens consistently exited the machine in
the same direction as the swing of the striker (front). When tested at a lower
MANAHAN AND STONESIFER ON TOTAL ABSORBED ENERGY 183
temperature, the low energy specimens often exited counter to the striker swing direction
(to the back). For these rear exiting specimens, the energy difference was about 5% less
than for the room temperature specimens that exited from the front. This leaves a
difference of about 1 to 2 J. It was also observed that occasionally a super high energy
specimen would not break into two pieces. For these specimens, the difference between
encoder energy and integrated striker energy was near zero.
Three dimensional (3D) finite element simulations were performed which made use
of two models. The grid for one model represented the pendulum, U-hammer, and
striker o f a Tinius Olsen model 84 test machine (see Figure 1). The center of percussion
for this grid was within 0.25 mm (0.01 in.) of the designed center of strike (CS). The
grid for the other finite element model represented just the instrumented striker portion
of the hammer. The finite element model of the entire pendulum included the effect of
deformations and vibrations in the pendulum assembly and was easily altered so as to
determine the effect of shifting the strike location relative to the pendulum's center of
percussion. The detailed finite element model of the striker was useful for comparing
dynamic load cell response to static load cell calibration behavior and for studying the
effects of non-uniform contact pressure distributions between the specimen and the
striker. Input to these finite element models was either the force versus time history from
an actual test, or an idealized impulse defined by:
/ "x
f~---Fmax sinl-~t1 for 0 "(t <~tt (I)
f = 0 f o r t > t,
By varying tt, this idealized impulse was used to determine the effect of impulse duration
(or frequency) on the load cell behavior.
Vibrational Energy in the Pendulum
The encoder-based energy gives a very reliable indication of the total energy loss in
the experiment. This encoder-based energy loss includes the desired Charpy test energy
as well as losses due to windage and friction (until corrected), any secondary impacts
with the broken specimen, and vibrational energy that remains in the pendulum
assembly. The striker energy integration, on the other hand, does not include any
vibrational energy left in the pendulum, and does not reflect energy transfers between the
striker and the specimen that take place after the specimen breaks. A study was made to
determine if the tendency for integrated striker energies to be smaller than encoder
energies for low energy specimens could be explained by vibrational energy left in the
pendulum.
The full pendulum finite element model was used for this study. A series of analyses
were made using an idealized impulse loading with impulse times (tt) ranging from 2 to
5000 ~tsec. It was important to study such a wide range of impulse times because Fourier
decomposition of Charpy load versus time histories results in a similarly wide range of
frequencies. Furthermore, the pendulum vibration amplitudes and energies excited by an
impact depend on the impulse duration as a result of their ability or inability to excite the
pendulum's natural modes and frequencies of vibration. ASTM Standard Test Methods
-2
J
/
("\
\
Figure 1-Finite Element Grid of the Charpy Test Machine Pendulum

for Notched Bar Impact Testing of Metallic Materials E23 allows Charpy test machines
to have their center of percussion (CP) above or below the CS by as much as 1% of the
distance between the pendulum's axis of rotation and the CS. Since shifting the strike
location relative to the CP would also be expected to affect the tendency to excite
vibrations in the pendulum, the finite element study considered three strike locations.
One location was essentially at the CP, one was 1% (8.9 mm) above the CP, and the
other was 0.57% (5.1 ram) below the CP. The lower point was not shifted the full 1%
because the striker does not extend far enough to allow this. Due to assumed symmetry
in the geometric modeling and impact force loading, torsional and lateral pendulum
vibrations inherently were assumed not to exist.
Figure 2 shows the results of the finite element analyses which explored the effect of
impulse duration and impact location on pendulum vibrational energy. The peak impact
force (Fm~) was kept constant at 620 kN (7000 lb) for all analyses. With peak force kept
constant, the nominal impact energy was proportional to the impulse duration. It can be
seen that the pendulum and striker vibrational energy is very dependent on both impulse
duration and strike location. Figure 2a shows that the least vibrational energy (~0.05 J) is
left in the pendulum for all three strike locations when impulse durations are around 250
~sec. With impacts at or below the CP, the peak vibrational energies (1.2 and 1.7 J)
occurred near an impulse duration of 33 ~tsec. While the impacts at 1% above the CP
had a local peak at 33 ~tsec (0.6 J), the highest computed vibrational energy was 2.9 J for
an impulse duration of 5000 ~tsec. Figure 2b shows the same vibrational energy results
as Figure 2a except that the vibrational energy is normalized by the nominal impact
energy. For impulse durations of 250 ~tsec or larger, the vibrational energy in the
pendulum that remains after the impact is 0.2% or less of the total impact energy.
However, for impulse durations of less than 250 ~tsec, the portion of the impact energy
that goes into pendulum vibrational energy continually increases until 100% is
approached for about a 2 ~sec duration.
Based on the results of Figure 2, it is concluded that while ductile specimens, such as
the NIST high and super high materials, might be expected to induce as much as 3 J of
pendulum vibrational energy, this energy is negligible compared to the total Charpy
energy being measured. On the other hand, for the NIST low energy material, the total
impact duration is about 200 ~tsec. It therefore appears from Figure 2 that vibrational
energy should be near zero both in absolute and relative terms. However, one must recall
that the impulses used in this study were smooth idealized impulses. While the total
impact duration of the low energy specimens may be 200 ~tsec, the effect of the rapid
load decrease due to brittle fracture is more accurately represented by idealized impulses
of 10 to 40 ~tsec duration. For these impulse durations, vibrational energies in the 0.5 to
1.5 J range are expected. This explains most of the observed difference for low energy
specimens which exit the back of the test machine.
Pendulum vibrational energies were computed for finite element simulations in which
force histories from each of the three NIST specimen types were used as input. With
impact at the CP, the vibrational energy left after impact was 0.26 J, 0.41 J, and 0.64 J
for the low, high, and super high specimens, respectively. These results are reasonably
consistent with the results of Figure 2. Impacting at 1% above the CP led to 0.14 J, 0.77
J, and 1.4 J, respectively, and as predicted in Figure 2, the low energy specimen
Figure 2-Pendulum Vibrational Energies for Various Impulses and Strike

Locations (Fmax =620 kN) (top) Vibration Energy Magnitude Versus
Impulse Duration (bottom) Normalized Vibration Energy Versus Impulse
Duration
vibrational energy decreased while the other two increased. It was found that artificially
shortening the unloading time associated with the brittle fracture of the low energy
specimen (5 ~tsec instead of 20 txsec) raised the vibrational energy from 0.26 J to 0.44 J.
It was concluded that the pendulum vibrational energy is an important factor in the
observed encoder versus striker energy differences. The finite element simulations
showed that the pendulum vibrational energy effect typically increases the encoder based
energy relative to the integrated striker energy by 0 to 1 J, but could perhaps increase the
energy by as much as 2 J in some cases. This effect therefore helps to explain the
observed 2 to 3 J difference between the encoder and integrated energies for the brittle
specimens, but cannot explain the entire difference. The vibrational energy effect does
not help to explain the observed energy differences for the more ductile specimens since
for those specimens the integrated energy is larger than the encoder-based energy.
Further, most of the vibrational energy for a ductile specimen is transferred from the
hammer to the specimen as a result of plasticity damping.
Striker Calibration
Calibration of the striker load cell is done using a static loading. Therefore, it is
assumed inherently that static loading of the striker results in similar strains in the striker
as dynamic loading. Using finite element simulations, it was shown that inertial effects
would not significantly reduce the load cell accuracy even for such dynamic events as
unloading due to brittle fracture and ringing of the specimen during initial loading. This
section addressed the effect of changes in contact pressure distribution on the striker.
Questions of interest included:
9 How sensitive is the load cell to contact pressure distribution?
9 Do geometric imperfections of specimens that meet ASTM E23 lead to significant
differences in the contact pressure distribution?
9 Do deformations of the striker and pendulum assembly lead to significant differences
in the contact pressure distribution?
9 Do large specimen deformations lead to significant differences in contact pressure
distribution as a result of pendulum rotation?
9 Does pinching of the striker tip by very ductile specimens alter the indicated load?
9 How sensitive is the load cell calibration to alignment of the calibration load frame?
Initially, contact with the specimen is limited to the center line of the striker contact
surface. As the specimen deforms, the contact pressures become more evenly spread
over the contacting surface. For very ductile specimens, the comers of the ASTM E23
standard striker contact surface can embed themselves in the specimen to the extent that
further bending of the specimen causes a pinching of the leading edge of the striker. This
range of potential contact pressure configurations was simulated by three idealized
contact pressure distributions. The first was for a line contact at the striker centerline
that extends the entire specimen height. The second was for a uniform pressure over the
entire nominal contact region. The third was for two lines of contact at the outer edges
of the contacting surface.
Forces for this first part of the contact pressure study were uniform with regard to
vertical position on the contacting surface. It was found that the indicated load was less
than the applied load by 0.4%. The indicated load was larger than the applied load by
0.8% when the uniform contact pressures were replaced by forces concentrated at the
edges. The static calibration is performed with a specially machined and hardened
calibration specimen. Due to the lack of plasticity in the calibration specimen, the
contact pressures during calibration are largely concentrated near the striker contact
surface centerline. This means that the calibration is most applicable to specimens with
little plasticity before fracture. The indicated load error for a very ductile specimen then
would be less than 1.2%. The trend for larger indicated forces with more ductile
behavior is consistent with the observed differences between eneoder-based energy and
integrated energies from the striker force history (an error in load cell indicated force
results in a nearly proportionate error in the integrated energy). A 1.2% indicated load
error for the high energy specimens would lead to a 1 to 2 J energy error. A 1.2%
indicated load error for the super high energy specimens would lead to a 2 to 3 J energy
error. These estimated upper bound energy errors are slightly smaller than the 3 to 4 J
difference noted above for high and super high specimens.
The next phase of the contact pressure distribution study examined the effect of force
variations in the vertical contact surface direction (along the striker surface parallel to the
notch). Such variations can arise from many sources, including: non-square specimen
cross-section; elastic deformation of the striker due to the impact load; elastic bending of
the pendulum due to the impact load; rotation of the pendulum during a ductile test; and
misalignment and imperfect machine geometry. The ASTM E23 test standard states that
the machining tolerance for specimen squareness is 10 minutes for adjacent sides. The
striking edge is to be parallel to a perfectly square specimen to within 1:1000 (3.4
minutes). The anvil is to be perpendicular to the specimen support to within 9 minutes.
The finite element model of Figure 1 was used to compute rotation of the striker
contact surface during numerous impulse loadings. Figure 3 summarizes the contact
surface rotational behavior as a function of the impulse duration. The peak force was
620 kN (7000 lb) for all analyses. A static load of 620 kN was found to induce a 0.0022
radian (0.13 degrees or 8 minutes) rotation of the contacting surface due to elastic shear
and bending deformation in the striker. This 8 minute rotation due to striker deformation
is comparable to the ASTM E23 specimen machining tolerance of 10 minutes. Positive
rotations were arbitrarily defined so as to promote loss of contact at the bottom of the
contact region. For the very shortest impulse loadings, inertial effects limited the rotation
of the contact surface. For impulse durations between 20 and 100 ~tsec, the contact
surface rotations were found to be comparable in magnitude to those from a static
loading and are again primarily due to local elastic deformation of the striker. For
impulse durations greater than about 200 ~tsec, the contact surface rotation due to the
overall rotation of the pendulum is larger than the rotation due to local elastic
deformation of the striker. For the longest impulse load duration of 5000 j.tsec, the
pendulum rotation is about -0.03 radians (1.7 degree or 100 minutes). This is about 14
times the rotation due to striker deformation and is opposite in direction. The low energy
specimens have a total impulse time between 100 and 200 ~sec, while the high and super
high energy specimens have impulse durations between 3000 and 5000 ~tsec.
The range of striker contact region rotation due to local deformation (indicated by
curves with circular and triangular points in Figure 3) reflects the dependence of the
Tup Rotation
0.005
0.000
[=
'u
g
t-
-0.005
-0.010
-0.015
\ I
<
I
0
o
-0,020
-0.025
~ o
Ill
tup contact zone maximumrelativeto penduluml
[ k t
tup contact zone ntnimumrelativeto pendulum II
-0.030
maximumpendulumrotation I
-0,035
1
I ll]
10
I I lllllll
100 1000
I
10000
Impulse Duration (microseconds)
Figure 3-Striker Contact Surface Rotation Due to Impulse Loads of Various Duration
(Fmax = 620 kN)
Uniform Load Over a Fraction of the Nominal Contact Region
20
10
E 0
-10
8 -20
,.J
-30 /
A
I contactat top of contact zone
-40 1 a contactat bottomof contact zone
-~0 ' l I I
0 0.2 0,4 0.6 0.8 1 1.2
Contact Area/Nominal Area
Figure 4-Striker Load Cell Error Due to Uniform Contact Pressure Distributions that
Cover Less than the Full Nominal Striker Contact Region
rotations on time. It was found that the dynamic rotations could not be correlated with
those for static loading based on instantaneous load levels. This means that inertial
effects on rotation behavior are significant and that any effect of rotations on dynamic
load cell behavior will not be automatically corrected by rotations that occurred during
the static calibration procedure. Due to the effect of contact surface rotations on load
cell behavior, it was considered necessary to assess the effect that force distribution
variations, such as might be induced by striker contact surface rotations, would have on
load cell accuracy.
The baseline contact pressure distribution was a uniform contact pressure over the
entire nominal contact area. Two types of contact pressure variations were simulated
with the detailed striker finite element model. The first was a linearly varying
distribution that covered the entire nominal contact region. The pressure was maximum
at either the top or bottom edge of the contact zone and decreased linearly to zero at the
opposite edge. The second v~'riation simulated a uniform pressure distribution that
covered only a portion of the ." ,~ainal contact region. For this case, the contact region
width was varied from a line contact at either the top or bottom edge to a case with
contact over 75% of the nominal contact region. It was found that the linear distribution
cases covering the entire nominal contact region resulted in an indicated load about 1%
above the actual load when the peak pressure was at the bottom of the contact region and
resulted in an indicated load about 1% below the actual load when the peak pressure was
at the top of the contact region. By linear superposition of these linear pressure
distributions with the baseline uniform pressure case, it can be concluded that any linear
pressure distribution will result in load cell accuracy of 1% or better.
Figure 4 shows the results of the load cell simulations that considered the effect of
uniform contact pressures existing over a portion of the full nominal contact region. This
type of distribution is representative of behavior in which loss of contact occurs at either
the top or bottom edge of the striker but plasticity in the specimen limits the peak contact
pressure and distributes the contact force over a finite area. The baseline contact
behavior (uniform pressure over entire nominal contact region) is represented by the
common point at the right ends of the two curves. Surprisingly, for the case in which the
contact region covers 75% of the nominal contact area, the indicated loads are 6 to 8%
above the actual applied load. For the case in which the contact region covers 50% of the
nominal contact region, the indicated loads are within 1.5% of the actual load. The
errors increase in magnitude as the contact regions are further reduced, with errors for
line contacts at the top and bottom edges of the nominal contact region being -46 and
-41%. The limiting case of a line load will not be reached for a real specimen since
plasticity tends to spread the load over a finite width region.
The tendency in Figure 4 for indicated loads to be larger than actual loads for contact
areas larger than 50% and smaller than actual loads for the areas less than 50% could
help to explain the observed differences between encoder-based energy and the
integrated energy. Loss of contact due to contact surface rotation would be expected to
be more dramatic (i.e., smaller contact area) for specimens exhibiting little plasticity.
The behavior of Figure 4 would therefore be consistent with the observed differences if
the low energy specimens had effective contact areas less than 50% and the higher
energy specimens had effective contact areas greater than 50%.
As described above, since contact pressure distributions and rotations of the contact
surface can have a significant effect on the load cell behavior, static calibration
procedures should produce similar behavior as the dynamic loading that occurs during
testing. The striker load cell was found to be relatively insensitive to lateral pressure
distribution variations and therefore, it was concluded that calibration procedures could
use a carefully machined and hardened calibration specimen that tends to concentrate
loads at the striker centerline. In applying loads to the striker during static calibration,
much care must be taken that the applied loads are aligned with the striker contact
region. Any offset results in differences between the applied load and the striker reaction
force due to reaction forces at the pendulum's axis of rotation. Perhaps even more
importantly, however, is that an offset will affect the amount of contact surface rotation
and thus the contact pressure distribution. A numerical simulation found that applying a
calibration force to the hammer at 44.5 mm above the CS led to a difference in applied
force and striker reaction force of 5.2% and increased the contact surface rotation by a
factor of 4.6 (34 minutes instead of than 8 minutes). Applying the force at 13 mm below
the CS resulted in a 1.5% force difference and the contact surface rotation was -2
minutes (direction reversed from perfectly aligned calibration force). It was concluded
that control of the applied calibration force location to within 3 mm of the CS results in
applied force errors and additional contact surface rotations that are tolerable.
Energy Integration
The force versus time record obtained by the instrumented striker can be used to
compute work done on the specimen only if the specimen displacement is also known.
Displacements can be calculated if the pendulum velocity at the instant of first contact
with the specimen is known. This is done by using the measured force to compute the
deceleration history via Newton's second law (a = F/m), integrating the deceleration to
get the velocity history, and then integrating again to get displacements. The initial
velocity enters as a constant of integration. The initial velocity is obtained by equating
the kinetic energy just before impact to the 'initial potential energy of the pendulum less a
windage and friction correction ( 1/2mv2 = mgh - E - E f ) and solving for velocity. This
procedure involves two assumptions. It is inherently assumed that the pendulum and
striker are rigid since displacement at the specimen contact is assumed to be the same as
that at the pendulum's CS. The second assumption is that the pendulum is struck at its
CP.
Striking the pendulum at the CP makes the mechanics equations relating impact force,
pendulum deceleration, velocity, and kinetic energy very convenient since the equations
reduce to those for a point mass (as outlined above). The mass that is used in these
simplified equations is a pseudomass (not the true mass). It is measured by weighing the
pendulum at its CP while the pendulum is horizontal and supported at its pivot point. If
the impact is not at the CP, the simplified equations begin to entail some error since it is
then necessary to consider mass moments of inertia (/) and angular velocities. The
magnitudes of the errors associated with using the simpler equations are governed by the
ratio rcJrcp, where rc~ is the distance from the axis of rotation to the CS and rcp is the
distance from the axis of rotation to the CP. The ASTM E23 standard states that this
ratio must be between 0.99 and 1.01. A study was done to assess the effect of not having
a rigid pendulum and striker and of having the CS different from the CP.
Interestingly, it was found that while many of the mechanics relations are affected by
not having the CS at the CP, the relationship between pendulum angles before and after
strike and the total energy loss remains perfectly accurate provided the pseudomass is
obtained by weighing the pendulum at its CS and not at its true CP. The initial impact
velocity on the other hand is affected by not striking at the CP. The actual initial velocity
is ~ times the initial velocity resulting from application of the simple equations.
Due to the square root and the ASTM E23 +1% limitation on rcs/rcp,the maximum initial
velocity error associated with use of the simple equations is 0.5%. The actual
deceleration of the pendulum is equal to rdr~p times the deceleration resulting from
application of the simple equation (F ~ ma). Since initial velocity and accelerations are
affected by rJr~p not being equal to unity, the energy resulting from integration of the
instrumented striker force data will also be affected.
The effect of the rJr~p corrections on Charpy energies obtained by integrating striker
force data is to increase the computed energy if striking below the center of percussion
and to decrease the computed energy if striking above the center of percussion. The
effect of the acceleration correction is to offset the effect of the velocity correction.
Since the acceleration correction is smallest for low energy specimens, the largest
integrated energy error is found for low energy specimens. Through numerical
experimentation, it was found that if the CS is within 1% of the CP, the upper bound
energy error associated with using the simple equations would be 0.5%. However, since
the locations of the CS and CP can be easily determined, the above velocity and
acceleration corrections can be applied and even this small error can be avoided.
Next, a study was made to determine the effect of striker and pendulum deformations
on the displacements used in the energy integration. Finite element calculations showed
that actual dynamic striker contact surface displacements could differ from idealized
rigid body displacements by as much as 10 to 20% (see Figure 5). It can be seen from
Figure 5 that the displacement difference reaches a maximum near peak load and then
decreases to zero as the load decreases. A load versus displacement behavior that results
from correcting the rigid pendulum displacements with deformations based on stiffness
from a static finite element analysis is shown in Figure 5. It can be seen that the
correction based on static deformation behavior tended to be too large compared to
displacements from a dynamic finite element simulation. More importantly, however, it
was found that the static displacement correction made no difference in the total
integrated energy. This is due to the correction displacement being precisely
proportional to the load. When actual dynamic finite element displacements were used
in the energy integration, there was an effect on the integrated energy. Therefore it can
be concluded that this energy effect of deformation related displacements is entirely due
to inertial effects. It was found that using the idealized rigid pendulum displacements
always led to the integrated energy being larger than if actual dynamic displacements
were used. Table 1 shows the increase in integrated energy that resulted when the
idealized rigid pendulum displacements were used instead of actual dynamic
Low EnergySpecimen(AL-106)
35
A
:~2o
v
t. Zi/ I. . . . . . . . . . . . . .
t0
5-
[ ...... ~p ntco,ion I ' ['~.

0.0 0.2 0.4 0.6 0.8 1.0 1.2
Olsplacemnt(ram)
Figure 5-Effect of Pendulum and Striker Deformation on Specimen Load

versus Displacement History
Table 1-Additional Integrated Energy when the Striker

and Pendulum are Assumed to be Rigid
Strike Energy Error near Energy Error for
Specimen Location Peak Load (J) Entire Test (J)
AL-106 0.2" below CP 1.1 0.5
AL-106 at CP 1.2 0.4
AL-106 0.35" above CP 0.8 0.4
AH-147 0.2" below CP 1.2 0.8
AH-147 at CP 1.2 0.8
AH-147 0.35" above CP 1.0 1.5
SH-398 0.2" below CP 1.8 1.3
SH-398 at CP 1.8 1.3
SH-398 0.35" above CP 1.3 2.4
displacements. Energy effects of using rigid body displacements in the integration range
from about 0.5 J to 2.5 J, with higher energy specimens having the higher energy
differences. Except for higher energy specimens with impact significantly above the CP,
the energy difference around peak load is larger than the final energy difference. While
the static deformation correction did not affect the overall integrated energy, it did result
in a difference for energies up to peak load. Therefore, the static deformation energy
correction could be useful if the total energy is to be partitioned.
Work at the Bearing
Energy loss at the bearing can be of two forms. The frictional loss is commonly
corrected along with windage loss. Bearing friction and windage loss are reasonably
determined by measuring the losses during a swing of the pendulum without a specimen
and then scaling the loss based on the ratio of the swing arc during an actual test to that
from the free swing arc. Losses also can occur as the result of radial play at the bearings.
This second type of energy loss will not occur during a free swing because there are no
vibrations in the pendulum. Theoretically, if the pendulum was rigid and the impact was
at the pendulum's CP, still there would be no loss at the beatings due to radial play since
no forces would be generated at the bearing. However, the pendulum is not rigid, and
significant forces can be generated at the bearing due to the impact even if the impact is
at the CP. Finite element simulations have found vibrational bearing reactions in the 0.4
to 2 kN range. This range of forces acting through the ASTM E23 maximum allowable
radial play of 0.075 mm gives an energy dissipation of 0.03 J to 0.15 J. Since this force
is cyclic, each cycle will involve additional energy loss. A 4000 btsec test (representative
of super high energy specimens) might involve as many as two full cycles at the bearing
thus resulting in four times the above energies being dissipated (0.12 J to 0.6 J). A low
energy specimen test takes so little time that less than half a cycle occurs, and thus the
above energies (based on a half cycle) would be reduced. These energy losses are
conservative if bearing radial play is less than the allowable. These energy losses would
be included in the Charpy energy if the encoder was being used. Obtaining Charpy
energy from integrating forces from an instrumented striker would not include the
bearing related energy loss.
Geometry Effects
Imperfect specimen geometry or imperfect specimen position is generally expected to

result in more opportunity for secondary impacts between the broken specimen halves
and the pendulum. Any such secondary impacts will affect the encoder energy, but will
not affect the energy integrated from the striker load cell. Significantly improved
agreement between encoder and striker energies was found in the present study when the
broken specimen halves were ejected from the rear (counter to striker motion) rather than
from the front. It is believed that the rear ejection path was accompanied by fewer and
perhaps no secondary strikes. The fact that there was significantly less scatter in the
encoder energies for the rear ejected specimens tends to support this conjecture.
Specimen Kinetic Energy
Kinetic energy of the broken specimen is a part of the Charpy energy and will be
reflected in both encoder energy and integrated striker based energies. However, the
kinetic energy in the broken specimen may result indirectly in energy differences
between the encoder energy and the integrated energy due to the tendency for a more
energetic specimen to have more secondary impacts with the pendulum. The amount of
kinetic energy that is left in a specimen after fracture is affected by three factors. The
first is the amount of strain energy stored in the specimen, anvils, and striker just before
fracture. The second is the speed of the striker at the onset of fracture, and the third is
the amount of energy absorbed by the fracture process.
Finite element simulations show that 4 to 7 J of strain energy exist in a standard steel
specimen for loads in the 23 to 31 kN range. The stored strain energy is proportional to
the square of the load at fracture. Strain energy in the pendulum is about I J. Although
not computed, strain energy in the anvils is probably no more than 1 J. All of this strain
energy (5 to 9 J) is available for conversion into specimen kinetic energy upon fracture,
but, as in most energy conversion processes, the conversion is not 100% efficient. As
addressed above, much of the strain energy in the pendulum remains in the pendulum as
vibrational energy. This probably also is the case for the anvil. The strain energy in the
specimen just prior to fracture gets converted into translational kinetic energy, rotational
kinetic energy, vibrational energy, and heat energy (due to fracture processes, plasticity,
and friction).
The speed of the pendulum at fracture affects specimen kinetic energy since the
central portion of the specimen is going at the same speed as the pendulum just before
fracture. Since low energy specimens slow the pendulum the least, the speed transfer
effect is greatest for the low energy specimens. The kinetic energy imparted by this
transfer of pendulum speed to the specimen depends on whether the specimen breaks. If
the specimen breaks and no spring action due to the release of strain energy is included
(i.e., coefficient of restitution is zero), the combined kinetic energy of the specimen
halves was estimated to be less than 0.2 J. If the specimen does not break and no spring
action is considered, the estimated specimen kinetic energy increases to 0.6 J. A
coefficient of restitution of 1 (elastic impact) results in the final specimen velocity being
about twice the pendulum velocity thus resulting in the 0.2 and 0.6 J values increasing to
0.7 J and 2.6 J. The tossing energy experiments of Chandavale and Dutta [1] that used
prebent specimens to measure energy loss associated with tossing unbroken specimens
found an energy of 1.9 J. This value implies a coefficient of restitution of about 0.57
which means that about 1.5 J of the total loss was transferred to the specimen as kinetic
energy and 0.4 J was dissipated via other mechanisms (e.g., heat, vibration). This
experiment is therefore consistent with the above estimated specimen kinetic energy
range of 0.6 to 2.6 J. It should be noted however, that the tossing energy experiments of
[1] do not involve the large amounts of stored strain energy that exists in a real Charpy
test (there being no reactions from the anvils) and therefore the measured tossing
energies are probably the lower bound of tossing energies for actual tests.
The energy absorbed by fracture is believed to have the most significant effect on the
specimen kinetic energy since energy absorbed by fracture processes is not available for
conversion to kinetic energy. For a very low energy specimen, the energy absorbed by
the fracture process has been estimated to be less than 1 J. This is just a small portion of
the 5 to 9 J of strain energy that was identified above, and therefore the potential for large
kinetic energies is high. For example, if 5 J of strain energy is converted into pure
translational specimen kinetic energy, the specimen halves would have a velocity of 15.2
m/sec (about 2.8 times the pendulum velocity prior to impact). For very ductile
specimens, the potential for energy loss due to plasticity essentially is unlimited.
Specimen velocities in this case are limited largely to the pendulum velocities at the end
of the fracture event.
Summary and Conclusions
The impetus behind this study was that low energy Charpy specimens resulted in
encoder energies 2 to 3 J greater than the instrumented striker's integrated energy, while
high and super high energy specimens resulted in encoder energies 3 to 4 J less than the
instrumented striker's integrated energy. It is concluded that the observed differences
between the encoder and striker integrated energies are due to a combination of factors.
Table 2 summarizes the findings of this study in terms of six phenomena. The extent to
which each affects the Charpy energy is summarized, and the extent to which each
phenomenon is consistent with the observed experimental behavior is noted. The results
have shown that the observed differences in energies are due primarily to a combination
of the first four phenomena (vibrational energy, secondary impacts, partial loss of
specimen contact through its effect on load cell accuracy, and pendulum deformation
through its effect on displacements used in the integrated energy). The order given in
Table 2 is believed to be the order of significance for the lower energy specimens. For
the high and super high energy specimens, it seems likely that the pendulum deformation
and partial loss of contact phenomena may be more important than secondary impacts
and vibrational energy.
Reference
[1] Chandavale, R. G. and Dutta, T., "Correction of Charpy Impact Values for
Kinetic Energy of Test Specimens", Pendulum Impact Machines Procedures and
Specimens for Verification, ASTM STP 1248, T. A. Siewert and K. Schmieder,
Eds., American Society for Testing and Materials, Philadelphia, PA.
Acknowledgments
The authors are grateful to NIST for providing the test specimens used in this study.
Table 2 -Summaryof CharpyEnergyEffects
phenomenon energy effect notes
vibrational energy 9 can add 0.5 to t.5 J to encoder energy 9 most significant for low energy specimens
9 does not affect striker integrated energy 9 partially explains observed energy difference for low energy
specimens
9 is not consistent with observed energy difference for high energy z
specimens -i-
z
secondary impacts 9 can add 0.5 to 1 J to encoder energy 9 most significant for low energy specimens
9 partially explains observed energy difference for low energy z
9 does not affect striker integrated energy cJ
specimen 0o
-I
9 is not consistent with observed energy difference for high energy 0
z
specimen m
09
pendulum 9 no direct effect on encoder energy 9 2% (~0.5 J) error for low energy specimens m
deformation
9 0.5 to 1 J increase in integrated energy due to rigid 9 0.5% error (~1 J) for high and super high energy specimens 0
pendulum assumption z
9 consistent with observed difference for high and super high energy
specimens 0
r---
contact surface 9 no effect on encoder energy 9 <50% loss of contact leads to higher indicated load
rotation/
9 as much as -20% to +8% error in striker loads 9 >50% loss of contact leads to lower indicated load
separation 0
9 may be consistent with observed energy differences for both low
and high energy specimens rn
CP not at CS 9 no direct effect on encoder energy 9 not a significant factor in current experiments since CS very near m
CP z
m
9 <0.5% error on striker integrated energy
9 increases vibrational energy effects .<
work at bearing 9 < 0.5% error in encoder energy based on maximum ASTM E23 radial beating play and peak
vibrational forces at the bearing from finite element analyses with
9 no effect on striker integrated energy
CS not at CP t.D
.,.,4
Toshiro Kobayashi, 1Naoya Inoue, 1 Shigeki Morita, 1 and Hiroyuki Toda
On the Accuracy of Measurement and Calibration of Load Signal in the

Instrumented Charpy Impact Test
Reference: Kobayashi, T., Inoue, N., Morita, S., and Toda, H., " O n the Accuracy
of Measurement and Calibration of Load Signal in the Instrumented Charpy
I m p a c t Test," Pendulum Impact Testing: A Century of Progress, STP 1380, T. ,a~.
Siewert and M. P. Manahan, Sr., Eds., American Society for Testing and Materials,
West Conshohocken, PA, 2000.
Abstract: Significant variation in measured load of an instrumented Charpy impact test

is observed if specimen materials or sizes are changed from the standard specimen. Load
calibration factors for different materials and sizes of specimens are evaluated by means
of both experiments and finite element analysis. Elastic finite element analysis shows
that the strain fields in the instrumented striker in contact with different materials and
sizes of specimens are not exactly the same as those for the standard specimen. As a
result of experiment and analysis, the decrease in stiffness or size of specimens results in
a decrease in the calibration factor. Those are attributable to strain localization near the
region in which the strain gage is attached. Therefore, the measured load for low-
stiffness materials or miniaturized specimens is overestimated when the decrease of
calibration factor is disregarded. Those results strongly suggest that the system must be
calibrated for the different materials and sizes of specimens to obtain accurate toughness
data.
Keywords: load calibration, instrumented Charpy impact test, fracture toughness, finite
element analysis
Introduction
One of the authors has already developed the CAI (Computer Aided Instrumented
Impact Testing) System, where dynamic fracture toughness parameters are obtained
simply from the analysis of the load-deflection curve of a single precracked specimen
[1]. In the test, therefore, it is important to record an accurate impact load. Generally one
can obtain the measured load in the instrumented Charpy impact test by multiplying the
strain signal from attached strain gages on the instrumented striker by a load calibration
factor assuming a linear relationship between the strain gage signal and applied load.
Recently, the instrumented Charpy impact test is used for the evaluation of toughness of
many kinds of materials such as steel [2], aluminum alloys [3], polymer and an alloy for
Professor, research associates, and associate professor, respectively, Toyohashi

University of Technology, Toyohashi, Aichi, Japan, 441-8580.
198
KOBAYASHI ET AL. ON LOAD SIGNAL 199
dentistry [4] and with miniaturized specimens [2,5]. In those cases, a significant
variation in the calibration factor has been reported because the Charpy specimen was
changed from the standard steel specimen to another material or geometry [6,7]. Though
a lot of methods of load calibration are proposed [8,9], there is no report taking into
consideration the change of material or geometry of the specimen. The elucidation of the
mechanism that the load calibration factor changes by material or geometry of the
specimen is important to measure accurate impact load and to enact the standard of load
calibration method.
This paper describes the result of the load calibration test when the material and
geometry of the specimen is changed from the standard steel specimen. Then, the results
of the strain distribution in the instrumented striker simulated by finite element analysis
explain the mechanism of the change in a load calibration factor. Moreover, the
influence of the position of the strain gages on the measured load is evaluated.
Experimental and Analytical Procedures
Load Calibration Test
Generally one can obtain the measuring impact load in the instrumented Charpy
impact test by multiplying the strain signal from the attached strain gages on the
instrumented striker by a constant load calibration factor. Therefore, the load calibration
test was carried out in order to determine a load calibration factor from the relationship
between applied load and strain signal from the attached strain gages on an instrumented
striker. Figure l(a) is a schematic drawing of a specimen for the load calibration test.
Such materials, ASTM A508 class 3 steel for reactor pressure vessels, tool steel
(Japanese Industrial Standards number JIS SKD40), ductile cast iron (JIS FCD400),
ASTM 6061 aluminum alloy and ASTM 5083 aluminum alloy were used in order to
evaluate the material effect of a specimen on a load calibration factor. Table 1 shows the
Young's modulus of materials used in this work. Moreover, the thickness of the
specimen made of A508 steel was changed to 10, 5, 3, and 1.5 mm in order to evaluate
the effect of the specimen sizes on a load calibration factor. An instrumented Charpy
impact test machine with 98J capacity was used for the load calibration.
Figure 2 shows a schematic drawing of a load calibration test. The semi-conductor
strain gages were attached on both sides and 16 mm position from the tip of
instrumented striker. Figure 3(a) shows the circuit diagram of the instrumented striker
Figure 1 - Schematic illustrations of (a) the

specimen .for the load calibration and (b)
the specimen for fracture toughness test.
Table 1 - Young's modulus of materials

used for the specimen for the load
calibration.
Material Young's modulus, GPa
A508 210
SKD40 206
FCD400 150
6061 70
5083 68
for measuring the impact load. The semi-conductor strain gages were connected as two
active bridge circuit. This bridge circuit is two times as sensitive as a single active
bridge circuit [1]. The instrumented striker was loaded with the compressive load P
through the specimen under the static condition. From the linear relationship between
the applied compressive load P and the output voltage V from the bridg e circuit, the load
calibration factor C can be calculated as
C=ae/AV (1)
In this calibration test, 6 kN and 3 kN of maximum compressive load were applied in
the case of where the specimen thickness 10 to 3 mm and 1.5 mm, respectively.
Load cell ~
Figure 2- Schematic drawing of the load

calibration test for the instrumented Charpy
test in accordance with JIS(japanese
Industrial Standards) B7755-1993, impact
testing machines for metallic materials
instrumentation.
Instrumented Charpy Impact Test
Figure 1(13) shows a three-point bending specimen for the instrumented Charpy
impact test. The material of the specimen was 5083A1 and the thickness B and the width
Wof the specimen was 10 mm. The specimen was notched with the standard V-notch of
2 mm deep. The impact loading velocity Vo was selected as 3.7 m/s. It satisfies the
energy condition of equation (2) in order to make that the hammer speed reduction
should be less than 20% during fracture of the specimen [10].
Eapp _>3E, (2)
where Eapp is applied energy and Et is total absorbed energy.

Figure 3 shows the diagram of the CAI system used in this work [1]. The high-
speed amplifier (Figure 3(c)) amplified the load signal and displacement signal. Then
the displacement signal was transmitted to the A/D converter (Figure 3(d)) built in a
computer. The frequency response and the maximum sampling points of this system are
800 kHz and 32,768 words, respectively. It is possible to obtain the measuring load by
multiplying the load signal by a load calibration factor and the displacement by
multiplying the displacement signal by a displacement calibration factor.
Figure 3 - Schematic diagram of the CAI (Computer

Aided instrumented Charpy Impac0 testing system.
Finite Element Analysis
To calculate strain fields in the instrumented striker when specimens of different

materials or sizes contact it, the ANSYS non-linear finite element code was used for
those analyses. Figure 4(a) shows a whole finite element model of the instrumented
Charpy hammer and the specimen for the load calibration test. The hammer arm was
disregarded in the model. The model was three-dimensional 1/4 size using the
symmetric condition. Eight-noded brick elements were used in modeling the hammer
and the specimen: The total number of the elements is 2,984 in those models. Figure
4(b) shows an enlarged view around the instrumented striker. Especially, the elements
were refined in the regions of a tip of the striker and the strain gage point. The contact
Figure 4 -Finite element models of the

Charpy hammer with a specimen for the
load calibration. (a) Whole elements and
(b) an enlarged view around the
instrumented striker.
elements were formed at contact points between a surface of the instrumented striker
and the specimen to calculate contact forces. The contact element that is a tetrahedral
consists of the target surface being a triangular element on the surface and the contact
node on the other surface. The normal contact forces f, in a direction normal to the target
surface is given by the Penalty function method [11] as
f,= fKog if g~O

if g >0 (3)
K, = nE
where K, is contact stiffness, g is gap size between the target surface and the contact
node, E is Young's modulus, n=0.1 is contact stiffness coefficient decided by the
preliminary analysis for satisfying convergence and analysis cost. It occurs for the
Penalty function method that the contact node overlaps with the target surface. In this
analysis, maximum penetration size was 94 ~tm when an applied stress was 0.5 kN. And
the tangential forces fs in a direction parallel to the target surface is given by elastic
Coulomb friction theory as
L=-~L
=L if sliding (4)
~f~fs= nf L if sticking
where nf= static friction factor, /~s/dynamic friction factor, /zd. Full-Newton-Raphson
method was used for the convergence calculation. Two specimens thickness were
analyzed 10 and 3 mm. The instrumented striker and the specimen were modeled as a
linear elastic solid. The elastic modulus of the instrumented striker was taken to be 210
GPa. The elastic modulus of the specimen used were 210, 70 and 45 GPa. Maximum
applied load Pm was 6 kN which was equal to the experimental load calibration test in
the case of specimen thickness of 10 mm.
Results and Discussion
Effect of the Load Calibration Factor on the Experimental Results
Figure 5 shows the relationship between Young's modulus E of the specimen and
the load calibration factor C when applied load is 6 kN. In this figure, C is normalized
with the load calibration factor for the specimen made of A508 steel C~o8. It seems that
C linearly decreases with E of the specimen. In those calibration tests, the load
calibration factor for the specimen made of FCD400 CFCD400and 5083 CA5o83decrease
3% and 8% compared to CA5o8,respectively. Although this cause must be investigated
more in its detail hereafter, shape of contact points(surface) between specimen and
striker may be affected according to Young's modulus of specimen and results in a
change of the strain distribution near the gage position. Figure 6 shows typical impact
load - displacement curves with V notched 3-point bend specimen made of 5083
aluminum alloy at 225 K. Those results of Figures 5 and 6 mean that the curve (b) which
is applied to CA5o8gives an overestimate of 8% for the total absorbed energy and impact
load when the material effect on a load calibration factor is disregarded. Figure 7 shows
1.00 AS08 o
a$ F C D 4 ~
J SKD40
0.95
m
O
5083
0.90 ~ '
50 100 150 200
Young's modulus, E/GPa
Figure 5 -Effects of materials of the specimen
on the load calibration factor, C, for the
instrumented Charpy test. Specimen thickness
is 10 mm. C are normalized with a load
calibration factor of the specimen made of
A508, CAsos.Applied load is 6 kN.
the relationship between C and thickness of the specimen B in the case of a specimen
made of A508 steel and 6061 aluminum alloy. It is clear from Figure 7 that C linearly
decreases with thickness of the specimen, which is made of both A508 steel and 6061
aluminum alloy. C ASO8for the B=3 mm decreases 10% and C A508for the B=l.5 mm
decrease 11% compared to Cm08 for the B=10 mm. Significant variation in the
calibration factor for the instrumented Charpy impact testing machine is seen if the
specimen materials or sizes are changed. Those results of calibration tests suggest that a
change of strain field around an attached strain gage occurs complex contact mechanism
of the instrumented striker with the specimen for calibration test.
Many types of specimens of different materials and sizes from the standard
specimen made of steel are used to evaluate the properties, such as the aging
degradation of the structural materials, neutron radiation embrittlement of the nuclear
reactor and the rolled materials in the instrumented Charpy impact test. In those tests,
measured impact load is calculated on the assumption that C does not change by the
materials or sizes of the specimen. However, Figures 5 and 6 show that materials and
sizes of the specimens affect a load calibration factor, therefore the instrumented striker
must be calibrated for the different materials and sizes of specimens to obtain accurate
toughness data.
*~$1.00
(b) Applied CAsoS
~4 0.95
~2 e~
o 0.90
~ ~ ~ ~ illoy
8 0.85 I t
10 20 30 0 4 8 12
Thickness of a specimen, B / •
Displacement, d / mm
Figure 6 - Typical impact load - Figure 7 - Effects of the thickness o]

displacement curves of the V notched 3- specimens, B, on the load calibration
point bend specimen of 5083Al alloy at factor, C, of A508 and 6061AL C is
225K. The curve (b) which is applied normalized with a load calibration factor
the load calibration factor for A508 of a specimen thickness of 10 mm and
steel, CA5OSis overestimated 8%. made of A508 steel, CA5oss=I~
Effects of Materials and Sizes of the Specimens on the Load Calibration Factor
Figure 8 shows reduction rate Rc of predicted load calibration factor CFEMfrom

computed strain at the strain gage point using the finite element analysis compared to
the experimental calibration factor C. CFEM decreases 2% with change of specimen
material from A508 steel to 5083 aluminum alloy. And CFnra decreases 10% with the
Figure 8 - Predicted and experimental

reduction rate of C, Rc in the case or
comparing with the CAsosn=l~obtained by
the experiment.
thickness of specimen from 10 mm to 3 ram. It seems that Rc is underestimated for the
reasons of idealization of a hammer model and overlaps of the contact elements, but a
decreasing tendency of the predicted CFEMcorresponds to the experimental C. Therefore,
it is demonstrated from the analytical result that C decreases as much as the specimen
rigidity decreases.
Effect of the Specimen Materials on the Strain Distribution in the Instrumented Striker
Figure 9 shows contour maps of predicted strain in x direction ex within the

specimen made of (a) A508 steel, (b) 5083 aluminum alloy and (c) Mg alloy as a result
of 6 kN load. It can be seen that local strain is developed largely around the contact
point with lowering the rigidity of specimen. Figure 10 shows contour maps of predicted
strain in x direction ex within the instrumented striker at the same load step as Figure 9.
The strain ex around the tip of the instrumented striker develops under the influence of
strain localization in the specimen. Especially, it can be seen that the contour line of ex=-
0.0015 shows the characteristics of strain localization. Figures 9 and 10 indicate that the
~ -0.0015 -~-
~X
-0.0060 -0.01145
Figure 9 - Contour maps of the strain in x

direction, ex, within the specimen for the
load calibration (y=O). Young's modulus, E,
is (a) 210 GPa, (b) 70 GPa and (c) 45 GPa.
Applied load, P, is 6 kN.
mismatch of the shape of the specimen and the tip of instrumented striker is a cause of
the development of the strain. Therefore, even if the applied load is the same, the load
calibration factor C is lowered with the decrease of Young's modulus E of the specimen
due to the increase of the strain around attached the strain gage.
Effect of the Specimen Sizes on the Strain Distribution in the Instrumented Striker
Figure 11 shows contour maps of predicted strain in x direction ex withi~ the

instrumented striker in the cases of (a) B=10 mm and (b) B=3 mm as a result of a 6 kN
load. In those figures, a part of the hammer except for the instrumented striker is not
shown. A strain distribution changes complicatedly due to the decrease of the specimen
thickness. The strain is localized in the center of the instrumented striker with the
decrease from B=10 mm to B=3 mm. It can be seen that the strain is concentrated at the
corresponding region to the width of the specimen. The maximum compressive strain
emx in the case of (b) reaches 2.7 times in comparison with emxin the case of (a). In the
upper part of the instrumented striker, the tensile strain is developed with the decrease of
the specimen thickness. The maximum tensile strain in the case of (b) reaches 1,2 times
in comparison with emxin the case of (a).
The i:esults of Figure 11 show that compressive strain around an attached strain
gage is developed with the decrease of the specimen width, and the load calibration
factor is decreased in the same manner as our instrumented striker. However, the load
calibration factor may be increased, if the strain gage is attached on the upper part of the
-0.0003
(a) ~ . 0 _ 0 0 6
I /--- ~ -0.0021
" "/
-0.0003
(b) ~ - 0 . 0 0 0 6 -0.0021
-0%0
-0.0069 .,
-0.0003 //
(c) ~ 0 0 0 6 -0.0021
Y ,-0.004
:00'oo
I .x
Figure 10 - Contour maps of the Figure 11 - Contour maps of the strain
strain in x direction ex, within the in x direction ex, within the instrumented
tip of the instrumented striker striker. Specimen thickness, B, is (a) 10
(y=O). Young's modulus, E, is (a) mm and (b) 3ram. Applied load, P, is 6
210 GPa, (b) 70 GPa and (c) 45 kN.
GPa. Applied load. P. is 6 Idg.
instrumented striker. The situation of the instrumented striker differs from the standard
load cell, in which the output is the same no matter how specimen material and
geometry are changed. The experimental and analytical results indicate that the material
and the geometry of the specimen influence the strain fields in the instrumented striker,
and change of the load calibration factor. When this phenomenon is disregarded, the
measured load consequentially produces the error. Of course, sensitivity of the
instrumented striker for the change of specimen materials and sizes, it is also largely
affected by the design of the instrumented striker. In the future research, the design of
instrumented striker using a simulation, such as finite element analysis is required to
develop an instrumented striker which is less affected by specimen materials and
geometry.
Effect of the Strain Gage Position on the Load Calibration Factor
Results of Figures 10 and 11 indicate that the materials and the size of specimens
influence the strain fields in the instrumented striker, and change the load calibration
factor. It can be seen in Figures 10 and 11 that strain field in the instrumented striker
changes largely as much as to be close to the contact point with the specimen. Figure 12
shows that predicted load calibration value CFEMas a function of the distance from a tip
of the instrumented striker dx, when applied load is 6 kN. Where C is normalized with a
load calibration factor in the case of the specimen made of A508 steel, B=10 mm and dx
=16 mm; CAs0aB=l~ dx =~6. All CFEU decline as much as the decrease of dx, and the
inclination is as steep as to be close to the tip of the instrumented striker. Experimentally
it has been found that the more the instrumented striker becomes sensitive, the more the
strain gage position approaches the tip of the instrumented striker. One must be careful,
2.0
Z 1.5
1.0
0.5
~'
,~-
~
~
9
~.-~
~
~ B--3
.4508
i
B =10
10 20 30 40
Distance, d x~X 10"3m
Figure 12 - Predicted load calibration value,
C as a function o l d x. dx is distance from a tip
of the instrumented striker. C is normalized
with a load calibration factor in the case o]
A508, B=IO mm and d~=16 mm ," CASOSB=lO'd~=16
Applied load, P,, is 61dV.
therefore, that the C increases, as the strain gage position becomes closer to the tip of the
instrumented striker.
The results of Figure 12 show that the load calibration factor is largely affected by
the strain gage position. In the current standard of ASTM and ISO, there is no regulation
on the accurate strain gage position for instrumentation (sometimes 11-15 mm from the
tip is recommended). Therefore, sensitivity against a change of the materials or sizes of
specimens is different in each instrumented Charpy impact testing machine. Those
results strongly suggest that the system must be calibrated for the different materials and
sizes of specimens to obtain accurate toughness data.
Conclusions
Load calibration test for the instrumented Charpy impact testing machine was
carried out with some materials and sizes of specimen. The effect is the change of a load
calibration factor on the toughness data was investigated. The finite element analysis
was used in order to clarify the mechanism in which the load calibration factor changes.
The effect of the strain gage position on the load calibration factor was quantitatively
evaluated from the result of the analysis. The following conclusions can be drawn from
this study.
(1) In the calibration tests, the load calibration factor linearly decreases with the
Young's modulus or thickness of the specimen. The load calibration factor for the
specimen made of ductile cast iron (JIS FCD400) and 5083 aluminum alloy decreases
3% and 8% compared with that for A508 steel, respectively. And the load calibration
factor for the specimen thickness of 3 mm and 1.5 mm decreases 10% and 11%
compared with that of 10 mm, respectively. These results mean that the measured load
gives an overestimate for the total absorbed energy and impact load when the material
and the size effects on a load calibration factor are disregarded.
(2) Finite element analysis indicates that even if the applied load is the same, a
compressive strain is localized with the decrease of Young's modulus E of the specimen
due to the mismatch of the shape of the specimen and the tip of instrumented striker.
And then the strain is localized in the center of the instrumented striker with the
decrease of specimen thickness. Therefore, the load calibration factor decreases.
(3) The materials and the sizes of specimens influence the strain fields in the
instrumented striker, and make change in the load calibration factor. A compressive
strain in the instrumented striker changes largely enough to be close to the contact point
with the specimen. Of course, sensitivity against a change of the materials or sizes of the
specimen is different in each instrumented Charpy impact testing machine. These results
strongly suggest that the system must be calibrated for the different materials and sizes
of specimens to obtain accurate toughness data.
Reference
[1] Kobayashi, T., Yamamoto, I., and Niinomi, M., "Introduction of a New Dynamic
Fracture Toughness Evaluation System," Journal of Testing and Evaluation, Vol. 21,
1993, pp. 145-153.
[2] Sugiura, N., Isobe, E., Yamamoto, I., and Kobayashi, T., "Effect of Specimen Size
in Static and Dynamic Fracture Toughness Testing Reactor Vessel Steel A508cl.3,"
ISIJ International, Vol. 35, 1995, pp. 419-425.
[3] Hafiz, M. E, and Kobayashi, T., "Metallurgical Factors Affecting Impact Toughness
of Eutectic A1-Si Casting Alloy," Z. Metalkd, Vol.89, 1998, pp. 445-449.
[4] Takahashi, S., Niinomi, M., Hukui, H., Kobayashi, T., and Hasegawa, J., The
Journal of the Japanese Society for Dental Materials and Devices, Vol.15, 1996, pp.
577.
[5] Yamashita, M., Viswanathan, U. K., Yamamoto I., and Kobayashi, T., "Service-
induced Changes in the Microstructure and Mechanical Properties of a
Cr-Mo-Ni-V Turbine Steel," ISIJ international, Vol. 37, 1997, pp. 1133-1138.
[6] Marur, P. R., Shimha, K. R. Y., and Nari, P. S., "A Compact Testing System for
Dynamic Fracture Studies," Journal of Testing and Evaluation, Vol.23, 1995, pp.
267-274.
[7] Kalthoff, J. F., Walle, E. van, and Wilde, G., "Variation of the Sensitivity of
Instrumented ISO/DIN and ASTM Tups and their Influence on the Determination
of Impact Energies in Tests with Ductile Steels," Evaluating Material Properties by
Dynamic Testing, ESIS20, Edited by E. van Walle, 1996, pp. 25-35.
[8] Winkler, S., and BoG, B., "Static Force Calibration of Charpy Impactors,"
Evaluating Material Properties by Dynamic Testing, ESIS20, Edited by E. van
Walle, 1996, pp. 37-44.
[9] Wilde, G., Covic, M. and Gregor, M., "Instrumented Impact Testing in Radioactive
Hot Cells, Equation, and Software, "ESIS20, Edited by E. van Walle, 1996, pp. 89-
96.
[10] Kobayashi, T., Yamamoto, I. and Niinomi, M., "On the Accuracy of Measurement
of dynamic Elastic-Plastic Fracture Toughness Parameters by the Instrumented
Charpy Test, "Engineering Fracture Mechanics, Vol. 26, 1987, pp. 83-94.
[11] Simo, J. C., Wriggers, P., Schweizerhof, K. H., and Taylor, R. L., "Finite
Deformation Post-Bucking Analysis Involving Inelasticity and Contact Constraints,
" International Journal for Numerical Methods in Engineering, Vol.23, 1986, pp.
779-800.
C.N. McCowan, 1D.P. Vigliotti, 1 and T.A. Siewert1
Evaluation of ABS Plastic Impact Verification Specimens
Reference: McCowan, C. N., Vigliotti, D. R, and Siewert, T. A., "Evaluation of

ABS Plastic Impact Verification Specimens," Pendulum Impact Testing: A Century
of Progress, STP 1380, T. A. Siewert and M. R Manahan, Sr., Eds., American
Abstract: Valid comparison of impact test energies reported by various laboratories and
over time depends on consistent performance of impact test machines. To learn what
information verification specimens might reveal about the impact machines used to test
plastics, and to learn the critical parameters that must be controlled to minimize the
scatter in the data, we tested a batch of acrylonitrile-butadiene-styrene plastic (ABS
plastic) specimens. Both Charpy and Izod sample configurations were tested. We
concentrated on determining the variability in impact strength, but also studied the effects
of room temperature aging, exposure to sunlight, and clamping pressure (for Izod testing)
on the impact strength of the ABS plastic specimens. We conclude that: (1) the
variability in impact strength for the ABS plastic impact specimens is comparable to
metal specimens which are now used as verification specimens to test large capacity
impact machines; (2) the ABS plastic specimens did not show significant signs of aging
during our tests; and (3) clamping pressures will have to be specified for verification
testing of ABS plastic Izod specimens.
Keywords: Charpy impact testing, impact testing, Izod impact testing, plastic impact
specimens, plastic impact verification specimens
Introduction
To help ensure valid comparisons of impact data, a requirement for the use of
verification specimens was added to the Standard Test Methods for Notched Bar Impact
Testing of Metallic Materials (ASTM E 23). This occurred because the metals impact
testing community discovered that verification tests of impact machine performance were
able to detect certain energy loss mechanisms that could not otherwise be observed during
traditional physics-based measurements of machine performance (pendulum period,
i Materials Research Engr., Engineering Technican, and Supervisory Metallurgist,

respectively NIST, Materials Reliability Division, 325 Broadway, Boulder CO 80303.
210
McCOWAN ET AL. ON ABS PLASTIC IMPACT VERIFICATION 211
mass, mechanical friction, windage, etc.), and these mechanisms resulted in significant
differences between the results o f impact machines. This study evaluates the applicability
o f ABS plastic verification specimens (acrylonitrile-butadiene-styrene plastic) for testing
the low capacity impact machines used by the plastics industry. Nationally standardized
verification specimens are not available for testing the performance o f this class o f impact
machine.
Few studies on performance issues for plastics impact machines could be found. For
our purposes, the most useful report was the one used to support the precision statement
in A S T M (D 256 - 93a), Standard Test Methods for Determining the Pendulum Impact
Resistance o f Notched Specimens." This report describes a round robin that included 6
different plastics and 25 laboratories) It indicates that both the materials and the
laboratories contribute significantly to the uncertainty in the data. This report also
indicates that the impact strength for specimens produced from ABS plastic has low
variability, as shown in Table 1.
The low coefficient o f variation (CV) for the ABS plastic, which is the ratio o f the
standard deviation to the impact strength, compares well with the 0.04 values typical o f
metal impact verification specimens now being produced. So, in terms o f variability, the
ABS plastic appears to be a good candidate for use as an impact verification material.
This report presents Izod and Charpy impact data for ABS plastic specimens to
further confirm the low variability in impact strength. In addition, the effects of room
temperature aging, exposure to sunlight, and clamping pressure (for Izod testing) on the
impact strength o f the ABS plastic specimens are evaluated.
Table 1- Izod impact data summary from a previous study. 1 Only the laboratories that
tested 10 specimens are included here.
Impact Strength StandardDeviation StandardError Coefficient
Lab of
J/m (ft-lbf/in) J/m (tt-lbf/in) J/m (fi-lbf/in) Variation
1 599.97 11.24 24.02 0.45 7.47 0.14 0.04
2 557.80 10.45 9.07 0.17 2.67 0.05 0.02
4 591.96 11.09 10.68 0.20 3.20 0.06 0.02
5 580.22 10.87 13.88 0.26 4.27 0.08 0.02
6 564.74 10.58 13.35 0.25 4.27 0.08 0.02
7 639.47 11.98 17.08 0.32 5.34 0.10 0.03
9 583.96 10.94 13.35 0.25 4.27 0.08 0.02
10 587.69 11.01 28.29 0.53 9.07 0.17 0.05
11 614.38 11.51 14.41 0.27 4.80 0.09 0.02
12 584.49 10.95 19.75 0.37 5.34 0.12 0.03
13 547.13 10.25 20.82 0.39 5.34 0.12 0.04
14 545.52 10.22 19.22 0.36 5.87 0.11 0.04
23 541.25 10.14 8.01 0.15 2.67 0.05 0.01
24 547.12 10.25 8.54 0.16 2.67 0.05 0.02
Avg [ 577.55 [ 10.82 [ 15.48 I 0.29 [4.80 I 0.09 10.03 I
Material and Procedures
The Izod and Charpy samples used for this study were all produced from the same
batch of ABS plastic. They were cast in molds to produce geometries that meet the
requirements of ASTM D 256. The notches were ground, and the dimensions of the
specimens were measured prior to testing. In the case of Izod specimens, two specimens
were cast at a time. These two specimens are distinguished from one another as "gate"
and "dam" specimens, which refers to the end of the mold from which the specimens
were taken.
Izod and Charpy specimens were made in thicknesses of 3.2 and 6.4 mm (0.125 and
0.25 in.) for the study. The measured thicknesses of the individual specimens were used
to calculate the impact strength. The notch depths and radii of the samples were all found
to be within the specifications of ASTM D 256. The 3.2 and 6.4 mm Izod samples were
clamped in the specimen vice using 5 and 8.5 N.m (45 and 75 Ibf-in.) of force
respectively, applied using a torque wrench on a 12.7 mm diameter bolt (0.5 in. - 14).
All testing was performed at room temperature. The specimens were stored and tested
at 21 4-1 ~ (70 ~ The specimens were not conditioned at 50 +5% relative humidity
prior to testing. The nominal humidity of the laboratory in which the specimens were
stored and tested was low (less than 30% relative humidity).
We used a pendulum impact machine with a maximum capacity of about 20 J. The
machine was designed according to the requirements of D 256.
Results
A s - R e c e i v e d I z o d Specimens - The impact data for the as-received Izod specimens are
given in Table 2. For the 6.4 mm (0.25 in.) thick specimens, the samples from the gate
end of the mold had a higher impact strength than the specimens from the dam end of the
mold. The same trend was found for the 3.2 mm (0.125 in.) samples: the gate samples
had a higher impact strength than the dam samples. The differences due to specimen
location were anticipated, and the results show that considering dam and gate specimens
as separate groups helps to reduce unnecessary scatter for the material.
The standard deviations in impact strength for the four groups of as-received
specimens are low, confirming that the ABS plastic is a good candidate for use as an
impact verification material. The gate specimens had higher variation in impact strength
than the dam specimens, but all the as-received specimens have CV values comparable
to those of the steel specimens used for impact verification specimens (0.04 or less).
A s - R e c e i v e d Charpy Specimens - The tests results on the as-received Charpy
specimens, given in Table 3, again show low variation in the impact strength for the ABS
specimens. The 3.2 mm specimens have more variation in impact strength than the 6.4
mm specimens, but both thicknesses have fairly low CV values. A practical
consideration here is that the thicker specimens may seat more firmly against the anvils in
the Charpy test, and this reduces the scatter in the test.
Table 2- Initial lzod data for as-received impt ct specimens.

ID Cases Mean Standard Coefficient
of Variation
Impact Strength Deviation
Jim (ft.lbf/in) Jim (ft-|b|hn)
6.4 mm gate 25 222.48 (4.17)) 7.91 (0.15) 0.04 (0.04)

6.4 mm dam 25 216.78 (4.06) 4.43 (0.08) 0.02 (0.02)
3.2 mm gate 25 233.21 (4.37) 6.83 (0.13) 0.03 (0.03)
3.2 mm dam 24 190.70 (3.57) 4.44 (0.08) 0.02 (0.02)
Table 3- Data Corthe as-received Charpy impc :t specimens.

ID Cases Mean Standard Coefficient
of Variation
Impact Strength Deviation
Jim (ft-lbf/in) Jim (ft-lbf/in)
3.2 mm Charpy [25 241.54(4.53) 12.26 (0.23) 0.05 (0.05)

6.4 mm Charpy ]25 229.40(4.30) 4.65 (0.09) 0.02 (0.02)
Room Temperature Aging Data - The Izod data from the 30, 90, and 600 day aging
groups and data from the as-received 3.2 mm gate specimens are given in Table 4. There
appears to be a slight decrease in the impact strength of the specimens aged for 30 and 90
days, compared to the as-received specimens. However, the standard deviations of the
three groups overlap, and the 30 and 90 day results are almost identical. The statistics for
the three groups combined (0, 30, and 90 day) show only slightly more variation in
impact strength than each group considered separately: (1) average impact strength of
228.5 J/m, (2) standard deviation of 7.37 J/m, and (3) coefficient of variation o f 0.03.
The impact strength of the specimens aged for 600 days increased by more than 10%
(delta %) and this indicates that aging had a significant effect on the impact strength of
these ABS specimens. This result is somewhat surprising. Aging would generally be
expected to degrade the impact properties of plastics. Unfortunately, since only one
machine was used for these tests and the performance of this machine could not be
checked with verification specimens, we are not convinced that the results of the 600 day
Table 4- Izod impact data for aged specimens. The as-received specimens
are estimated to be I week old 0 aging.
ID Aging Cases Mean Standard Coefficientof Variation
(Days) Impact Strength Deviation
Jim (ft-lbf/in) J/m (ft-lbf/in)
3.2mm gate 0 25 233.21 (4.37) 6.83 (0.13) 0.03 (0.03)

3.2mm gate 30 25 225.67 (4.23) 6.27 (0.12) 0.03 (0.03)
3.2mm gate 90 24 226.60 (4.25) 6.48 (0.12) 0.03 (0.03)
3.2 mm gate 600 25 256.80 (4.81) 6.60 (0.12) 0.03 (0.03)
tests are necessarily reliable enough to accept. These tests will have to be repeated in a
more comprehensive study (with humidity control) to convincingly show the effects of
aging on the impact strength of the ABS specimens, if any, and to better document any
change in impact properties over time.
Specimens exposed to Sunlight - Izod and Charpy specimens were exposed to
sunlight on a window sill and periodically rotated for 30 days prior to testing. The impact
results, Table 5, show the average impact strength of the sunlight exposed specimens is
slightly lower than that of the as-received specimens. Again, the difference is not
significant considering the standard deviations.
Table 5- Izod Char ~act data for specimens exposed to sunlight.

ID Exposure cases Mean Standard Coefficient of Variation
Time Impact Strength Deviation
(Days) Jim (ft.lbf/in) J/m (ft-lbf/in)
Charpy (3.2 ram) 0 25 241.54 (4.53) 12.26 (0.23) 0.05 (0.05)

Charpy (3.2 ram) 30 25 238.44 (4.47) 8.54 (0.16) 0.04 (0.04)
Izod dam (3.2 ram) 0 24 190.70 (3.57) 4.44 (0.08) 0.02 (0.02)
Izod dam (3.2 mm) 30 24 189.90 (3.56) 15.62 (0.1 l) /o.o3 (0.03)
Table 6- Izod Clamping Force.
ID Cases Clamping Force Mean Standard Deviation
N.m (in-lbs) Impact Strength
I]/m (ft-lbf/in)
3.2 mm gate 1 Just touching !243.38 (4.56)
3.2 mm gate 1 1.1 (10) 217.09 (4.07)
3.2 mm gate 5 2.3 (20) 247.69 (4.64) 10.77 (0.20)
3.2 mm gate 1 3.4 (30) 232:16 (4.35)
3.2 mm gate 1 4.5 (40) 235.06 (4.40)
3.2 mm gate 4 5.7 (50) !225.44(4.22) 1.66 (0.03)
3.2 mm gate I 6.8 (60) 216.96 (4.07)
3.2 mm gate 1 7.9 (70) 215.91 (4.05)
3.2 mm gate 3 9.0 (80) 210.75 (3.95) 3.74 (0.07)
Clamping Force in Izod Tests - The force 270 l l l l l l l l l
with which the ABS specimens were 260[-

E
clamped for Izod testing had a significant 2501- 9
effect on the impact strength of the
2401-
specimens, as shown in Figure 1 and Table 6.
At low clamping forces (less than 3 N'm), 23~
0
the scatter in the results and the impact
~ o
strength increased. At clamping forces
E 21~- 0
greater than 6 N.m, it appears that the effect
0~ I I I l r 4 1 1 r
of clamping on scatter in the impact strength 20 1 2 3 45678910
decreases. Clamping Force, Nm
Figure 1- Effect of the clamping force
usedfor Izod specimens.
Discussion
The average impact strengths for the groups of ABS plastic impact specimens that
were tested ranged between about 190 and 240 J/m. Converting to impact energy, the
range is from about 0.6 to 0.8 J. This is a useful energy for a low energy verification
specimen, considering that the machines used to test plastics have low capacities. The
impact machine used for this study, for example, has a low range of about 0 to 3 J, and
weights can be added to the pendulum to increase the capacity up to about 20 J.
The variation in impact strength for the ABS specimens was low, and this is important
for a material being considered for use as a verification specimen. The variation must be
low enough to allow a small sample size to be used when testing the performance of a
machine. Typically the CV values for the ABS material were 0.04 or less, and previous
experience with metal impact verification specimens indicates that this variability would
allow a sample size of around 5. The sample size for a verification test depend on the
specific requirements of the test, but clearly the ABS plastic has low variation.
The age of the ABS plastic specimens did not have a significant effect on the impact
strength for short aging times, but the impact strength increased for the specimens that
were aged for 600 days. These results imply that the impact strength of the specimens
does not remain stable over time, but further testing is required to confirm these results.
The impact strength of the specimens did not appear to be too sensitive to sunlight
exposure. The sunlight exposure tests were somewhat simplistic, but they indicate that
the specimens could be left out on a sunlit desk for several days prior to testing without
adverse effect.
The impact strength of the ABS Izod specimens decreased significantly as clamping
pressure was increased on the specimen. This is apparently a common effect for some
plastics, so we included these preliminary tests here to help determine how much the
clamping pressure affected the impact strength of the ABS plastic specimens (D 256, note
7). The results show that clamping pressure is an important variable and would have to
be specified for ABS Izod verification specimens.
Overall, the impact strength of the ABS specimens was consistent and this makes the
material a good candidate for the production of verification specimens for low capacity
impact machines. The affect of aging on the ABS plastic will have to be considered and
there are other considerations in choosing a specimen to verify the performance of impact
machines. A verification specimen should test specific aspects of the machine
performance that are difficult or impossible to verify by static measurement alone.
Experience in verifying the performance of large capacity impact machines, used for
metals, has shown that testing hard specimens (HRC 45 or greater) is the best way to
verify that a machine is properly mounted. The very rapid loading of the machine
associated with the impact of these hard specimens is needed to reveal a mounting
problem for the machine. Softer specimens, on the other hand, are found to better show
the effects of anvil and striker condition on the machine performance. For the lower
capacity impact machines used to test plastics, factors that affect the test results and the
machine material interactions should be carefully considered to help ensure that
appropriate and useful verification specimens are chosen. It is likely that several different
types of specimens will be needed to adequately verify the performance of the low
capacity impact machines.
Conclusions
(1) The variability in impact strength for the ABS plastic impact specimens is
comparable to that of metal specimens which are now used as verification specimens to
test large- capacity impact machines.
(2) Clamping pressures will have to be specified for verification testing of ABS plastic
Izod specimens.
(3) The effects of aging and relative humidity need more study.
Acknowledgments
We are grateful to Dave Scarier of Bayer Corporation for producing the ABS impact
specimens used in this study.
References
1. Research Report RR: D20-1134, available from ASTM Headquarters, Conshohocken,

PA.
Appendix I
The Groups are identified as follows:

- i indicates initial, as-received, test group;
- a indicates aged for either 30 or 90 days (a30 or a90);
- 4 and8 indicate specimens thickness (4 = 6.4 mm or 1/4 in. and 8 = 3.2 mm or 1/8 in.);
- u v indicates exposure to sunlight.
T a b l e A I . 1 - IzodData
46 a90g8 g147 229.045 4.29l
Case Group ID J/m ~-lbffin
47 a90g8 g148 214.366 4.016
l a30g8 gl01 220.238 4.126 48 a90g8 g149 220.451 4.130
2 a30g8 g102 228.511 4.281 49 a90g8 g150 229.792 4.305
3 a30g8 g103 223.253 4.182 50 id4 dl 210.630 3.946
4 a30g8 g104 217.515 4.075 51 id4 dlO 215.647 4.040
5 a30g8 g105 225.469 4.224 52 id4 dll 214.206 4.013
6 a30g8 g106 219.864 4.119 53 id4 d12 213.459 3.999
7 a30g8 g107 230.059 4.310 54 id4 d13 213.939 4.008
8 a30g8 g108 228.511 4.281 55 id4 d14 220.558 4.132
9 a30g8 g109 228.031 4.272 56 id4 d15 216.181 4.050
10 a30g8 gll0 222.586 4.170 57 id4 d16 212.925 3.989
11 a30g8 g l l l 223.440 4.186 58 id4 d17 219.811 4.118
12 a30g8 gl12 226.002 4.234 59 id4 d18 216.448 4.055
13 a30g8 gl13 225.469 4.224 60 id4 d19 225.629 4.227
14 a30g8 gl14 215.754 4.042 61 id4 d2 2 1 4 . 1 5 3 4.012
15 a30g8 gl15 235.984 4.421 62 id4 d20 218.636 4.096
16 a30g8 gl16 229.952 4.308 63 id4 d21 219.010 4.103
17 a30g8 g117 232.354 4.353 64 id4 d22 226.323 4.240
18 a30g8 g118 222.907 4.176 65 id4 d23 223.974 4.196
19 a30g8 gl19 241.322 4.521 66 id4 d24 219.811 4.118
20 a30g8 g120 215.807 4.043 67 id4 d25 212.605 3.983
21 a30g8 g121 224.775 4.211 68 id4 d3 210.896 3.951
22 a30g8 g122 221.732 4.154 69 id4 d4 218.209 4.088
23 a30g8 g123 219.330 4.109 70 id4 d5 216.448 4.055
24 a30g8 g124 228.938 4.289 71 id4 d6 220.398 4.129
25 a30g8 g125 233.849 4.381 72 id4 d7 214.206 4.013
26 a90g8 g126 226.590 4.245 73 id4 d8 215.220 4.032
27 a90g8 g127 221.305 4.146 74 id4 d9 210.096 3.936
28 a90g8 g128 225.682 4.228 75 id8 d26 193.068 3.617
29 a90g8 g129 234.169 4.387 76 id8 d27 197.605 3.702
30 a90g8 g130 232.515 4.356 77 id8 d28 183.460 3.437
31 a90g8 g131 220.131 4.124 78 id8 d29 188.424 3.530
32 a9098 g132 232.301 4.352 79 id8 d30 194.776 3.649
33 a90g8 g133 230.913 4.326 80 id8 d31 193.762 3.630
34 a90g8 g134 224.988 4.215 81 id8 d32 183.567 3.439
35 a90g8 g135 220.932 4.139 82 id8 d33 185.435 3.474
36 a90g8 g136 222.319 4.165 83 id8 d34 188.584 3.533
37 agOg8 g137 221.625 4.152 84 id8 d35
38 a90g8 g138 230.273 4,314 85 id8 d36 187.090 3.505
39 a90g8 g139 241.856 4.531 86 id8 d37 198.940 3.727
40 a90g8 g140 215.594 4.039 87 id8 d38 196.751 3.686
41 a90g8 g141 231.767 4.342 88 id8 d39 187.677 3.516
42 a90g8 g142 230.593 4.320 89 id8 d40 188.157 3.525
43 a90g8 g143 229.579 4.301 90 id8 d41 188.958 3.540
44 a90g8 g144 230.913 4.326 91 id8 d42 185.489 3.475
45 a90~8 ~145 220.771 4.136 92 id8 d43 189.972 3.559
Case Group ID J/m ~-lbffin
93 id8 d44 191.307 3.584 139 ig8 g41 241.744 4.529

94 id8 d45 191.307 3.584 140 ig8 g42 230.524 4.319
95 id8 d46 191.894 3.595 141 ig8 g43 225.373 4.222
96 id8 d47 190.613 3.571 142 ig8 g44 227.086 4.254
97 id8 d48 197.605 3.702 143 ig8 g45 228.591 4.282
98 id8 d49 188.424 3.530 144 ig8 g46 221.780 4.155
99 id8 d50 194.563 3.645 145 ig8 g47 222.800 4.174
100 ig4 gl 226.750 4.248 146 ig8 g48 234.484 4.393
101 ig4 glO 215.647 4.040 147 ig8 g49 238.802 4.474
102 ig4 gll 214.046 4.010 148 ig8 g50 237.762 4.454
103 ig4 g12 213.512 4.000 149 ig8 g60 236.198 4.425
104 ig4 g13 214.046 4.010 150 uvd8 dlO1 192.641 3.609
105 ig4 g14 231.554 4.338 151 uvd8 d102 179.350 3.360
106 ig4 g15 223.761 4.192 152 uvd8 d103 183.247 3.433
107 ig4 g16 226.109 4.236 153 uvd8 d105 190.506 3.569
108 ig4 g17 233.048 4.366 154 uvd8 d106 187.997 3.522
109 ig4 g18 227.871 4.269 155 u v d 8 d107 192.961 3.615
110 ig4 g19 232.942 4.364 156 uvd8 d108 195.150 3.656
111 ig4 g2 231.821 4.343 157 uvd8 d109 201.075 3.767
112 ig4 g20 225.415 4.223 158 uvd8 dllO 190.666 3.572
113 ig4 g21 231.554 4.338 159 uvd8 dill 191.307 3.584
114 ig4 g22 232.301 4.352 160 uvd8 dl12 193.655 3.628
115 ig4 g23 231.340 4.334 161 uvd8 dl13 191.627 3.590
116 ig4 g24 227.390 4.260 162 uvd8 dl14 185.542 3.476
117 ig4 g25 217.729 4.079 163 uvd8 dl15 194.616 3.646
118 ig4 g3 210.843 3.950 164 uvd8 dl16 178.816 3.350
119 ig4 g4 218.316 4.090 165 uvd8 dl17 194.456 3.643
120 ig4 g5 216.715 4.060 166 uvd8 dl18 192.001 3.597 ~
121 ig4 g6 219.917 4.120 167 uvd8 dl19 182.766 3.424
122 ig4 g7 214.046 4.010 168 uvd8 d120 185.382 3.473
123 ig4 g8 215.113 4.030 169 uvd8 d121 189.812 3.556
124 ig4 g9 210.309 3.940 170 uvd8 d122 199.943 3.746
125 ig8 g26 226.910 4.251 171 uvd8 d123 186.770 3.499
126 ig8 g27 231.821 4.343 172 uvd8 d124 190.026 3.560
127 ig8 g28 243.350 4.559 173 uvd8 d125 187.357 3.510
128 ig8 g29 243.831 4.568
129 ig8 g31 236.625 4.433
130 ig8 g32 229.632 4.302
131 ig8 g33 245.058 4.591
132 ig8 g34 228.618 4.283
133 ig8 g35 230.326 4.315
134 ig8 g36 234.308 4.390
135 ig8 g37 227.908 4.270
136 ig8 g38 228.116 4.274
137 ig8 g39 243.649 4.565
138 ig8 g40 235.002 4.403
Table A1.2: Charp~r Data 45 i8 70.000 258.136 4.836

Case Group ID J/m ft-lbf/in 46 i8 71.000 252.958 4.739
1 i4 76.000 227.123 4.255 47 i8 72.000 229.205 4.294
2 i4 77.000 227.817 4.268 48 i8 73.000 249.702 4.678
3 i4 78.000 223.280 4.183 49 i8 74.000 251.090 4.704
4 i4 79.000 225.522 4.225 50 i8 75.000 246.713 4.622
5 i4 80.000 226.483 4.243 51 uv 251.000 231.447 4.336
6 i4 81.000 227.550 4.263 52 uv 252.000 242.283 4.539
7 i4 82.000 224.134 4.199 53 uv 253.000 241.429 4.523
8 i4 83.000 226.643 4.246 54 uv 254.000 223.440 4.186
9 i4 84.000 225.415 4.223 55 uv 255.000 241.535 4.525
10 i4 85.000 222.373 4.166 56 uv 256.000 221.732 4.154
11 i4 86.000 241.322 4.521 57 uv 257.000 243.777 4.567
12 i4 87.000 226.109 4.236 58 uv 258.000 249.862 4.681
13 i4 88.000 232.568 4.357 59 uv 259.000 252.104 4.723
14 i4 89.000 236.358 4.428 60 uv 260.000 243.991 4.571
15 i4 90.000 234.970 4.402 61 uv 261.000 241.749 4.529
16 i4 91.000 229.152 4.293 62 uv 262.000 236.411 4.429
17 i4 92.000 227.924 4.270 63 uv 263.000 237.585 4.451
18 i4 93.000 229.098 4.292 64 uv 264.000 247.087 4.629
19 i4 94.000 229.899 4.307 65 uv 265.000 246.019 4.609
20 i4 95.000 237.425 4.448 66 uv 266.000 237.799 4.455
21 i4 96.000 232.568 4.357 67 uv 267.000 239.721 4.491
22 i4 97.000 227.390 4.260 68 uv 268.000 223.067 4.179
23 i4 98.000 230.326 4.315 69 uv 269.000 238.706 4.472
24 i4 99.000 234.169 4.387 70 uv 270.000 229.846 4.306
25 i4 100.000 229.365 4.297 71 uv 271.000 237.372 4.447
26 i8 51.000 219.864 4.119 72 uv 272.000 252.425 4.729
27 i8 52.000 239.347 4.484 73 uv 273.000 238.706 4.472
28 i8 53.000 250.556 4.694 74 uv 274.000 227.337 4.259
29 i8 54.000 234.276 4.389 75 uv 275.000 235.450 4.411
30 i8 55.000 251.731 4.716
31 i8 56.000 248.475 4.655
32 i8 57.000 235.557 4.413
33 i8 58.000 254.880 4.775
34 i8 59.000 244.151 4.574
35 i8 60.000 207.907 3.895
36 i8 61.000 241.375 4.522
37 i8 62.000 233.155 4.368
38 i8 63.000 239.827 4.493
39 i8 64.000 245.165 4.593
40 i8 65.000 234.116 4.386
41 i8 66.000 224.134 4.199
42 i8 67.000 254.346 4.765
43 i8 68.000 238.813 4.474
44 i8 69.000 253.118 4.742
Impact Test Procedures
M. P. Manahan, Sr.l, F. J. Martin 2, and R. B. Stonesifer 3
Results of the ASTM Instrumented/Miniaturized Round Robin Test Program
Reference: Manahan, M. E, St., Martin, E J., and Stonesifer, R. B., "Results of the
ASTM Instrumented/Miniaturized Round Robin Test Program," Pendulum Impact
Testing: A Century of Progress, STP 1380, T. A. Siewert and M. R Manahan, St., Eds.,
Abstract: Two related standards arc currently being developed by ASTM Subcommittee
E 28.07.07. The first is focused on test procedures for instrumented impact testing. It is
intended that ASTM Standard E 23 will eventually reference the instrumented test
standard for tests conducted using conventional Charpy V-notch (CVN) specimens. The
second new standard covers miniaturized Charpy V-notch (MCVN) testing. The
Instrumented/Miniaturized Charpy Round Robin Test Program was established to support
development of these standards. The goal of the program is to test CVN and MCVN
specimens in the transition region and in the upper-shelf region using an instrumented
striker system. The test procedure for the round robin is prescribed in the draft standards.
A total of six specimens are being tested in the upper-shelf region and six in the transition
region for two materials.
Keywords: impact testing, instrumented striker, absorbed energy, miniaturized Charpy

testing, round robin test program
Introduction
Eleven organizations, representing Japan, the United States, and Europe, are
participating in the round robin test program. Table 1 lists the participants that are
providing data for the CVN portion of the round robin and provides information on the
testing equipment used in the program. Similar information for the MCVN portion of the
program is given in Table 2. At the time of the writing of this paper, not all of the
participants had completed the round robin testing. However, since more than half of the
test results were submitted to Subcommittee E28.07.07 by the November 1998 deadline,
a decision was made to publish the interim results of the program.
Two well-characterized materials were chosen for the program. The National Institute
for Standards and Technology (NIST) has provided two heats of 4340 steel (material
similar to that used for test machine verification). Oak Ridge National Laboratory
(ORNL) has provided the other material, which is an A533B (HSST-03) reactor pressure
1 President, MPM Technologies, Inc., 2161 Sandy Drive, State College, PA 16803.
2 Research Engineer, MPM Technologies, Inc., 2161 Sandy Drive, State College, PA
16803.
3 Owner, Computational Mechanics, Inc., 1430 Steele Hollow Road, Julian, PA 16844.
223
Table 1 - Round Robin Participants and Test Machine Characteristics for CVN Tests
1 ] 3 i 4 i 7 i 8 .[ 9 i 11
Nuc.Res. ENEL i VTI? Mfg. JRC-IAM- i SCK-CEN, i MPM Tech.
Organization Institute, Research, CIEMAT, Tech., Petten, ,~ Belgium i Inc.,
Czech R. Italy, Spain Finland NL i USA
I
Machine Pendulum i Pendulum i Pendulum i....Pendulum i Pendulum i Pendulum i Pendulum

Hammer C-hammer i U-hammer i C-hammer i C-hammer '!" C-hammer i C-hammer i U-hammer
Machine Mfr. rinius OlseniTinius Olseni Wolpert Wolpert I TONI-MFL iTinius Olsen
Machine
368 i 358 i 300 i 300 i 300 i 300 i 400
..C.ap.a.~.!t.~..(JL
Impact Vel.
(m/s) 5.1 i 512 i 5'53 i 5'43 i 5.52 i 5.42 i 5.47
.....Span (mm) 40 i 40 i 40 i 40 i 40 i 40 i 4o
Striker Rad. i 8
2
(mm)
Anvil Rad.
(ram)
Strain Gage
i , i
i Semi- i
i i 1 1
i ,
Semi-
i 1
.....(.s..~,.)..Z~pe ..... i conductor i i Metal wire i Metal wire conductor Metal wire
SG Position i....L~ft.,..r~ht..., i i.T.gp,...b....o..ttomi Backside [.. L.e..ft,..risht..j. Left, r!.~h.t....
SG Distance
(cm)
i 1.5 i i 1.5 i 1 i 15 i0762
Frequency
Limit (kHz).. i lOOi 300 i 300 i 1000 i 1000 i 1000
Max Data
Points lo24 i 4ooo i 8000 ii 10000 i 8000 i g192 i 18182
Gen. Yield 2-region fit i 2-region fit i 2-region fit i visual i 2-region fit i 2-region fit i Hooke's law
Method .........:...~A)..............i (B) i (A) ! i..............(A.). [. (B).............i. intersect
Impact Vel. Light sensor l PE i PE i PE ! PE i PE
Method(C) (No Loss) i (No Loss)i (No Loss)i (No Loss)I(W&F Loss)
Notes: Partici rants 2 (Imperial College, UK), 5 (Tohoku University, Japan), 6 (NIST, USA) and
10 (ORNL, USA) had not submitted data by the time of publication of this paper. Blank entries
indicate data not reported.
(A) Region one is a line parallel and coincident with the rising part of the second peak. Region
two is a fitted curve to the right of the General Yield Point. The intersection of the two
curves defines General Yield.
(B) Region one is a linear fit to the initial load oscillations. Region two is a non-linear fit up to
the maximum load. The intersection of the two curves defines General Yield.
(C) PE (No Loss) refers to calculation of the impact velocity using the pseudo mass of the
hammer, the acceleration of gravity, and the potential energy without correction for windage
and friction. PE (W&F Loss) refers to a revision of the PE (No Loss) procedure to include
windage and friction corrections.
MANAHAN ET AL. ON ROUND ROBIN TEST PROGRAM 225
T a b l e 2 -Round Robin Participants and Test Machine Characteristics for Miniature

Charpy V-Notch Tests
......p_.arfieipant ! 1 i 3 i 4 i 7 .L 8 .] 11
Nuc.Res. i ENEL i T VTT Mfg. i JRC-IAM- i MPM Tech.
Organization Institute, i~ Research, i C1EMAT,spain[ Tech., [ Petten, i Inc.,
Czech R. i Ital)l l Finland ] NIL i USA
Machine Pendulum i..D...rol~..we..!.gh.t.iPendulum i Pendulum i Pendulum i Pendulum
Hammer C-hammer ~ i C-hammer i C-hammer i C-hammer i U-hammer
Machine Mfr. Roel Amsler{ CEAST ! Wolpert i WolDert i Woloert !Tinius Olsen
Machine i i ~ i
...CaEa.ci.ty....(j.).... 50 i 34 ~ 25 i 50 i 15(D) i 400
Impact Vel. i i i i i
(m/s) 3.83 i 3.96 i 3.85 i 3.85 i 3.85 i 5.47
iiilSp..an..(mm.)..... 19.3 i 19.3 i 19.3 i 19.3 i 19.3 i 19.3
Striker Rad. i i 2 i 2 2 4
(mm) 2 i 4 ! i i [
Anvil Rad. ! ~ i i i
(mm) 0.5 i 0.5 i 0.5 i 1 i 0.5 i 0.5
"Strain Gage i Semi- i i i
.....(S.G.)..Tzp..e.... i conductor i i Metal wire i Metal wire i Metal wire
SGPosition i Lateral i i Side faces i Backside i Leftr..r!.~h.t.."
SG Distance [ 2.4 i i 1 i 1 i
..............(cm)............. i i i i i 0"762
Frequency i 100 i i ~ i
..L!.mit...(~.z).... ! i 300 i 300 i 1000 i 1000
Max Data i i ~ i i
Points 20000 i 4000 i 8000 i 10000 i 8000 1 18182
Gen. Yield 2-region fit i 2-region fit i i Visual i 2-region fit i Hooke's law
Method (A) i .(B).............i ! i ..(..A...). i intersect
Impact Vel. PE i PE i i PE i PE ~ PE
Method(C) (NoLoss) i (No Loss) i i (No Loss)i (NoLoss)i(W&FLoss~
Notes: Partici ,ants 2 (Imperial College, UK), 5 (Tohoku University, Japan), 6 (NIST, USA) and
10 (ORNL, USA) had not submitted data by the time of publication of this paper. Participant 9
(SCK-CEN, Belgium) did not participate in MCVN testing. Blank entries indicate data not
reported.
(A) Region one is a line parallel and coincident with the rising part of the second peak. Region
two is a fitted curve to the right of the General Yield Point. The intersection of the two
curves defines General Yield.
(B) Region one is a linear fit to the initial load oscillations. Region two is a non-linear fit up to
the maximum load. The intersection of the two curves defines General Yield.
(C) PE (No Loss) refers to calculation of the impact velocity using the pseudo mass of the
hammer, the acceleration of gravity, and the potential energy, without correction for windage
and friction. PE (W&F Loss) refers to a revision of the PE (No Loss) procedure to include
windage and friction corrections.
(D) Weights added to increase test machine capacity to 50 J.
vessel plate material. This choice of materials provides one material with an inherently
low data scatter and one with a relatively large data scatter. An additional advantage of
this choice of materials is that most of the tests can be performed at room temperature to
achieve transition region behavior and upper-shelf behavior. Only one series of tests had
to be performed at elevated temperature to achieve upper-shelf behavior. However, the
elevated test temperature was specified well into the ductile fracture temperature region.
This approach has the advantage that test temperature variation effects are essentially
eliminated from the results. In addition, the absence of time constraints on specimen
placement for room temperature tests provided participants the opportunity to verify
proper specimen centering on the supports. This paper summarizes the test results
obtained by the participants for both the CVN and MCVN tests.
Test Procedure
The goal of an ASTM round robin is to have participant laboratories follow the same
test procedure so that the validity of the test procedure and the data accuracy can be
evaluated. The test procedure for the current round robin is described below. Since some
participants have 8 mm strikers and others have 2 mm strikers, the choice of an 8 mm or
2 m m striker was left as an option since other studies have shown negligible effects of
striker radius at low and intermediate energy ranges (up to - 175 J)[1].
In accordance with ASTM Test Methods for Notched Bar Impact Testing of Metallic
Materials (E 23), conventional Charpy V-notch (CVN) specimens (10 m m x 10 mm x 55
mm) have been tested. Nominally half scale miniature Charpy V-notch (MCVN)
specimens (-4.83 mm x 4.83 mm x 25 mm) have also been tested. The miniature
specimen sizes are slightly below 5 mm in cross section to allow them to be machined
from the broken halves of conventional specimens. The A S T M E 23 anvil geometry was
required for conventional specimen testing. In particular, tests were conducted using a 40
m m span with anvils having a 1 mm radius. Participants were asked to test the miniature
specimens using a scaled ASTM anvil geometry (19.3 mm span with anvils having a 0.5
mm radius) although not all of the participants were able to produce a miniature anvil
meeting the round robin requirements, as indicated in Table 2.
CVN tests were conducted in accordance with the test requirements of ASTM E 23.
Guidance concerning instrumented testing was provided in the draft ASTM standard
given in Reference [2]. Similarly, guidance for MCVN testing was provided in
Reference [3]. Participants were encouraged to complete the entire test matrix (given
below) although several participants chose to complete only portions of the matrix.
As mentioned earlier, uncertainties due to alignment and temperature variation were
minimized for tests conducted at room temperature. The upper-shelf energy tests for the
A533B material were conducted at 150~ Since this test temperature was chosen so that
the test is conducted well into the upper-shelf, the effects o f heat loss during bath transfer
were not expected to have a large effect on the data. The test matrix is summarized
below:
A533B Material Transitional Fracture Behavior Upper-shelf Behavior

CVN 6 tests at 20~ + I~ 6 tests at 150~ + I~
MCVN ..... 6 tests at 150~ + I~
4340 Material Transitional Fracture Behavior Upper-shelf Behavior

CVN 6 tests at 20~ + I~ 6 tests at 20~ + I~
MCVN ..... 6 tests at 20~ + 1~
The impact velocity was specified in the range of 3 to 6 m/sec for all tests. For
test machines capable of adjusting the impact velocity, the participants were requested to
keep the impact velocity as close to 5.5 m/sec as possible. In order to avoid issues related
to data acquisition accuracy, it was recommended that at least 1,000 data points should be
recorded for each millisecond of acquisition, or for systems with lower storage capability,
it was requested that the participants acquire as many data points as possible. The data
storage capacities are summarized in Tables 1 and 2.
Materials
The A533B material was provided by Oak Ridge National Laboratory (ORNL). This
material was chosen because it is representative of the scatter obtained in testing reactor
grade steels. The material is an ASTM A533, grade B, class 1 plate designated as plate
03 by the Heavy Section Steel Technology (HSST) program at ORNL. The material has
been well characterized by ORNL and the data and fabrication history are presented in
Reference [4]. As discussed in the reference, the through thickness Charpy, drop weight,
and fracture toughness properties do not vary significantly from the 1Athickness (T) to
the 90 T positions.
The LT and TL data in Reference [4] were fit to provide reference data for comparison
with the round robin tests. The software used to fit the Charpy data [5] in this study
provides two alternative fitting functions. The first is a hyperbolic tangent function with
four fitting parameters. The second is a polynomial of order two (three fitting
parameters). In addition to fitting the mean energy versus temperature trend, the software
also simultaneously fits the data with a three-parameter Weibull statistical distribution.
The orientation of specimens tested in the round robin is TL. However, there are more
ORNL data for the LT orientation than for the TL orientation. While the TL data points
are sufficient to obtain a reasonable mean trend behavior, the amount of data is not
sufficient to obtain a reliable statistical behavior for the material. Therefore the LT data
were used to determine the Weibull exponent and the results used to fit the TL data. The
results of the fitting are given in Table 3 and these data are compared with the results of
the round robin testing later in the paper.
Table 3 - Plate 03 Charpy Energies at the Two Round Robin Test Temperatures
Oak Ridge Plate 03 LT (RW) Data

Hyperbolic Tangent Fitting
Temperature Energy (J)
(C) 1% 5% 50% 95% 99%
20 53 60 78 94 100
150 148 153 165 176 180
Polynomial Fitting
(C) 1% 5% 50% 95% 99%
20 48 56 76 94 101
150 152 155 163 171 174
Oak Ridge Plate 0 3 TL (WR) Data

Hyperbolic Tangent Fitting
(C) 1% 5% 50% 95% 99%
20 42 47 59 70 74
150 114 117 127 135 138
Polynomial Fitting
(C) 1% 5% 50% 95% 99%
20 35 41 55 69 74
150 117 120 126 131 134
The 4340 specimens provided by NIST were machined from double vacuum melted
AISI-SAE 4340 steel bars. This material was chosen because of its inherently low data
scatter. The material preparation has been optimized by NIST for use in the Charpy test
machine verification program. The steel also meets the requirements of AMS 6414 and
has P, S, V, Ni, Ti, and Cu contents as low as can be achieved. The bars are normalized
at 350 C and hardened to approximately 35 HRC. The goal is to produce Charpy bars
having a minimum of large carbides in the structure and with the most uniform carbide
precipitation as possible. Since NIST has optimized the process for conventional CVNs
only, the MCVN specimens were machined from the CVNs provided by NIST using an
electrical discharge (EDM) wire machine.
NIST tested twenty of the low energy CVNs (designated LL68) and twenty of the high
energy CVNs (designate HH71). These tests were conducted on the NIST reference
machines. The results for the LL68 tests indicated an absorbed energy of 23.7 J with a
2or = 3.3 J. Similarly, the HH71 tests resulted in an absorbed energy of 122.2 J with a 2or
= 13.8 J. The uncertainties for this heat of material are comparable to and only slightly
higher than those of NIST verification lots.
Results
The data submitted by the participants are summarized in Tables 4 through 6 and in
Figures 1 through 7. As shown in Figure 1, the reference dial energy mean (NIST data)
and the population mean (round robin participants) are in close agreement for the 4340
material. Similarly, the instrumented striker data given in Figure 1 for the 4340 material
are in close agreement with the NIST reference data for all participants. As indicated in
Table 5, tests on the low energy 4340 material exceeded the load limit of the
instrumented test system used by participants 3 and 9. This experimental limitation
resulted in the energy, brittle fracture load, and peak load data being under predicted or
not measured at all. Overall, the measured means and 2 a levels are in good agreement
for both the dial and instrumented energy measurement for the low energy 4340 tests.
As shown in Figure 1 for the low energy A533B material tests (top figures), the
reference and population means are not in close agreement. This is believed to be due to
the fact that the sample blocks were cut from different locations within the HSST 03
plate. One particularly notable feature of the Figure 1 A533B plots is the large scatter for
the A533B energies. Examination of the data shows that the instrumented striker results
track the dial closely, that is, a low dial reading corresponds with a low instrumented
striker result. It has been concluded that the scatter is due to the statistical nature of the
trigger particle fracture mechanism inherent in reactor pressure vessel steels. It is
interesting to note that the scatter is larger for all of the C-hammer machines than that of
the U hammer machines (Participants 3 and 11). Additional work will be required in the
future to determine the cause of the low scatter for the U hammers machines as shown in
Figure 1. The overall conclusion for the A533B material tests conducted in the transition
region is that, in spite of the large scatter, the instrumented striker data is in good
agreement with the dial energies (see Table 4).
Figure 2 presents a comparison of the dial and instrumented striker energies for CVN
tests conducted on the upper-shelf. For the 4340 material, the instrumented, dial, and
reference data are all in good agreement. The reference data for the A533B material does
not agree with the population mean for the reasons discussed earlier. Overall, the dial
and instrumented striker data for upper-shelf CVN tests for all of the participants are in
very good agreement and the scatter is typical of that observed for these materials
(2~-12%).
The MCVN upper-shelf energy data are summarized in Figure 3 and in Table 6. The
instrumented striker and dial energies reported for the A533B material are in close
agreement, and the scatter is typical. However, the scatter for the 4340 MCVN
instrumented data appears to be unacceptably large. Closer examination of the load-
deflection data for participant 3 showed exceptionally large signal oscillations.
Discounting the data for participants 1 (participant 1 did not submit load-deflection
curves) and 3 would result in reasonable agreement between the average instrumented
striker data and the average recorded dial energies. The instrumented data for
Participants 4, 7, 8, and 11, which are in agreement, are on average lower than the dial
energies. It is important to note that the MCVN data reported by participants 1,3,4,7, and
8 was taken using low energy capacity test machines. Hammer vibration could cause
Table 4 - Summary of CVN Data for A533B Materials
__1 i 3 i 4 i 7 i 8 i 9 i 11
CVN Data: Transitional Fracture Behavior - A 5 3 3 B Material
Striker Rad. (mm) 2.00 i 8.00 [ 8.00 i 2.00 i 2.00 i 4.00 [ 8.00 i 8.00
.......Ira_pact Vel. (m/s) 5.10 j 5.12 [ 5.53 ] 5.43 i 5.52 [ 5.42 i 5.47 i 5.20
Dial Energy_ (J) 83.50 i 74.50 j 76.71 i 69.00 i 88.23 i 74.58 i 74.49 ~ 55.00
Dial Energy SD (J) 18.72i 5.53 i12.70 i_20:_76__i - 20.54
- i 23"49 i 11.30 i 7.0
Instr. Striker (J) 84.77 75.16 i 75.75 i 67.20 i 85.74 I 74.58 i 74.49 !
Instr. Striker SD (J) 19.06 i 5.33 i 13.45 i 22.02 i 20.52 i 23.49 i 11.30 i
Instr.-Dial Diff. (J) 1.27 i 0 . 6 6 ! - 0 . 9 6 i -1.80 i -2.49 i ,o.o [ ,o.o i
Gen. Yield (kN) 14.381 12.14! 13.93 i 14.42i 12.59 i 13.34 i 12.17 i
Max. Load (kN) 18.81 i 18.50 [ 19.29 [ 17.34i 18.20 I 19.34_] 19.19[
*PL Deflect. (mm) i 3.20 i 3.22 i 2.88 i 2.55 i._2.91 ] 3.28 1
Brittle Fract. (kN) 17.37 i 17.94 [ 18.08 [ 16.38[ 16.92 i 18.97] 18.62 L - - -
Brittle Arrest (kN) 7 . 5 6 i 4.25 i 10.61 i5.22 i8.44 i7.652 i 9.31 i
C V N Data: U p p e r - S h e l f Fracture Behavior - A 5 3 3 B Material
Striker Rad. (mm) 2.00[ 8.00 i 8 . 0 0 i 2.00 i 2.00 j 4.00 i 8.0 I~8.00
ImpactVel. (m/s) 5.10 i 5.12 i5.53 i 5.43 i 5.52 i 5.42 [._5.47 i 5.20
Dial Energy (J) 128.2 i 134.2 1131.8 [ 127.6 j 130.5 i 128.3 i 130.0! 126.0
Dial Energy SD (J) 5.64 i 5.72 i5.18 i 5.41 i 3.91 i 6.19 i 4.60 i 2.75
Instr. Striker (J) 129.8 i 132.0 i 136.1 i 128.0i 131.9 i 128.3 i 130.0 i
Instr. Striker SD (J) 6.05 T 5.19 5.38 i- ~-~- ~ 5"96 i 6"19 i--4-:60-i
Instr.-Dial Diff. (J) 1:63 -2.16 i4.29 _____i 0.40 ! 1.43 i ]'0.0 I t0.0 i. . . . .
Gen. Yield (kN) 11.63 i 10.42 I 11.38 11.84 1 12.59i 11.2 I10.26.1 . . . . . . . .
Max. Load (kN) 16.19 i+' 17.43 [17.35 16.22 16.02i 17.371 17.231
*PL Deflect. (mm) i 3.59 [3.66 i[ 3 . 4 3 3.09 i 3.29 ! 3.57 i
Notes:* PL Deflect refers to the measured deflection at peak load. -~ The procedure used
by Participants 9 and 11 employed a scaling of measured loads to match the instrumented
energy with the dial energy.
Table 5 - Summary o f CVN Data f o r 4340 Materials
Participant I 1 i 3 !~ 4 i 7 !~ 8 9 i 11 Ref.
C V N Data: Transitional Fracture Behavior - 4340 Material
Striker Rad. (ram) 2.00 i 8.00 i 8.00 i 2.00 2.00 4 . 0 0 i 8.00 8.00
ImpactVel.(m/s) 5.10 i 5.12 i 5.53 5.43 5.52 5.42 { 5.47 5:47
.........p ialEne~gy (_J).... 24.50123.54i24.20 23.54 25.36 24.70 L25.72 23.69
Di~l Ene~gy._SD (J) __1.52 ~i 1.25 ,i 1.41 0.54 1.84 1.86 I 1.85 1.56
Instr. Striker (J) 26.80--~19.8 1-24.05 ' 2-2-46 ~'2i--54 ......? 2.5.72 . . . . .
Inst-r-'_S_trike(SD (J~)i~2963_i' ~i_~0~98-iii12~4_5 ~_-_-_~ii-_-_~
In~tr~TDial pjff ~ . (_J) 2.30 ....... '-3.7 -9:15 ~..71._08=3.83 .......... t ...........~0.0 .................
Gen. Yield (kN) 28.75 I
.......M_ax:Lgad(_kN) 33.17 t 2 9 . 9 31.96i 35.60i35.05 t ............
9PLDeflect. (ram) 0.85 0.88 i 0.80 [ __0.89_~_~0.93
Brittle Fract. (kN) 3_2.62,_'f_29.6 , 29.63 [ 35.60 1 31.45 t ~36.0 ...........
Brittle Arrest (kN) 4.90 0.15 0.00 i 0.00 i 1.03
C V N Data: Upper-Shelf Fracture Behavior - 4340 Material
Striker Rad. (mm) 2.00 8.00 8.00 2.00 i 2.00 i 4.00 i 8.00 [ 8.00
Im~Pa_c_tVel. (__m/s) __5_.10_5_.12 5.5_3 ._5:_43 [ f . 5 2 i_5.42 j_5.47 ~_5r47
..... DialEnergz(J ) _.[24.0 [124.0 1!6.4 [!_14.6~.1_19.7 ]__1_!7.2~1_.!5.06 [_!22.2
pialEne_r~S__D(J) 8.39 7.56 6.24 i 6.80 i 3.62 i_2.58 i_5.26~_6_90
Instr. Striker (J) _l127.0_~_122.L+llS~8[_.U6.Li.l19.7___] 117.2] 115:_06~..........
fi_nstr. StrikerSD(J)l 8.29
Gen. Yield ( k N ) [ 22.61

I : a ad- kN>A25.28
6.44 7.12 [ 6.69 [ 4.14 [ 2.58~ 5.26 [
17.09 21.36 [ 21.04 { 20.40 I 18.87 18.503~I

24.25 ~ . 4 9 ~ 2 4 . 6 2 2 23.877 25.17 25.24 i . . . . . . .
Notes:* PL Deflect refers to the measured deflection at peak load. ~'A portion of
measured loads for each specimen of Transitional Fracture Behavior 4340 Material were
out of range for Participants 3 and 9. :~ The procedure used by Participants 9 and 11
employed a scaling of measured loads to match the instrumented energy with the dial
energy.
Table 6 - Summary of MCVN Data for A533B and 4340 Materials
Participant I 1 I 3 i 4 i 7 1 8 i 11
MCVN Data: Upper-Shelf Fracture Behavior - A533B Material
Striker Rad. (mm) 2.000 i 4.000] 2.000 i 2.000 i 2.000 i4.000
~ a c t Vel. (m/s) 3.830 ~ 3.960 t 3.850 I 3.850 i 3.850 i 5.470
Dial Energy__(_J) 10.78 i I 10.98 j 10.90 I 11.29 i 10.73
Dial Energy SD (J) 0~3-25~ ~ 0.603 ._0.292_~ 0.186 i 0.231
Instr. Striker (J) 11.22 I__11.64~ 10.84 11.00 i 10.40 i 10.73
Instr. Striker SD (J) 0.360 ~ 0.581 ~I 0 . 6 2 3 - 0 . 2 9 2 I0.185 i 0.231
Instr.-Dial Diff. (J) 0.433 i i-0.138~ 0.100 i-0.942 ~ 0.00
Gen. Yield (kN) 2.445 ~-~90-iii125.547 i 2.482 i 2.530 i, 2,580
Max. Load (kN) 3.263 ~ 3.668 I 3.358 ~3.392_ i 3.254 i 3.399
*PL Deflect. (mm) 1.307 ] 1.498 i 1.272 ,! 1.380 i 1,325
MCVN Data: Up] per-Shelf Fracture Behavior - 4340 Material
Striker Rad. (mm) 2.000 i 4.000 L2.000_~_
i i 2.000, [ 2.000 iI 4.000
......._Impact Vel. (m/s) 3.830 i 3.960 i 3.850 I 3.850~ 3.850 T 5.470
Dial Energy_ (J) 11.12 | 10.62 i 10.78 i 11.28 ] 10.84
Dial Energ.y SD (J) O.172
Instr. Striker (J) 11.75~ 11.63 ] 9..76 [ ~ 7 4 i ~ : 1 3 ~ 10.84
Instr. Striker SD (J) 0.176 0. ...~. .- ~- (+-~
) 7 -. ~. .4. 1--4- .... -------+
~ 5~'0~..... [-0.17-7-
Instr.-Dial Diff. (J) 0.633 I-0.867~-0.040 -1.147] 0.0t
Gen. Yield (kN) 3.975 3 . 4 8 2 _ i ~ 4 . 0 ~ _ i i ~ - ~ - i 4.0-0-'7-j 3.741
Max. Load (kN) 4.735 f 5.500 i 4.477 [ 4.848 ] 4.663 j 4.971
*PLDeflect.(mm) i 0.818 i 0,733 i 0.748 i 0.732 i 0.786
Notes:* PL Deflect refers to measured deflection at peak load. ? The procedure used by
participant 9 employed a scaling of measured loads to match the instrumented energy
with the dial energy.
CVN Specimens : Transitional Fracture Behavior - A533B Material CVN Specimens : Transitional Fracture Behavior - A533B Material
120
/
UJ
m
~ fO0
o 1~176 , I 1
60 . . . . 60 . . . . m ~ ~ ~
>
=- 40 f " . . . . . . ". . . . . . . . . . . . . . 1". . . . "'"" ' Z
= >
I ." _-- ." -- Population (Mean • "1-
;~ 20 | ~ " Reference (Mean • 20 1 ~ . Reference (Mean • " >
Z
0 . . . . . . . ~ Laboratory(Mean • 2(;) I ~ Laboratory (Mean •
0 ~ m
1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11
>
Participant Participant .F-
0
CVN Specimens : Transitional Fracture Behavior - 4340 Material CVN Specimens : Transitional Fracture Behavior - 4340 Material Z
35 0
C
Z
30 0
LU ~J
0
OJ
~) 20 1 Z
-H
- -- : -- Population (Mean • 2o) m
3~15. ~ f5- .... Population (Mean • 2o)
:::'---~:'::...-~ Reference (Mean •
~ Laboratory (Mean • 2G) :::::--::::'~ Reference (Mean • 20)
~ 10- "13
=
~ Laborato~(Mean~.a)
m 0
a 5- :7O
E
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
Participant Participant
r,o
Figure 1 - Comparison of Dial and Instrumented Striker Energies for CVN Tests Conducted in the Transition Region co
co
r~3
CVN Specimens : Upper Shelf Fracture Behavior - A533S Material CVN Specimens : Upper Sheff Fracture Behavior - A533B Material CO
~ 140 .......... "13

.'_-_'_-_- ~--&'.-'_-_'_-'_T:T-_E'_-_T--_'-'-_'t m
Z
ILl m
120 0
C
100 100 I--
C
~ ~0. .... Population (Mean • 2(;) 9= "~ 80 " ~ " ~ Population (Mean • 2(']
0
::'::m:::~:...~" Reference (Mean • ::::::::::: Reference (Mean • 2(;)
{ Laboratow(Mean~r.2(;) Laboratory (Mean • 2(;1
w 40. ~" 40 0
r~ "-4
20. _~ 20 ITI
03
0
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11 fi3
>
0
CVN Specimens : Upper Shelf Fracture Behavior - 4340 Material CVN Specimens : Upper Shelf Fracture Behavior - 4340 Material m
z
160 160 9 C
140 '
o
120
"13
~ 100
0
c~. 80- "= ~ 80 . . . . . Population (Mean • 2(;)
9 -- ~ ~ Population (Mean • 2(;) :73
m
:::::----~::::.--~ Reference (Mean • 2(;) ::::.-~-:::::.~ Reference (Mean • 2G)
60- 60 ~
~ Laboratow(Mean-'-~(;) { Laboratow(Mean-,'9o) m
~ 40- _s 4o.
a 20- 20 ~
0
2 3 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
Figure 2 - Comparison of Dial and Instrumented Striker Energies for CVN Tests Conducted on the Upper-shelf
MCVN Specimens : Upper Shelf Fracture Behavior - A533B Material MCVN Specimens : Upper Shelf Fracture Behavior - AS33B Material
14 14
. . . . . . . . . . . . . . . . . . . . .
12 ~ 12
. . . . . . . . . ~ ......... 3 ............ i11 . . . . . . . . . . . . . . . . . . . . . . . . .

w 10 -~ 10
8 .~A 8
,,=,~
6 ~ 6 - -- : ~. Population (Mean • 2o)
.~ ~ .~ Population (Mean • 2(
~ Laboratory (Mean r >
~ Laboratory (Mean :P.2a)
4- 4- Z
i11 E >
p -r
2
>
Z
m
0
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
>
Participant Participant F
0
MCVN Specimens : Upper Shelf Fracture Behavior - 4340 Material MCVN Specimens : Upper Shelf Fracture Behavior - 4340 Material Z
14- 14- 0
C
Z
12 . . . . . . . . . .
12!
m 10 ~o 10. 0
i-
8 84 Z
(n 6 84 m
3~ 6 .... Population (Mean • 2a) "~" ~- Population (Mean 9 2a)
~ Laboratory (Mean r ~ Laboratory (Mean :P..2a)
0= 4 "0
ul _E
0
N 2 2
>
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
PO
Figure 3 - Comparison of Dial and Instrumented Striker Energies for MCVN Tests Conducted on the Upper-shelf cnc~
t~
CVN Specimens : Transitional Fracture Behavior - A533B Material CVN Specimens : Transitional Fracture Behavior - A533B Material
18 25
16
-0
2o . . . . . . . ----~ . . . . . . . . . . . . . . . . . ~- ...... m
141 Z
. . . . . . . . . . . . . . . . . . . ..~_. -~ . . . . . . . . . .
12 C
T-
3 15
10- C
m~
_v
8-
g --" : -" Population (Mean • 2(~) 10 - - ~ ~ ~ Population (Mean • 2G)
r 6 ~ Laboratory (Mean .2o) ~ Laboratory (Mean +~o')
r3 c~
4- 5-
-q
2- m
0 "--t
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
CVN Specimens : Transitional Fracture Behavior - A533B Material CVN Specimens : Transitional Fracture Behavior - A533B Material c)
177
Z
25- 20 -'t
C
18- ..<
_-- ~ ~. Population (Mean • 20)
20" 16- ~ Laboratory (Mean .,.~o) o
14
o
. . . . . . . . . . . . . . . . . . . .
15 :73
o
2D
10- ." _-- " _-- Population (Mean • 2(~) ........ I ........................ I"n
~ Laboratory (Mean :P.2o)
5 -
2-
0 0
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
Figure 4 - Instrumented Striker Loads f o r CVN Tests Conducted on A533B Material in the Transition Region
CVN Specimens : Transitional Fracture Behavior - 4340 Material CVN Specimens : Transitional Fracture Behavior - 4340 Material
35 45
4O
30
35
25
~ 30 0 i
3
-o 20 ~, 25
~ 15 ~ 20
--_ : _-- Population (Mean • 2c .... Population (Mean • 2o)
g ~" :iS E
~ Laboratory (Mean .-k2r ~ Laboratory (Mean r ) )>
~ 10
Z
10
"I"
5 >
Z
0 m
2 3 4 5 6 7 8 9 10 2 3 4 5 6 7 8 9 10 11 -I
PaNic/pant Participant )>
CVN Specimens : Transitional Fracture Behavior - 4340 Material CVN Specimens : Transitional Fracture Behavior - 4340 Material o
Z
40, 10 33
0
40 9 C
Z
8 0
3s 0
7 0
3o !
(30
6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
~A 25. Z
r
5, .-f
r 20- m
- _--~"~ Population (Mean = 20), 4 : " : -- Population (Mean • 20)
m 15 " ~ Laboratory (Mean +9o)
13 ~ iboratory (Mean :~.2o) "13
10 2, 30
0
5 G')
~0
)>
0 (~ O O
E
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
PO
Figure 5 - Instrumented Striker Loads for CVN Tests Conducted on 4340 in the Transition Region CO
PO
CVN Specimens : Upper Shelf Fracture Behavior - A533B Material CVN Specimens : Upper Shelf Fracture Behavior - A533B Material CO
CO
16 20
/
_ ..... ~ - - ~ . . . . . . . . . . . . . . . . . ~- .....
18- "13
14
. . . . . . . . ._ . . . . . ...... - _~_.~. - I"11
16! Z
U
=. 14 C
.3
~ 10- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r-
C
(o : " : : Population (Mean • 2o)

a_ e- : - " - Population (Mean • 2o') x s ~
~E ~ Laboratory (Mean :~.o.)
~ Laboratory (Mean • 2(;) 6- 0
4 .-t
4
---t
2 2- m
--I
0
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
CVN Specimens : Upper Shelf Fracture Behavior - 4340 Material CVN Specimens : Upper Shelf Fracture Behavior - 4340 Material m
Z
30 30- C
25 25! . . . . . . . . . "r . . . . . . . . . ~" . . . . . . K" .....

. . . . . . . . . . . . . . . . . . 0
"11
-IJ
20 20
3 o 9
_.R
.-A15 if)
c = ~ 15
: ~. : "- Population (Mean • 2(;) m
- ~ Population (Mean • 2a)
10 ~ Laboratory (Mean • :~ 10- ~ Laboratory (Mean r
5- 5-
0 ,i 0
2 3 4 5 S 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
Figure 6 - Instrumented Striker Loads for CVN Tests Conducted on the Upper-shelf
MCVN Specimens : Upper Shelf Fracture Behavior - A533B Material MCVN Specimens : Upper Shelf Fracture Behavior - A533B Material
4.5
3.5 4
3.5
3
-g
_,o 2.s
~ 2.5
el i
~ ~" Population (Mean • 20) 9 " ." ~ Population (Mean • 2G
~ 1.s
o ~ Laboratory (Mean r ~; 1.5 ~ Laboratory (Mean r >
1 Z
1 >
"I-
0.5 0.5 >
Z
0 i'n
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11 .-I
Participant >
Participant r-
MCVN Specimens : Upper Shelf Fracture Behavior - A533B Material MCVN Specimens : Upper Shelf Fracture Behavior - 4340 Material 0
Z
4.5 7-
0
C
4 6- Z
3.5 ,~
, ~ Q 5
3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
~ 2.5 Z
=J< .-I
.S_~ 2 m
: -- " ~ Population (Mean :k 2o) - -- : .~ Population (Mean • 20) (n
:! --I
s; 1.5 ~ Lal~ratow (Mean:0.~) ~ Laboratow (Mean P.2o)
2 .-Q
1
o
1 U~
0.5
0 0
2 3 4 5 6 7 8 9 10 11 2 3 4 5 6 7 8 9 10 11
Figure 7 - Instrumented Striker Loads for the MCVN Tests Conducted on the Upper-shelf CO
(D
lower measured energies for the instrumented measurements and several of the
participant's data support this hypothesis (participants 4 and 8). The effects of hammer
vibration on energy measurement are discussed in Reference [6], and the data of Figure 3
are consistent with conclusions of [6].
The instrumented striker loads for CVN tests conducted in the transition region for the
A533B material are given in Figure 4 and Table 4. The results for general yield load,
peak load, and brittle fracture load were consistent for all of the participants. Participant
8 experienced large scatter for general yield load and this scatter could not be analyzed
because the instrumented signals were not submitted. As shown in Figure 4, the scatter
in arrest load is significantly larger than that for the general yield load, peak load, and
brittle fracture load. A large part of the scatter is due to the inherent uncertainty in the
fracture process itself. Examination of the load-deflection curves indicates that part of
the scatter is due to the procedure used by some participants, which is to extrapolate a
curve fitted to the post-brittle ductile tearing data and intersect this curve with the nearly
vertical unloading line of the brittle fracture event to define the crack arrest load. This
approach may not be desirable because there is no physical relationship between post-
brittle tearing and unstable crack propagation. Other participants have defined the crack
arrest load as the lowest load recorded at the end of the brittle fracture event.
The instrumented striker loads for CVN tests conducted in the transition region for the
4340 material are given in Figure 5 and Table 5. The results for peak load, brittle
fracture load, and arrest load are in good agreement. Since this material undergoes brittle
fracture shortly after peak load, the crack arrest load is nearly zero for most tests. Since
this material reaches peak load shortly after general yield, most participants were not able
to determine the general yield load because of limited data for curve fitting.
Figure 6, and Tables 4 and 5 present the instrumented striker loads for CVN tests
conducted on the upper-shelf. Similar data are presented in Figure 7 and Table 6 for
MCVN tests conducted on the upper-shelf. As shown in the figures, the agreement
among the participants is good and the uncertainties are relatively low.
The round robin testing has been performed on materials with widely differing data
scatter. The 4340 material has very low scatter from specimen to specimen while the
A533B material shows large transition region scatter which is typical of reactor pressure
vessel steels. The draft ASTM test procedureS used in this round robin have yielded
results that show good agreement among the various laboratories. Accordingly, it is
concluded that the standards should proceed to full ASTM Committee ballot.
It has been concluded also that several modifications and additions should be made to
the existing draft standards to clarify the procedures. In particular, it has been observed
that excessive vibrations in the instrumented signal may result in exceptionally large
variations in the instrumented striker loads. These large oscillations are believed to be
caused by insufficient test machine stiffness. The upper-shelf material behavior is
characterized by dynamic oscillations during initial loading which are damped out
significantly due to plasticity by the time at which peak load is reached. Therefore,
instrumented signals that exhibit large load oscillations after peak load indicate
insufficient test machine stiffness. The instrumented draft standard should be modified to
include this caution.
Two of the participants exceeded the load capacity of their instrumented test system
when testing the 4340 material. This lead to under prediction of the total absorbed
energy and incorrect estimation of the peak load, brittle fracture load, and deflection to
peak load. The instrumented standard should be modified to include a caution, which
requires verification that the instrumented striker calibration range has not been
exceeded. This clarification is particularly important in cases where test machines are not
equipped with dial gages or optical encoders for independent energy measurement. In
such cases, the load-time curve must be inspected for evidence of load signal saturation.
The draft standard defined the brittle fracture arrest load as "The force at the end
(arrest) of unstable crack propagation..." and prescribed that this force "is determined as
the force at the intersection of the steep drop of the force-displacement curve and the
smoothed curve through oscillations of the subsequent part of the force-displacement
curve." This determination was originally adopted because it provides a convenient
means for automatic determination of the arrest load. However, as mentioned earlier,
there is no physical basis for this determination. It is proposed that the draft standard be
modified to define the crack arrest load as the lowest load measured at the end of the
brittle fracture event unloading to ensure that an accurate and conservative crack arrest
load is measured.
References
[1] Nanstad, R. K. and Sokolov, M. A., "Charpy Impact Test Results on Five Materials
and NIST Verification Specimens Using Instrumented 2-mm and 8-mm Strikers,"
ASTM STP 1248, 1995, pp. 111-139
[2] Proposed ASTM Standard Method for Instrumented Charpy V-Notch and
Miniaturized Charpy V-Notch Impact Tests on Metallic Materials, Draft 6, July
1998
[3] Proposed ASTM Standard Method for Impact Testing of Miniaturized Charpy V-
Notch Specimens, Draft 6, July 1998
[4] ORNL Characterization of Heavy-Section Steel Technology Program Plates 01, 02,
and 03, NUREG/CR-4092, April 1985
[5] "FRACTURE/FIT: Weibull Based Fracture Fitting Code," MPM Technologies, Inc.,
1997
[6] Manahan, M. P., Sr., and Stonesifer, R. B., "The Difference Between Total Absorbed
Energy Measured Using An Instrumented Striker and That Obtained Using and
Optical Encoder", Pendulum Impact Testing: A Century of Progress, ASTM STP
1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American Society for Testing
and Materials, West Conshohocken, PA, 1999
Enrico Lucon 1
European Activity on Instrumented Impact Testing of Subsize Charpy V-Notch

Specimens (ESIS TC5)
Reference: Lucon, E., "European Activity on Instrumented Impact Testing of

Subsize Charpy V-Notch Specimens (ESIS TC5)," Pendulum Impact Testing: A
Century of Progress, STP 1380, T. A. Siewert and M. P. Manahan, St., Eds.,
Abstract: In the frame of the activities of the European Structural Integrity Society
(ESIS) and within the Technical sub-committee 5 on "Dynamic Testing at Intermediate
Strain Rates," a working group was formed in 1991 to address the subject of subsize
specimen impact testing. In parallel with drafting a test procedure, this group has
organized and carried out an interlaboratory round-robin exercise aimed at validating the
procedure and clarifying several aspects of the methodology. The round-robin involved
thirteen labs, two of which were from the U.S., and was subdivided into two phases. This
paper presents an overview of the activity of this ESIS working group, with specific
emphasis on the round-robin results, which will be presented and discussed.
Keywords: ESIS, subsize Charpy V-notch specimens, instrumented impact testing, test
procedure, round-robin
Nomenclature
b Length of the specimen ligament, below the notch (mm)

Fgy In an instrumented impact test, force corresponding to general yield of the
specimen (kN)
Fm In an instrumented impact test, maximum force sustained by the specimen
(kN)
FATT50 Temperature corresponding to 50% shear fracture appearance in a Charpy test
(oc)
KV Dial energy absorbed during the test, as read from the machine indicator (J)
NDTT Nil Ductility Transition Temperature, measured according to ASTM E 208-
95a (~
np02 Yield strength, in a tension test (MPa)
RUTS Ultimate tensile strength, in a tension test (MPa)
T41J Temperature corresponding to 41 J of absorbed energy in a Charpy test (~
T68J Temperature corresponding to 68 J of absorbed energy in a Charpy test (~
1 Senior Researcher, Reactor Materials Research, SCK.CEN, Boeretang 200, B-2400

Mol, Belgium.
242
LUCON ON EUROPEAN ACTIVITY 243
USE Upper Shelf Energy (asymptotic value of absorbed energy reached in fully
ductile conditions) in a Charpy test (J)
W Specimen width (mm)
Wm In an instrumented impact test, absorbed energy corresponding to the
maximum force sustained by the specimen (21)
wt In an instrumented impact test, total absorbed energy, corresponding to the
termination of the test (J)
The use of subsize Charpy V-notch specimens for instrumented impact testing has
been gaining more and more widespread popularity in the scientific community,
following the increasing need to estimate in a reliable way the mechanical properties of
service-exposed or irradiated plant components. Indeed, this has to be achieved without
sampling large quantities of material from the component, if this has to be maintained in
operation. Furthermore, in the case of tests on irradiated material, the available space in
irradiation facilities is usually rather limited.
Instrumented testing of subsize impact specimens represents a convenient tool for
characterizing impact and fracture properties (such as FATTs0, T41J,USE, etc.) using
limited amounts of material or, alternatively, machining small specimens out of
previously tested broken specimens. An unambiguous test procedure is therefore needed
to carry out tests in an efficient and consistent way, as well as reliable correlations with
full-size specimen data, in order to derive significant assessments of a material's fracture
properties in the absence of a "conventional" approach to material characterization.
The ESIS Technical sub-committee on "Dynamic Testing at Intermediate Strain

Rates"
In the frame of the Technical Committee 5, Fracture Dynamics, of the European

Structural Integrity Society (ESIS), the sub-committee named "Dynamic Testing at
Intermediate Strain Rates", chaired by H. McGillivray (Imperial College, London) and
formed by approximately 20 active members, has been working for several years on the
drafting of test procedures concerning various types of dynamic mechanical tests, such as:
impact tests on [1] V-notched and [2] fatigue pre-cracked Charpy-V specimens, [3]
dynamic tensile and [4] dynamic compression tests.
The final objective of the work is the submission of such documents, following
validation through internal round-robin exercises, to international standardizing bodies
(such as ISO) for their eventual transformation into official test standards. From this point
of view, the main achievement so far has been the adoption by ISO of the test procedure
on Instrumented Impact Testing of Charpy V-notch Specimens of Metallic Materials,
produced by the sub-committee in 1996 in its final version [1], which is soon to become
the ISO 14556 standard "Steel - Charpy V-notch Impact Test - Instrumented Test
Method".
The Working Group "Instrumented Impact Testing of Subsize Charpy V-Notch

Specimens"
In 1991, enough interest on the topic of subsize specimen testing was found
amongst members of the sub-committee, and a working group was therefore set up for the
preparation of a suitable test procedure. The author of this paper was entrusted with
coordination of the group.
The document produced, which was initially based to a large extent on the
analogous test procedure for full-size test pieces [1], has now reached the stage of Draft 7
[5]. The most important features of this version of this procedure are the following:
9 its structure reflects, as closely as possible, that of the ISO 14556 Draft Standard;
9 no correlation procedure with respect to standard-size specimens is recommended,
although the user is clearly warned that subsize specimens test results cannot be
directly used as an alternative to full-size data;
9 definitions of characteriztic values of force, displacement and impact energy are
exactly the same as for standard type specimens;
9 the reference test geometry is the Kleinstprobe (small specimen) mentioned by the
German standard DIN 50 115 "Prfifung Metallischer Werkstoffe -
Kerbschlagbiegeversuch" (April 1991), see (Figure 1); other specimen geometries (for
instance, the half-size, 1:2 scale, Charpy-V specimen) are shown in an Annex;
9 for pendulum-type machines with available energy not greater than 50 J, the deviation
between dial energy indication and absorbed energy given by the instrumentation
should not exceed • J;
9 the recommended upper frequency limit of the force measuring system has been set at
250 kHz;
9 particular emphasis is given to temperature control, due to the small size of the
samples; in view of this, in situ (on location) conditioning techniques are
recommended;
6 0 ~ __ 2 ~
V 1 __.0.1
/ 4___0.1
R=0.1 _+0.025
27___0.6 3_+0.1
Figure 1 - Geometry of the reference subsize Charpy V-notch specimen (dimensions in

mm)
the force at general yield (Fgy) is determined from the intersection of the straight line
fitted through the oscillations of the initial part of the force-displacement trace
(Hooke's line) and the fitted curve through the oscillations which follow the plastic
collapse of the ligament; a recommended procedure for determining Hooke's line [6] is
given in an Annex.
Outline of the Round-Robin Exercise
In 1994, a decision was taken to validate the procedure by means of an

interlaboratory exercise, with the aims of clarifying some ambiguous aspects of the
methodology and gaining some experience on this type of mechanical test.
The round-robin program has involved 13 laboratories (11 from 7 different
European countries and 2 from the U.S.); the participation of the American laboratories
represented the liaison of the activity of this working group with the corresponding
ASTM E28.07.08 sub-committee working on the same topic.
The material was an ASTM A 533 B Class 1 steel (AEA correlation monitoring
material - code designation JRQ); its mechanical properties are given in Table 1.
Table 1 - Mechanical properties of JRQ steel
Rp02 RUTS Elong. Red,area T41J T68J FATTs0 NDTT

(MPa) (MPa) (%) (%) (~ (~ (~ (~
467-487 624-635 25-27 76-77 -23 - -28 -13 - -21 +2 - +4 -15
The experimental activity was subdivided into two parts:

9 Phase 1 consisted of 3 to 5 tests per lab, conducted at room temperature using 3 m m x
4 mm x 27 mm specimens (Figure 1), using an impact speed of approximately 3 m/s;
9 Phase 2 was intended to allow all participants to investigate different aspects of the
experimental procedure and to study the influence of various parameters (temperature,
specimen and striker geometry, impact speed, side-grooving, etc.) on test results.
Different types of impact machines have been used to perform the tests, including
full-scale pendulums (i.e. with available energy _>300 J), small-scale pendulums, a small
drop-weight tower and a high-velocity servohydraulic machine.
Results of Phase I
Typical results of Phase 1 are shown in Figures 2 and 3, which present

characteriztic force and impact energy values reported by the participants.
F i g u r e 2 - Characteriztic values of force reported by Phase 1 participants (dashed lines:

average values; dotted lines = • standard deviation)
Figure 3 -Characteriztic values of impact energy reported by Phase 1 participants

(dashed lines: average values; dotted lines = 4- standard deviation)
A detailed analysis of Phase 1 results has been given in a final report [7]; this
includes the determination of the repeatability and reproducibility of the test method, in
accordance with the ISO Standard 5725:94 "Accuracy (trueness and precision) of
measurement methods and results".
The main remarks, which have emerged from the discussion of Phase 1 results
within the working group, can be summarized as follows:
1. Although different impact machines and test procedures have been employed by the
participants, no problems have emerged in applying the draft test procedure.
2. The scatter in the characteriztic values reported is reasonably low (in the range 4% to
8%) in the case of force and energy values but much higher for displacement values
(10% to 22%).
3. The influence of impact speed and upper frequency of the measuring system on test
data appears quite moderate, except for characteriztic values relevant to test
termination (displacement st, absorbed energy Wt).
4. Mean values of total calculated energy (Wt) are consistently lower than dial energy
values, although the difference always remains within +0.5 J.
Results of Phase 2
Nearly all participants chose to investigate the influence of temperature on test

results; additionally, a few labs concentrated on other parameters as well, such as impact
speed, specimen and striker geometry, span value and specimen side-grooving.
Influence of Temperature - Figures 4, 5 and 6 show the characteriztic values of
force at yield point (Fgy), maximum force (Fm) and total calculated energy (Wt) reported
by'participants as a function of temperature. Apart from a few anomalous data, reasonable
scatter was obtained, taking into account the different values of impact speed (ranging
from 2.6 to 3.9 m/s) and the problems connected with temperature control, which is
indeed a very critical aspect in the case of very small specimens.
Influence of Specimen Geometry - Although all labs used DIN 50 115 type
specimens, several tests were also performed using a different specimen geometry (half-
size, with thickness = 5 mm, width = 5 mm and length = 27.5 mm); this enabled some
attempts at normalizing energy values to be performed. The most successful was based on
fracture volume normalization (W-b2), as shown in Figure 7, where full-size specimen
data, available from the literature, are also reported. As expected, a shift in transition
temperature is evident, in that smaller-size test pieces tend to behave in a more ductile
manner; this was also confirmed by shear fracture measurements. Furthermore, DIN 50
115 specimens show a large scatter as compared to full-size and half-size specimens, due
to their reduced cross section which tends to emphasize the influence of local material
inhomogeneities.
Effect of Side-Grooving - The effect of side-grooving was investigated by one of the
labs on DIN 50 115 type specimens, tested at 3 m/s at different temperatures. The
comparison, in terms of shear fracture, with plain-sided specimen data Figure 8 clearly
shows that side-grooving significantly increases constraint conditions at the notch root,
shifting transition curves towards higher temperatures.
2.3
2.1
1.9
A
Z
'-'r 1.7 []
o
t,I. o
1.5
6 oo~ []
~ 1.3
I11
P 1.1
0
0.9
0.7
0.5 i i i i i i I i
-150 -100 -50 0 50 100 150 200 250 300

Temperature (~
Figure 4 - Values of force at general yield reported by Phase 2 participants
2.6
2
E
I,I.
e
&
,s
E
E
1.4
0.8 i i i i ~ i i i i i i i
-150 -100 -50 0 50 100 150 200 250 300

Temperature (~
Figure 5 - Values of maximum force reported by Phase 2participants

12
+
+
A
"3
10 + + *+
+
+ § *§ + ,
>~
0J
e-
"O
s
O
,,Q
§ .;
+++ + + +
O
I-
+ + +
~++§247
0
- 150 -100 -50 0 50 100 150 200 250 300
Temperature (~
Figure 6 - Values of impact energy reported by Phase 2 participants
Figure 7 -CompariSon between impact energy values from different specimen sizes,
normalized by fracture volume (VK.b2)
100
75 [ / 20% side-grooved
50
/~176
e-
25 /:/
+ r 1
0
-10, -50 0 50 100 150 200 250 300
Temperature (~
Figure 8 - Comparison between plain-sided and 20% side-grooved specimens
Calculated Energy Values VS Dial Readings The very satisfactory comparison -
between measured (KV) and calculated (Wt) energy data, obtained in Phase 1, was
thoroughly confirmed by Phase 2 results: with the exception of a few anomalous data at
high energy values, the reported W t lie within a +0.5 J tolerance band with respect to
machine dial readings (Figure 9).
10
. ,.-'/~tp,. -+
~" 8
~ 6
9z"~ ........
o 4
o
I,-
2
0 "I "''" , i r i , I , i , I i
0 2 4 6 8 10
Machine dial, KV (J)
Figure 9 - Comparison between total absorbed energy Wt and dial readings KV

A more detailed analysis of the results of Phase 2 is presently underway, along with
discussions at the occasion of the bi-annual meetings of the technical sub-committee;
preliminary elaborations are available in the form of a draft report [8].
Acknowledgements
The author gratefully acknowledges the priceless collaboration of all the working
group members, who substantially contributed to the development of the test procedure
and to the successful results obtained from the round-robin exercise.
References
[1] "Proposed Standard Method for the Instrumented Charpy-V Impact Test on Metallic
Materials," prepared by the Working Party European Standards on Instrumented
Charpy Testing of the ESIS TC5 Technical sub-committee on Dynamic Testing at
Intermediate Strain Rates, Draft 10, April 1996.
[2] "Proposed Standard Methods for Instrumented Pre-Cracked Charpy Impact Testing
of Steels ," prepared by the Working Party European Standards on Instrumented
Pre-Cracked Charpy Testing of the ESIS TC5 Technical sub-committee on Dynamic
Testing at Intermediate Strain Rates, Draft 9, March 1998.
[3] "Proposed Standard Method for Dynamic Tensile Tests," prepared by the Working
Party European Standards on Dynamic Tensile Testing of the ESIS TC5 Technical
sub-committee on Dynamic Testing at Intermediate Strain Rates, Draft 4, April 1997.
[4] "Proposed Standard Method for Dynamic Compression Testing of Metallic Materials
at Room Temperature," prepared by the Working Party European Standards on
Dynamic Compression Testing of the ESIS TC5 Technical sub-committee on
Dynamic Testing at Intermediate Strain Rates, Draft 2, November 1996.
[5] "Proposed Standard Method for Instrumented Impact Testing of Subsize Charpy V-
Notch Specimens of Steels," prepared by the Working Party European Standards on
Instrumented Charpy V-Notch Testing of Subsize Specimens of the ESIS TC5
Technical sub-committee on Dynamic Testing at Intermediate Strain Rates, Draft 7,
June 1997.
[6] Taylor, H., "Analysis of Procedures for the Determination of the Yield Force (Fgy)
for Instrumented Subsized Charpy-V Specimens," Small Specimen Test Techniques',
ASTMSTP 1329, W. R. Corwin, S. T. Rosinski, and E. van Walle, Eds., American
Society for Testing and Materials, West Conshohocken, PA, 1998, pp. 123-136.
[7] Lucon, E., "Round-Robin on instrumented Impact Testing of Subsize Charpy-V

Specimens: Results of Phase 1," Final Report, 2 April 1998.
[8] Lucon, E., "Round-Robin on Instrumented Impact Testing of Subsize Charpy-V

Specimens: Results of Phase 2," Draft Report, 2 April 1998.
Kikuo Kishimoto, 1 Hirotsugu Inoue, 2 and Toshikazu Shibuya 1
Dynamic Force Calibration for Measuring Impact Fracture Toughness us-

ing the Charpy Testing Machine
Reference: Kishimoto, K., Inoue, H., and Shibuya, T., "Dynamic Force Calibration
for Measuring Impact Fracture Toughness using the Charpy Testing Machine,"
Pendulum Impact Testing: A Century of Progress, STP 1380, T. A. Siewert and M. P.
Manahan, Sr., Eds., American Society for Testing and Materials, West Conshohocken,
PA, 2000.
A b s t r a c t : The Charpy impact test is one of the most popular techniques for assess-
ing mechanical properties of materials under impact loading. Although many studies
have evaluated the impact fracture toughness of materials by using the instrumented
Charpy testing machine, there still remains difficulty in obtaining the impact force
accurately. In this paper, a technique is developed to remove the effect of mechanical
vibration of the instrumented hammer on the output of the sensor. Details for practi-
cal application of this technique are also presented. An application of this technique
to measure the impact fracture toughness of PMMA is demonstrated. Factors which
affect the evaluation of the impact fracture toughness are discussed.
K e y w o r d s : instrumented Charpy test, impact fracture toughness, impact force, dy-

namic stress intensity factor, deconvolution, polymethyl methacrylate
Introduction
Pendulum impact tests such as the Charpy and Izod tests are standardized in
many countries. Although these testing methods were established about a hundred
years ago, their practical usefulness as simple methods for assessing mechanical prop-
erties of materials under impact loading has scarcely deteriorated. A principal reason
for this may be that the pendulum impact test can be conducted rather easily and it
enables useful relative characterization of materials. Such relative characterization is
often sufficient to show that a newly developed material exhibits higher performance
than existing materials.
However, with continuous demands for materials which exhibit higher perfor-
1Professors, Mechanical and Intelligent Systems Engineering, Tokyo Institute of Tech-
nology, 2-12-10-okayama, Meguro, Tokyo 152-8552, Japan.
2Associate Professor, Mechanical and Intelligent Systems Engineering, Tokyo Insti-
tute of Technology, 2-12-1 O-okayama, Meguro, Tokyo 152-8552, Japan.
253
mance under impact loading, critical or absolute characterization of materials has

grown more and more important. Much effort has been made by many researchers to
establish a reliable method for evaluating the impact fracture toughness of materials.
Hence it is now widely recognized that the impact fracture toughness is an adequate
parameter for absolute characterization of materials under impact loading. Use of a
Charpy testing machine is one of the most convenient methods because the testing
apparatus is officially standardized and readily available in many cases.
A study on the use of an instrumented Charpy testing machine and a pre-
cracked specimen for evaluating the impact fracture toughness was first reported
about thirty years ago [1, 2] and studies have been conducted ever since. The key
points for accurate evaluation of the impact fracture toughness are considered as
follows:
1. Instrumentation for the measurement of impact force and deflection,
2. Evaluation of the dynamic Stress Intensity Factor (SIF), and
3. Detection of the crack growth initiation.
For the second issue, a formula for the three-point bending specimen under quasi-
static loading (e.g. ASTM Test Method for Plane-Strain Fracture Toughness of Metal-
lic Materials, E 399) is frequently applied as an approximation. However, it has been
pointed out that dynamic analysis which takes inertia effects into account is essential
for evaluating the dynamic SIF correctly [3]. Kishimoto et al. [4] derived a simple
formula satisfying this requirement. Concerning the third issue, the crack growth ini-
tiation can be detected either by a rapid decrease in strain measured by a strain gage
attached near the crack tip or by disconnection of conductive paint strip applied just
ahead of the crack tip. Thus fairly satisfactory techniques are available with regard
to the second and third issues.
On the other hand, there still remains a problem relative to the instrumenta-
tion of the testing machine. Quantities to be measured are usually time variations of
impact force and load-point deflection of the specimen. Since the load-point deflec-
tion can be derived fairly accurately from the impact force if the pendulum hammer is
sufficiently rigid, the impact force is the most important quantity to be measured. In
most cases the impact force is measured either by using a special tup with a built-in
load cell or by simply attaching strain gages to the tup. However, the time history
of a signal obtained by a sensor is generally different from that of the true impact
force because stress waves travel from the striking edge to the sensor in a complex
manner. Nevertheless, the impact force is conventionally obtained by simply multi-
plying a conversion factor to the signal acquired by the sensor, which is essentially
valid only in quasi-static loading cases and is no more than an approximation. Many
discussions have been made on this issue and several techniques for reducing errors
due to this approximation have been proposed (e.g. Refs. [5 12]). Although some of
them are effective in specific cases, no technique which is not based on the quasi-static
approximation has been proposed yet. Therefore, a definitive technique for measuring
the impact force should be developed.
In this paper, a new technique for measuring the impact force in the instru-
mented Charpy test is developed. The basic principle and practical details of this
technique will be explained first. An application of this technique to the measure-
KISHIMOTO ET AL. ON DYNAMIC FORCE CALIBRATION 255
ment of impact fracture toughness of a polymeric material will be presented and

discussed.
M e a s u r e m e n t of I m p a c t Force
It can be assumed that the hammer of the Charpy testing machine deforms
only elastically during the impact test. Hence impact response such as strain, dis-
placement or acceleration at a certain point of the hammer is linearly dependent
on the impact force applied to the tup. Namely, the impact force and the response
can be considered as an input and an output of a linear system, respectively. The
input-output relationship of this linear system is expressed as
e(t) = /: h(t - ~-)f(~-)d~- (1)
where e(t) and f(t) denote the impact response and the impact force, respectively,
and h(t) is the impulse response function of this system. It is assumed that e(t) =
f(t) = h(t) = 0 for t < O. Taking the Fourier transforms of both sides of Eq. (1), the
convolution is transformed into a multiplication as follows:
E(w) = H(w)F(w) (2)
where the symbols in uppercase denote the Fourier transforms of the corresponding
ones in lowercase. If the transfer function H(w) is known in advance, the impact force
can be estimated by
1. Measuring the impact response in the test,
2. Evaluating the Fourier transform of the measured response,
3. Finding F(w) from Eq. (2), and
4. Evaluating its inverse Fourier transform.
It is well known that use of the Fast Fourier Transform (FFT) improves the com-
putational efficiency of this deconvolution process significantly. Care must be taken
against errors caused by discretization and truncation of the Fourier integral when ap-
plying the FFT. Inoue et al. [13] showed that windowing with an exponential function
is effective for reducing the truncation error.
As the transfer function is originally unknown, it should be identified by some
means. Since theoretical or numerical analysis of the hammer behavior under impact
loading is rather difficult, an experimental analysis or dynamic calibration is an ap-
propriate choice. If an impact force measurable by some means is applied to the tup,
the transfer function can be identified by
1. Measuring both the impact force and the response simultaneously,
2. Evaluating the Fourier transforms of them, and
3. Finding H(w) from Eq. (2).
Note that the transfer function does not depend on the impact force in principle. To
apply a measurable impact force to the tup, an impact hammer (typically used for
experimental modal analysis) or a slender rod instrumented with strain gages can be
employed. Inoue et al. [13, 15] showed that a better estimate of the impact force is
attained if the transfer function is identified from many pairs of force and response
fit)
g
S
Figure 1 - - Configuration of pre-cracked Charpy specimen.
data obtained by conducting calibration many times and by using the equation
E;(w)Ek(w)
= k (3)
where Fk(W) and Ek(W) denote the Fourier transforms of the impact force and the
response measured in the kth calibration, respectively, and the superscript 9 denotes
the complex conjugate. Note that all the impact forces as well as corresponding
responses in multiple calibrations are nominally the same but slightly different due
to experimental errors such as electric noise.
E v a l u a t i o n of D y n a m i c SIF
No exact formula is available for evaluating the dynamic SIF of the pre-cracked
three-point bending specimen. Although it is possible to evaluate dynamic SIFs for
every specimens by conducting FEM analyses, it is a time consuming task in practice.
Kishimoto et al. [3] derived a simple formula for evaluating the dynamic SIF of the
pre-cracked Charpy specimen (Figure 1) given by
~)IKs ft
K,(t) = - ~ ]o f(T)sin[wl(t - v)]dT- (4)
where Wl is the natural angular frequency of the first mode of vibration of the specimen
and Ks denote the static SIF for the same specimen given by [16]
Ks - 6Sf(t) v / ~ ( a / W ) (5)
4BW ~
where for S / W = 4
ql(a/W) = 1.090 - 1.735(a/W) + 8.20(a/W) 2 - 14.18(a/W) a + 14.57(a/W) 4 (6)
Equation (4) indicates that the dynamic SIF can be evaluated only by measuring the
impact force if dimensions, Young's modulus, and mass density of the specimen are
specified.
Figure 2 - - The tup of the Charpy testing machine used. A gage protection cover is
removed.
Instrumentation of Charpy Testing Machine
Testing Apparatus
The testing apparatus used was a Charpy testing machine (Tokyo Testing
Machine Mfg., CI-30) of 294.2-J (30-kgf.m) capacity in accordance with Japanese
Industrial Standard (JIS) Charpy Impact Testing Machines (B 7722). The tup of
this machine has a groove on each side for attaching a strain gage and also has slits
in order to enhance sensitivity of the gages to the applied impact force (Figure 2).
A semiconductor strain gage (Kyowa, KSP-2-120-E4) was attached to each of the
grooves in order to measure the strain response of the tup. These two gages were
connected in series to a strain amplifier (DC-300 kHz, - 3 dB). The output signals
were recorded by a conventional digital recorder with 12-bits resolution at a rate of 1
#s per sample.
Dynamic Calibration
Dynamic calibration for identifying the transfer function was conducted by

impacting a horizontally supported rod with the hammer (Figure 3). The rod used
was 10 mm in diameter (equal to the thickness of the standard specimen, that is B
in Figure 1), 700 mm in length and is made of SUS 304 stainless steel. The impact
end of the rod was supported by a jig made of Polymethyl Methacrylate (PMMA)
(Figure 4). This jig consists of three parts and is designed to satisfy the following
requirements:
1. The impact end of the rod is located exactly at the same position as the impact
face of the specimen is set in the test.
2. The center part of the jig has a groove so that it does not touch the tup when
the tup come into contact with the rod.
3. The center part of the jig is sufficiently softer and lighter than the rod so that
it does not affect the impact force induced between the tup and the rod.
Figure 3 The method of dynamic calibration of the hammer.
Figure 4 The jig for supporting the impact end of the rod in calibration. Dimensions
are in mm. The thickness of this jig is 10 ram, that is, equal to the width of the
standard specimen (W in Figure 1).
KISHIMOTO ET AL. ON DYNAMIC FORCE.CALIBRATION 259
Figure 5 - - Calibration to identify the transfer function between the impact force
acting at the end of the rod and the strain measured by the gages on the rod.
4. Both the rod and the center part of the jig are ejected after impact without
obstructing the hammer motion.
5. The hammer can pass through the gap between the two side parts of the jig
after impact.
On the other hand, the free end of the rod was supported by a V-block. Thus
calibration can be performed in the same manner as the testing of the specimen.
In calibration the impact force was measured using strain gages (Kyowa, KSP-
2-120-E4) attached at 100 mm from the impact end of the rod. The transfer function
between the impact force acting at the end of the rod and the strain measured by
the gages on the rod was identified by conducting another calibration, that is by
impacting the rod longitudinally with another rod (10-mm in diameter, 1-m long
and made of SUS 304) as shown in Figure 5 [17]. According to the one-dimensional
theory of longitudinal impact of rods, the impact force acting between the rods can
6 150
' (a)'
' (b)'
z
4 ~ 100
2 r 50
O
Q
1.1.
0 o
, i i
-2 -50 i I i
0 I 0 1 2
T i m e [ms] Time [msl
Figure 6 - - A typical result of dynamic calibration of the hammer when the release
angle was 30~ (a) Impact force, and (b) Strain response of the hammer.
be measured by strain gages attached at 100 m m from the impact end of 1-m rod
(Gage 2 in Figure 5). The strain response of 700-mm rod can also be measured
simultaneously. Therefore, the transfer function of the rod used for the calibration of
the hammer can be identified in the same manner described above. Once the transfer
function of the rod is identified, the impact force acting on the end of the rod can
be estimated from the measured strain in the same manner described above. As an
alternative to this technique, one may apply the method developed by Lundberg and
Henchoz [18] to measure the impact force acting on the end of the rod.
A typical result of calibration is shown in Figure 6 when the hammer was
released from an angle of 30 ~ The impact force is almost a rectangular pulse whose
duration is equal to the time for longitudinal waves in the rod to travel from the
impact end to the free end and return to the impact end. On the other hand, the
strain response of the hammer is a rectangular pulse with a sinusoidal vibration
superposed on it. This sinusoidal vibration corresponds to the transient vibration of
the hammer excited by the impact force. It is obvious t h a t multiplying a conversion
factor to this strain response does not give a correct impact force. In this study,
calibrations were conducted ten times under a fixed condition and ten pairs of force
and response d a t a were acquired to identify the transfer function according to Eq. (3).
If it is found by conducting calibration that the impact force and the output
of a sensor are similar, one may simply multiply a conversion factor to the measured
output in order to obtain the impact force. Therefore, dynamic calibration mentioned
here is worth conducting to check whether the output of the sensor is similar to the
true impact force or not, that is, whether the quasi-static approximation can apply
or not.
Measurement of Impact Fracture Toughness
Specimen
The material tested was PMMA. Young's modulus of this material was 3.06
G P a and the mass density 1.19 k g / m 3. The geometry of the specimen was as shown
in Figure 1. The dimensions were B -- W -- 10 mm and L -- 55 mm, which was in

accordance with the standard Charpy specimen specified in JIS Test Pieces for Impact
Test for Metallic Materials (Z 2202). This is because the testing machine employed
in this study was one for metallic materials (JIS Z 7722). Longer specimens (L = 80
mm and 125 mm) were also tested in order to examine the effect of overhang on the
evaluation of the impact fracture toughness. The span between the supports (S in
Figure 1) was 40 mm in accordance with JIS Z 7722.
Specimens were cut from a 10-mm thick sheet. After finishing the outer ge-
ometry, a pre-crack was introduced at the center of each specimen in the following
manner:
1. Machining a rectangular notch (0.1 mm in width and 4 mm in depth) using a
circular cutting blade,
2. Inserting a fresh razor blade (0.1 mm in thickness) into the notch, and
3. Striking the back of the razor blade with a hammer to initiate a natural pre-
crack. J
This method is in accordance with the tapping method specified in ASTM Test Meth-
ods for Plane-Strain Fracture Toughness and Strain Energy Release Rate of Plastic
Materials (D 5045). The razor blade was tapped with a modified pendulum hammer
of a Charpy testing machine for plastic materials (ASTM Test Methods for Impact
Resistance of Plastics and Electrical Insulating Materials, D 256). The modified pen-
dulum hammer was released from a fixed angle in order to avoid a scatter of the
pre-crack length. The pre-crack length was measured by using a microscope after the
impact test.
Impact Force
A typical result of impact test for 80-mm-long specimen is shown in Figure 7

when the hammer was released from an angle of 30~ Figure 7(a) shows measured
strain response of the hammer. A sinusoidal vibration seen in the calibration data
(Figure 6(b)) can be also found in this data. The impact force estimated from this
data and the transfer function identified above is shown in Figure 7(b). The estimated
impact force has two large peaks followed by some fluctuation. The two large peaks
indicate that a double impact occurred between the tup and the specimen. This is
due to mechanical interaction between the hammer and the specimen. The specimen
was accelerated by the first impact and gained a velocity faster~than the hammer.
After a short period, the specimen was overtaken by the hammer again and was
fractured by the second impact. On the other hand, the fluctuation after the peaks is
due to mathematical difficulties in the inverse analysis to estimate the impact force
[15, 19] and does not represent the correct variation of the impact force. However,
this fluctuation does not affect the impact fracture toughness since the dynamic SIF
at the crack growth initiation is derived from the variation of the impact force before
that time.
In this case, the time variation of the impact force does not differ so much from
that of the strain response of the hammer. However, a significant difference may be
found in other cases. Some examples are found in Refs. [14, 20]. In principle, the
10 !
(a)
i 400 i
(b)
~ 5
i 200
r
-5 I I I
-200 i I i
0 1 2 0 1 2
Time [ms] Time [msl
3
2
' (c)' 2000
~ i
(d)
E 1 =_ o o o
(d
0. 0 t-
--1 o
U-
~ -2
I I = I I i
-3 -1000
0 1 2 0 1 2
Time [ms] Time [ms]
Figure 7 - - A typical result of impact test for 80-mm-long specimen when the release
angle was 30~: (a) Strain response of the hammer, (b) Estimated impact force, (c)
Dynamic SIF, and (d) Strain near the crack tip.
difference becomes more significant, that is, the dynamics of the pendulum hammer
becomes more important as higher energy is required to break the specimen.
It should be emphasized that the true impact force acting between the striking
edge of the tup and the specimen can be oscillatory because of mechanical interaction
between them including inertial force of the specimen. The contribution of the inertial
force of the specimen to the impact force record should not be removed because the
dynamic SIF derived from Eq. (4) takes this inertial effect into account correctly.
Techniques for removing this inertial effect are required only when a quasi-static
formula is used for evaluating the dynamic SIF approximately.
Dynamic SIF
The dynamic SIF calculated from the impact force according to Eq. (4) is
shown in Figure 7. Since the crack length was assumed to be constant and the
overlapping of the crack surfaces was allowed in this calculation, the dynamic SIF
vibrates with the natural frequency of the specimen after reaching its maximum
value. However, this vibration does not affect the evaluation of the impact fracture
toughness because the crack growth initiates no later than the dynamic SIF reaches
its maximum value.
4 i 4 i
(a) (b)
3 3
E E
It. 2 0 g. 2
o
Oo 6o o
0 0 0 0 0 ~ 0000
"o
1
v o
I I I I I I I I
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.3 0.4 0.5 0.6 0.7 0.8
a/W a/W
4 i
o o
(c)
3
E o
o
13_ 2
"o
1
I i I I
0
O:3 0.4 0.5 0.6 0.7 0.8
a/W
Figure 8 Summary of evaluation of the impact fracture toughness of PMMA: (a)

L =50 mm, (b) L =80 mm, and (c) L =125 mm.
Impact Fracture Toughness
The strain measured by a strain gage at the vicinity of the crack tip is shown
in Figure 7(d). In the beginning part the variation of this strain is similar to that
of the dynamic SIF, which indicates that the calculated dynamic SIF represents the
stress field near the crack tip correctly. The crack growth initiation can be detected
by a rapid decrease in the strain. Therefore, the impact fracture toughness of this
specimen can be evaluated as the value of the dynamic SIF at the time corresponding
to the rapid decrease in the strain. Note that the crack growth initiates certainly
before the estimated impact force starts to fluctuate and no later than the calculated
dynamic SIF reaches its maximum value.
A summary of evaluation of the impact fracture toughness Kid is shown in
Figure 8. All results were obtained when the release angle of the hammer was 30~
that is when the impact velocity was 1.40 m/s. For the 55-mm-long specimen, the
impact fracture toughness is evaluated almost constantly regardless of the pre-crack
length of the specimen. The average for all specimens was 1.20 MPa.m 1/2. The quasi-
static fracture toughness of the same material was also evaluated and an average value
Kic = 1.23 MPa.mU2 was obtained. Since the impact velocity was not very high in
this experiment (/~ ~ 104 MPa.ml/2/s), the similarity between the impact and quasi-
static fracture toughnesses indicates the validity of evaluation of the impact fracture
toughness.
For the 80-mm-long specimen, the scatter of results appears slightly more
than for the 55-mm-long specimen. However, the average for all specimens was 1.25
MPa.m 1/2 which is very close to the average obtained for the 55-mm-long specimen.
Therefore, the impact fracture toughness was evaluated fairly correctly for this spec-
imen too.
On the other hand, a significant scatter is observed for the 125-mm-long speci-
men. A principal reason for this is that Eq. (4) does not consider the specimen length
L. As the specimen length L increases, namely as the overhang at each end of the
specimen increases, the evaluation of the dynamic SIF becomes less accurate. For
the 125-mm-long specimen, the overhang is 42.5 mm which is longer than the span
S = 40 mm.
From the results of this experiment, it may be said that the impact fracture
toughness can be evaluated accurately by the present technique if the length of the
specimen is less than twice the span between the supports. In order to verify this,
however, a more detailed examination will be necessary.
For the three-point bending specimen such as the pre-cracked Charpy spec-
imen, it is known that loss of contact at the supports affects the evaluation of the
dynamic SIF significantly [3]. Unfortunately, no observation was made at the sup-
ports in this study. Nevertheless, it may be considered that the loss of contact did not
occur in this study since the testing machine employed has a capacity much greater
than the energy required to break the specimen.
In this study, it is assumed that the anvils are rigid, that is, the supports of
the specimen are fixed. This assumption may be justified because the deformation
of the specimen is usually much larger than the deformation of the anvils. Since
there exist dynamic effects in the anvils as in the hammer, however, it is interesting
to instrument the anvils to measure the reaction forces. This will also be useful for
checking the loss of contact at the supports.
Conclusions
A new technique was developed for measuring the impact force in the instru-
mented Charpy impact test. This technique makes it possible to estimate the impact
force acting between the striking edge of the tup and the specimen from the measured
response of the hammer, that is, to remove the effect of the mechanical vibration of
the hammer on the output of the sensor. Details for practical application of this tech-
nique were also described. An application of this technique to measure the impact
fracture toughness of PMMA was demonstrated. Factors which affect the evaluation
of the impact fracture toughness were discussed.
References
[1] Radon, J. C. and Turner, C. E., "Fracture Toughness Measurements by In-

strumented Impact Test," Engineering Fracture Mechanics, Vol. 1, 1969, pp.
411-428.
[2] Turner, C. E., "Measurement of Fracture Toughness by Instrumented Impact

Test," Impact Testing of Metals, ASTM STP 466, American Society for Testing
and Materials, 1970, pp. 93-114.
[3] Kalthoff, J. F., "On the Measurement of Dynamic Fracture Toughnesses--A

Review of Recent Work," International Journal of Fracture, Vol. 27, 1985, pp.
277-298.
[4] Kishimoto, K., Aoki, S. and Sakata, M., "Simple Formula for Dynamic Stress
Intensity Factor of Pre-Cracked Charpy Specimen," Engineering Fracture Me-
chanics, Vol. 13, 1980, pp. 501 508.
[5] Venzi, S., Priest, A. H. and May, M. J., "Influence of Inertial Load in Instru-
mented Impact Tests," Impact Testing of Metals, ASTM STP 466, American
Society for Testing and Materials, 1970, pp. 165-180.
[6] Ireland, D. R., "Procedures and Problems Associated with Reliable Control of
the Instrumented Impact Test," Instrumented Impact Testing, ASTM STP 563,
American Society for Testing and Materials, 1974, pp. 3 29.
[7] Saxton, H. J., Ireland, D. R. and Server, W. L., "Analysis and Control of Iner-
tial Effects During Instrumented Impact Testing," Instrumented Impact Testing,
ASTM STP 563, American Society for Testing and Materials, 1974, pp. 50-73.
[8] Cheresh, M. C. and McMichael, S., "Instrumented Impact Test Data Interpreta-
tion," Instrumented Impact Testing of Plastics and Composite Materials, ASTM
STP 936, American Society for Testing and Materials, 1987, pp. 9-23.
[9] Cain, P. J., "Digital Filtering of Impact Data," Instrumented Impact Testing
of Plastics and Composite Materials, ASTM STP 936, American Society for
Testing and Materials, 1987, pp, 81-102.
[10] Hodgkinson, J. M. and Williams, J. G., "Analysis of Force and Energy Measure-
ments in Impact Testing," Instrumented Impact Testing of Plastics and Com-
posite Materials, ASTM STP 936, American Society for Testing and Materials,
1987, pp. 337-350.
[11] KarisAllen, K. J. and Matthews, J. R., "Load Damping Absorbers and the
Determination of Load/Displacement Data for Precracked Charpy Specimens,"
Pendulum Impact Machine: Procedures and Specimens for Verification, ASTM
[12] Mackin, T. J. and Tognarelli, D. F., "Design and Evaluation of a Verification

System for Force Measurement Using Instrumented Impact Testing Machines,"
Pendulum Impact Machine: Procedures and Specimens for Verification, ASTM
[13] Inoue, H., Shibuya, T., Koizumi, T. and Fukuchi, J., "Measurement of Impact
Force Applied to a Plate by the Deconvolution Method," Transactions of the
Japanese Society for Non-Destructive Inspection, Vol. 2, 1989, pp. 74-83.
[14] Inoue, H., Ishida, H., Kishimoto, K. and Shibuya, T., "Measurement of Impact
Load by Using an Inverse Analysis Technique (Comparison of Methods for Es-
timating the Transfer Function and its Application to the Instrumented Charpy
Impact Test)," JSME International Journal, Vol. 34(I), 1991, pp. 453-458.
[15] Inoue, H., Kishimoto, K., Shibuya, T. and Koizumi, T., "Estimation of Impact
Load by Inverse Analysis (Optimal Transfer Function for Inverse Analysis),"
JSME International Journal, Vol. 35(I), 1992, pp. 420-427.
[16] Tada, H., Paris, P. C. and Irwin, G. R., "The Stress Analysis of Cracks Hand-
book," Del Research, Hellertown, PA, 1973.
[17] Inoue, H., Ikeda, N., Kishimoto, K., Shibuya, T. and Koizumi, T., "Inverse
Analysis of the Magnitude and Direction of Impact Force," JSME International
Journal, Vol. 38(A), 1995, pp. 84-91.
[18] Lundberg, B. and Henchoz, A., "Analysis of Elastic Waves from Two-Point
Strain Measurement," Experimental Mechanics, Vol. 17, 1977, pp. 213-218.
[19] Inoue, H., Kishimoto, K., Shibuya, T. and Harada, K., "Regularization of Nu-
merical Inversion of the Laplace Transform for the Inverse Analysis of Impact
Force," JSME International Journal, Vol. 41(A), 1998, pp. 473-480.
[20] Inoue, H., Shibuya, T., Koizumi, T. and Kishimoto, K., "Measurement of Impact
Load in Instrumented Impact Testing," (in Japanese), Journal of the Japanese
Society for Non-Destructive Inspection, Vol. 39, 1990, pp. 390-395.
T. Varga I and E Loibnegger 2
Low Striking Velocity Testing of Precracked Charpy-type Specimens
Reference: Varga, T., Loibnegger, E, "Low Striking Velocity Testing of Precracked

Charpy-type Specimens," Pendulum Impact Testing: A Century of Progress, STP 1380,
T. A. Siewert and M. P. Manahan, Sr., Eds., American Society for Testing and Materials,
Abstract: The usual loading rate for impact testing is near 5 m/s. As it will be
demonstrated, force-deflection diagrams show strong oscillations, at least at the beginning.
Investigations on realistic loading rates showed, however, in most practical cases effective
loading rates below 0.1 m/s. Because this is true in nearly every case, a loading rate of
0.1 m/s seems to be sufficient for precracked specimens.
If the loading rate is reduced l~om about 5 m/s to 0.1 m/s, the force-deflection-diagram
can be evaluated without parasitic oscillations. Therefore measurement of fracture toughness,
Kid, Ja and the calculations of CTOD become easier.
Regarding fracture work and specific fracture work, the steel investigated shows no
significant difference if 5 m/s and 0.1 m/s impact velocity is compared. Fracture toughness
characteristics will become comparable when transition temperature shifts below 20~ are
negligible.
If a transition temperature shift above 15~ is of importance, the experimental results
have to be treated differently. A substitution will become not practicable without a function
converting the 0.1 m/s results to 5 m/s results and vice versa.
If the steels tested are sufficiently brittle, an extrapolation to the fracture toughness of
larger sections will become possible.
Keywords: precracked Charpy-type specimen, instrumented impact test, influence of

impact velocity, oscillations, transition behaviour, fracture toughness
1 Professor Dr., Head of Laboratory, Institute for Testing and Research in Materials
Technology-Vienna University of Technology (TVFA TU Wien), 1040 Vienna,
Karlsplatz 13, Austria.
2 Dr., senior researcher, TVFA TU WIEN, 1040 Vienna, Karlsplatz 13, Austria.
267
Nomenclature
A~ = dial energy
Aa = integrated energy to the estimated initiation or the onset of cleavage fracture.
Am = integrated energy to the maximum force of the force-deflection diagram.
A~t = total integrated energy of the force-deflection diagram
a0 = total crack size
aXE= dial energy divided by the area of the remaining ligament.
CTOD (&Z)= Crack-tip opening displacement at crack initiation or at the onset of cleavage
fracture. Evaluated in accordance with the ASTM E 1290 for study tests
JQ = J-Integral evaluated from A~
o'YS= yield strength
ors = tensile strength
A5 = percentage elongation after fracture
Introduction
Impact testing is performed in general using an impact velocity of 4 to 6 m/s. Under

these conditions oscillations disturb the force-deflection diagram of the instrumented test.
Furthermore, crack initiation is difficult to detect. If the impact velocity is reduced to 0.1 m/s,
the disturbing oscillations of the force-deflection diagram will be suppressed. The evaluation
of the test becomes much easier. Therefore a comparison of the testing, using on one hand
5 rn/s impact velocity and on the other hand 0.1 m/s as the loading rate, is demonstrated.
Material Tested
A modem fine grain, quenched-and-tempered carbon steel of the designation QStE

690 TM (plate thickness 15mm) was used, see Table 1 for the chemical composition and
Table 2 for mechanical properties.
Table 1 - Chemical composition of the steel QST 690TM (theplates used)
c I si [ P slAt Cr
mass content [%]
0.080 0.29 1.75 0.011 0.002 0.041 0.037
Ni MO Cu I v Ti Nb Nta
mass content [%]
0.520 [ 0.32 0.33 Io.o611 0.024 I 0.04 0.0054
VARGA AND LOIBNEGGER ON LOW STRIKING VELOCITY TESTING 269
Table 2 - Mechanical properties at the e t d of the plate

a) Tension test results:
Sample position ms ors A5

(N/mm~)
transversal 787 850 16.6
longitudinal 768 854 20.0
b) Impact values (Charpy-V): specimens not aged, transversal, at the end of the plate
Testing temperature: -40~ Av = 63, 65, 42 Joules
c) Bend tests: transversal 180~ bending angle
longitudinal 180~ bending angle
d) Apparent grain size approximately 15 Ixm.
Specimens and Preparation
If, instead of a notch radius of 0.25 mm (Charpy-V-notch) a precracked Charpy type

specimen according to Figure 1 is used [1] (crack plane orientation L-T), the plastic
deformation will be constricted to a much smaller area. The deformation rate, however, will
be increased very strongly. Because of the sharp precrack a fracture mechanics evaluation of
the Charpy-type test as a three point bend test will become possible.
zT,s'-.0~
I
ss:o~
q
. 9 .
7/)
Precracked Charpy-Ope specimen, when the milled notch o f O.05 mm radius is 1.0 mm deep
and followed by a fatigue precrack of 2 mm to 2.5 mm in addition. That meam" an initial
crack length m the range o f 3.0 m m < ao < 3.5 ram.
Fig. 1
There was a milled notch radius of 0.05 mm, which was very favourable for the initi-
ation of fatigue precracks. Fatigue precracks were intended to become 2.0 to 2.5 mm deep;
i.e. the depth of the notch and of the fatigue crack was together 3.0 to 3.5 mm. The maxi-
mum stress intensity factor range eLK in fatigue was at the beginning about 800 Nmm "s/2,
-3/2
which was reduced in very small steps down to about 300 Nmm . The precracking was
conducted on a Mierotron machine of 20 kN capacity for 20 to 25 minutes, the frequency
was around 205 cycles per second.
Experimental Arrangement
The tests were performed at the TVFA TU Vienna on an instrumented impact testing
machine, model Schnadt, according to ASK AN 425 Rev. 1 [2]. Tests with different stroke
velocities are possible on this machine of 100 J [3]. The force measurement was made by
strain gages on the striker 5m/s and on the so called chisel 0,1 m/s (cut-off frequency
100kHz). Deflection measurement was made with two different inductive displacement
measurement systems. The same machine has already been used for tests to investigate crack
initiation [4]. The pendulum used was compared with others in a round-robin test of DVM
about fatigue precracked, erosion notched and Charpy-V notched specimen. Our machine
has been shown as equivalent to a "reference machine" [5].
Instrumented Impact Tests
In the Figures 2 to 15 typical force-deflection diagrams, further macrographs of

fracture surfaces and magnified SEM pictures of the neighbouring areas of the fatigue crack
front are shown. The force-deflection diagram of a precracked Charpy-type specimen tested
at -30~ is presented in Figure 2. The loading rate was 5 m/s and rather large oscillations of
the force can be observed.
Force-deflection diagram of the precracked Charpy-Ope specimen no. 27, tested at -30~
with 5m/s loading rate. Thefracture surface can be seen in the macrograph.
Fig. 2
Figure 3 exhibits the SEM picture, the magnification being smaller in the upper picture
a) and larger in the lower picture b), With the higher magnification blunting becomes visible,
then the transition with some dimple fractured grains. Further cleavage fracture follows.
Charpy-gype specimen no. 27, magnified SEM pictures of the jattgue crack~'ont and neigh-
bouring areas: a) lower, b) higher magnification (see inc~cation)
Fig. 3:
Figure 4 shows, also at -30~ a different picture because the loading rate was reduced
to 0.1 m/s. Due to the reduction of the loading rate very little oscillation was found, at least in
the rise and the first drop of the force. The first drop also indicates the crack arrest which
could be evaluated here. Concerning Figure 5, both upper picture a) and picture b) represent
the size of the crack initiation in the centre of the impact specimen. The amount of ductile
dimple deformation seems to be somewhat higher in the specimen with the slower bend.
The crack opening displacement is apparently higher with 0.1 m/s than that with 5 m/s
loading rate. This can also be seen in the comparison of Figures 3 and 5.
Force-deflection diagram of the ,specimen no. 261, tested at -30~ with 0.1 mA' loading
rate. Thefracture surface can be seen m the macrograph on the right side.
Fig. 4
Charpy-Ope specimen no. 261, magnified SEM pictures of the fatigue crack front and
neighbouring areas with a lower magnification.
Fig. 5a
Charpy-(ype ~pecimen no. 261, magnified SFM pictures of the fatigue crack front and
neighbouring areas with a higher magnification (see indication).
Fig. 5b
The consequences of a further increase in testing temperature to about -15~ is seen in
Figure 6. The loading rate was again 5 m/s, rather large oscillations were observed The
fracture surfaces in Figure 7 show larger ductile areas than before.
Force-deflection diagram of the precracked Charpy-type specimen no- 215, tested at -15~
with 5m/s loading rate. The fracture surface can be seen in the macrograph on the right
side.
Fig. 6
Figures 7a and b show the fatigue crack front and neighbouring areas can be seen in
magnified SEM pictures.
Charpy-Ope ~pecimen no. 215, magnified SF~Vlpictures of the fatigue crack front and
neighbouring areas: a) lower, b) higher magnification (see indication).
Fig. 7
Cleavage cracks perpendicular to the crack front can be observed in the Figures 6 and
8. The large cleavage crack in Figure 8 resulted in a much higher ductility and much higher
force. Therefore the triaxiality was reduced by the cleavage crack.
This is also valid concerning resilience, The fracture surfaces again in Figure 9 show
larger ductile areas than before.
Force-deflection diagram of the precracked Charpy-Ope specimen no. 227, tested at -15~
with 5m/s loading rate. The fracture surface can be seen m the macrograph on the right
side.
Fig. 8
Figures 9a and b show the fatigue crack front and neighbouring areas can be seen in
Charpy-Ope specimen no. 227, magnified SEAt pictures of the fatigue crack front and
neighbouring areas," a) lower, b) higher magnification (see indication).
Fig. 9
The loading rate of 0.1 rn/s shows again in Figure 10 an increase in maximum force.
Cleavage was observed again. The cleavage crack has been arrested at a higher force than
half of the maximum. There was again very little oscillation during the rise and the first drop
due to cleavage observed.
Force-deflection diagram of the specimen no. 263, tested at -IO~ with 0.1 m/s loading
rate. Thefracture surface can be seen in the macrograph on the right side.
Fig. 10
Figure 11 shows a large increase of crack tip opening displacement both in the low and
the higher magnification pictures. The ductile, dimple zone has been increased. The cleavage
fracture follows immediately after the dimples.
Charpy-type specimen no. 263, magnified SEM pictures of the fatigue crack front and
neighbouring areas: a) lower, b) higher magnification (see indication).
Figll
At last at +20~ cleavage can be observed perpendicular to the crack front. This was
the case at both loading rates: 5 m/s and 0.1 m/s. At +20~ the expected increase of the
force-defleXion area was observed; few large initial oscillations and fiu'ther damped
oscillations and the continous decrease of the force-deflection curve in Figure 12 can be seen.
The micrograph as shown in Figure 13 is again typical: increase of crack tip opening
displacement can be seen followed by dimple cracking in both cases, Dimple formation is
followed by small deformation fracture,
Force-deflection diagram of the precracked Charpy. (ype ~oecimen no. 243, tested at +20 ~
with 5 m/s loading rate. The fracture ~trface can be seen in the macrograph on the right
side.
Fig. 12
Figures 13 show the fatigue crack front and neighhouring areas can be seen in
Charpy-(ype specimen no. 243, magnified SFAI pictures' of the fatigue crack front and
neighbouring areas: a) lower, b) higher magnification (see indicatiotO.
Fig. 13
At the loading rate of 0.1 m/s a rather smooth force deflection diagram can be
observed in Figure 14. If one looks at the fracture surface one will find additional cleavage
faeettes and dimple deformation structure followed by more or less ductile fracture in Figure
15. The cleavage fracture is apparently more extensive with 5 m/s than with 0.1 m/s.
Force-deflection diagram of the ~pecimen nO. 273, tested at +20~ with O.1 m/s loading
rate. Thefracture surface can be seen in the macrograph on the right side.
Fig. 14
Figure 15a+b show the fatigue crack front and neighbouring areas can be seen in
Charpy-type specimen no. 273, magnified SEM pictures of the fatigTte crack front and
neighbouring areas: a) lower, b) higher magnification (see indicatiotO.
Fig. 15
Description of Charpy-type Specimen Results
In Figure 16 the impact energy is, up to -40~ equal with both loading rates in a
scatter band. Above -40~ the curve for 5 rn/s begins to separate from that of 0.1 m/s loading
rate.
100
Av[J] , o.1 m/s
80
60
Y~ = -0"0014x4+ 0'0727x3"09496x2 + 48432x - 2'3712 X tt / ~ /
40
20
-- ~ ~" X .... due to seperate a r e a s of the fracture ligament
-80 -60 -40 -20 0 20 40

Temperature [ ~ ]
A v over temperature
Fig. 16
At -15~ a higher impact energy as usual has been observed on one of the three
specimens. The mean value of these 3 points is plotted also. Normally the cleavage crack
occurs in the fatigue crack front propagation direction, see the diagram and fracture surface
in Figure 6, specimen no. 215. If one compares this diagram with the diagram and fracture
surface of the specimen no. 227 where the cleavage crack occurs normal to the fatigue crack
front propagation direction, see Figure 8, the apparent decrease in triaxiality resulted in a
higher plastic deformation and therefore this specimen, marked with X, absorbed more
energy in comparison to the test of specimen no. 215 (Figure 6). Nearly all impact energy
values at 5 m/s are below of those at 0.1 m/s readings. The average difference, however, is
smaller than 12~ There are no significant differences observed in lower shelf values.
In Figure 17 the specific impact energy over temperature does not differ much f o m the
overall impact energy over temperature. Fracture toughness values like JQ o r CTOD (fit)
exhibit, however, a considerable difference beginning at about-75~ in the Figures 18
and 19.
282 PENDULUMIMPACTTESTING:A CENTURYOF PROGRESS
1,6
aKE[J/mm 2] 9 0.1m / s . ~
1,4
1,2
1,0
. //
0,8
0,6
0,4
0,2
A_=.=._..~ .=~-~ X ... d u e t o s e p e r a t e areas of the fracture hgament
0,0 I l I I I I I I I I I I I I I I I I I I I F I I I
-80 -60 -40 -20 0 20 40
Temperature [~
am over temperature
Fig. 17
200
JQ [kJIm2]
180 A 0.1 m/s
160 r 5.0 m/s A ~ "~
140
120 Yo.1= -0.0001xS + 0.0085x4 - 0.1688x 3 + 1.5923x~ - 4.8244x + 12.327 &
Ys.o= -0.002x4 + 0.1143xa - 1.9601xZ + 12.547x - 8.1012 J o

100 D = I
A A
60
40 = o == D
A Q DE]
2O
0 I I 1 I I i I I I J I I I I 1 I I I I I I I I I I
-90 -80 -80 -40 -20 0 20 40
Temperature[~
JQ over temperature
Fig. 18
0,16
8 o [ mm ]
0,14
A 0.1m/s = A / " /
J
0,12
: ,om, j f I
0,10 . . . .
Yo.1 = -1E-06x4 + 5E-05x3 - 0.0005x2 + 0.0029x + 0.0027 ~ ~

Ys~ -3E-06x4 + 0"0002x3 - 0"0023x2 + 0"0113x - 0"0093 9 ~ ~ =
0,08
0,06
0,04
A la m mi []
0,02
0,00 1 I 1 I I 1 I I I I I I I I I 1 I I I I I I I
-80 -60 -40 -20 0 20 40
Temperature [ ~ !
CTOD (6Q) over temperature

Fig. 19
Beginning at about -75~ the JQ values at 0.1 m/s seem to rise with increasing
temperature, see Figure 18. There is a rather small scatter to be observed at 5 m/s loading
rate, then with the loading rate of 0.1 m/s, however, the apparent start of transition is near to
-30~ Upper shelf seems to begin at approximately +40~ The value with 0.1 m/s was near
to 170 kJ/m2 whereas upper shelf lies for 5 m/s at about 100 kJ/m2.
Also CTOD(SQ), as derived by using the equation according to ASTM E 1290 Test
method for Crack-Tip-Opening Displacement (CTOD) Fracture Toughness Measurement in
function of temperature show similar features, see Figure 19. The difference in lower shelf
values begins at about -75~ The transition begins at about -40~ upper shelf is achieved at
about +40~ for 0.1 m/s and about +20~ for 5m/s. Scatter at 5 m/s loading rate is becoming
larger above -20~
The upper shelf value is about 0.14 mm with 0.1 m/s loading rate, whereas the upper
shelf value will be near to 0.09 mm for the 5 m/s rate.
The different transition temperatures of impact energy, specific energy JQ and CTOD
agree well up to about - 30~ At higher temperature the absorbed energy and the specific
energy become larger because of the plastic deformation work which is not taken into
account for JQ and CTOD.
The influence of the loading rate is depicted in the following Figures 20, 21 and 22. In
the first diagram the specific work of fracturing is shown. This steel exhibits very small
deviations in function of loading rate, the lines connecting the results at 0.1 m/s with those at
5 m/s are more or less vertical. That means that the loading rate does not reflect any large
influence in the temperature shift of the specific work. Therefore the change of the loading
does not reflect any large influence on this material.
10,0
f
~e
,~
1,0
0,1
a r e = 0.26
/
0.42 0.63 1.1 1.32 J / m m 2
0,0 I I I I I I I I I I I I I I I I I I I I I I I I
-80 -60 -40 -20 0 20 40
Temperature [~ ]
a~Bfor O.1 m/s and 5 m/s

Fig. 20
10,0
y
1,0
~e
._=
0,1
JQ = 25 40 60 80 90 kJ/m 2
0,0 I I I I I I I I [ [ I I I I I I I I I I I I
-80 -60 -40 -20 0 20 4O
Temperature [ ~ ]
J~for 0.1 m/s and 5 m/s
Fig. 21
10,0
1,0
E_
0,1
~Q = 0.02 m m 0.04 0.05 0.06 0.09 m m
0,0 I I I I I I I I I I I I I I I L I I I I I I I I I
-80 -60 -40 -20 0 20 40
Temperature [ ~ ]
CTOD (~Q)for 0.1 rn/s and 5 rn/s
Fig. 22
The J integral at initiation, depicted as JQ shows at all loading rates similar inclinations,
therefore the shift from 0.1 m/s to 5 m/s is constantly about +20~ The largest shift results
by comparing crack tip opening displacement values at least at the lowest value of 0.02 mm.
Here the shift is about +20~ whereas the higher crack tip displacements are restricted to
about +15~
It has to be mentioned that the crack initiation values were investigated according to
the thesis work cited in [4].
Acknowledgment:
The authors thank Mr. Wolfgang Engelke for conducting the experiments.
References
[]] Varga, T., "On Instrumented Testing of Charpy-V- and Precracked Charpy-
type Specimens of Weld Metal and HAZ," Jubiliiumskonferenz, ISIM,
Timisoara, Rum~inien, 1995.
[2] Varga, T., Njo, D.H. and Prantl, G., "ASK Procedure for Instrumented
Precracked Charpy-Type Tests," Proceedings C.S.N.I. No. 67, Specialist
Meeting on Instrumented Precracked Charpy Testing, Palo Alto, USA 1981.
[.3] Varga, T., "Loading Rate Influence on Precracked Charpy-type Tests," ESIS-
TCS-Sub-Committee, Miskolc, Hungaria, 1995.
[4] Loibnegger, E, " Z u r Rifleinleitung und Ihrer Bedeutung in Stahl," doc. thesis,
Vienna University of Technology, 1990.
[5] Btihme, W., "Instrumentierte Kerbschlagbiegeversuche, Ergebnisse eines
Ringtests zur Kerbform," Materialprafung 37 (1995), Nr 10, page 401-404.
M. P. Manahan, Sr. 1
In-situ Heating and Cooling of Charpy Test Specimens 2
Reference: Manahan, M. E, St., "In-situ Heating and Cooling of Charpy Test

Specimens," Pendulum Impact Testing: A Century of Progress, STP 1380, T. A.
Siewert and M. P. Manahan, Sr., Eds., American Society for Testing and Materials,
Abstract: This paper presents an innovative approach to CVN and MCVN testing: the
specimens are heated and cooled on the test machine itself. This approach is not only
cost-effective but is technically superior to methods requiring transfer of the test
specimen to the test machine from a thermal conditioning bath because the specimen is
very accurately centered and is thermally conditioned up to the moment of impact. The
system developed is capable of thermally conditioning both CVN and MCVN specimens
over the temperature range of-180~ ~ T _<315~ Advanced systems are under
development which will extend this temperature range. This paper presents data obtained
using the in-situ heating and cooling system and compares the results with specimens
which were heated and cooled using a liquid bath transfer approach. Specimens
instrumented with embedded thermocouples were used to characterize the heat loss
during bath transfer and to compare with the uniform temperature field produced by the
in-situ system. Measurements are also presented which show a factor-of-two
improvement in specimen alignment can be easily achieved with the in-situ system
centering tool.
Keywords: impact testing, Charpy test, instrumented striker, absorbed energy,

miniaturized Charpy testing, specimen thermal conditioning
Introduction
Charpy V-notch (CVN) tests to characterize the transition region and upper shelf
energy require heating and cooling of the test specimen. The conventional approach is to
use an apparatus separate from the test machine such as an oil bath to heat and an
alcohol/dry ice bath to cool the test specimens. ASTM E 23 requires that the specimens
be held in an agitated liquid bath within + 1~ for at least 5 minutes before transferring
the specimen to the test machine. More recently, some laboratories which test
radioactive specimens have developed gas heating and cooling chambers which circulate
gas around the test specimens to thermally condition them prior to transfer to the test
t President, MPM Technologies, Inc., 2161 Sandy Drive, State College, PA 16803.
2 The information contained in this paper is covered by US Patent Number 5,770,791 and
is the property of Dr. M. P. Manahan, Sr..
286
MANAHAN ON IN-SITU HEATING AND COOLING 287
machine. ASTM Test Methods for Notched Bar Impact Testing of Metallic Materials (E
23) requires that the gas circulate around test specimens within _+ I~ for at least 10
minutes prior to transferring them to the impact test machine. Irrespective of the means
used to thermally condition the test specimen, E 23 recommends the use of self-centering
tongs to place the specimen on the supports against the anvils and requires that the
specimen be broken within 5 seconds after removal from the thermal conditioning
medium. Allowing for an approximate 1 second hammer swing to contact the specimen,
the specimen must be transferred to the test machine within 4 seconds, and this is
achievable by individuals with moderate hand-eye coordination for CVNs using transfer
tongs. However, when testing miniaturized Charpy V-notch (MCVN) specimens, the
heat losses are unacceptably large over the time interval which can be achieved manually.
An example of the average temperature change measured for CVN and MCVN
specimens after removal from a liquid bath is given in Table 1. The data illustrate that
for CVN specimens a 3.5~ temperature increase has occurred at the specimen surface 5
seconds after removal from the -40~ bath. As expected, the internal specimen
temperature change is negligible. However, the surface temperature change is not
desirable because it can affect the crack initiation at the root of the notch.
As shown in Table 1, both the surface and bulk material temperature changes are
unacceptably large for the MCVN specimens. To overcome this, some laboratories have
built robotic transfer systems which can transfer the specimen to the test machine within
1-2 seconds. However, these systems are often unreliable and are very expensive. An
alternative to robotic transfer is to thermally soak the specimen on the specimen support
up to the moment of impact, and this approach is the subject of this paper. Although the
initial impetus for the development of this new technology was MCVN applications, the
improvement in temperature control and specimen alignment make the technology very
attractive for CVN testing as well.
Experimental Configuration
The experimental configuration is shown in Figure 1. Test specimens are thermally

preconditioned in-situ by flowing a thermally conditioned gas over the surfaces of the
specimen which contain the volume of material near the notch which influences fracture
properties (fracture process volume). Depending on the temperature range of interest, it
is desirable in some instances to thermally precondition the test fixture as well.
The key element of the approach is to ensure that the fracture process volume is
thermally conditioned. Figure 2 shows the extent of the plastic zone in a Charpy
specimen as a function of applied load. These data were obtained by performing three
dimensional finite element analyses [1]. Normalizing the applied load by the general
yield load makes the results applicable to other yield stress levels. It can be seen in the
figure that the plastic zone extent peaks at about one specimen height. Therefore, the
fracture process zone extends along the axis of the specimen 10 mm on either side of the
crack plane. This volume of material must be kept at the desired test temperature up to
the time when the striker contacts the specimen.
Table 1-Example of the Average Temperature Change in Miniature and

Conventional Charpy Specimens after Removalfrom a Liquid Bath
Average
Average Temperature
Time After Bath Temperature Change at
Specimen Removal from Temperature Change at Specimen
Type Bath (~ Specimen Crack Plane
(seconds) Surface Near Notch
(~ .... ............. ( ~
CVN -4O 3.5 0.2
MCVN 5 -40 12,1 3.8
Figure 1-Schematic Representation of ln-situ Heating and Cooling System for Charpy
Impact Test
In-situ Heating/Cooling Test Results
Measurements using specimens instrumented with thermocouples (see Figure 3) have

demonstrated that the desired test temperature can be reached within 10 minutes for most
test cases. In the most limiting case studied, a test specimen instrumented with
thermocouples started cooling at room temperature and reached a set point of-190~
(liquid nitrogen temperature) within 15 minutes. This is comparable to the time required
for a liquid bath to reach its setpoint. The desired test temperature (within + I~ was
maintained within the specimen up to the moment of impact. The current ASTM E 23
impact test standard requires the thermal conditioning bath to be held within + 1~ for at
least 5 minutes before the specimen is transferred to the test machine, and the specimen
must be struck within 5 seconds of removal from the bath. Thus, the current in-situ
system meets the ASTM requirement because the specimen is thermally conditioned to
within + 1~ right up to the time of impact. As mentioned earlier, the conventional bath
transfer approach results in changes in the specimen temperature after the specimen is
removed from the bath (an example is shown in Table 1). The magnitude of the
specimen temperature change after removal from a liquid bath is dependent on the bath
temperature. The larger the temperature difference relative to room temperature, the
larger the specimen temperature change.
The data obtained for the case where the specimen was cooled to -190~ is given in
Figure 3. The B1 thermocouple is an internal thermocouple which was placed at
approximately the edge of the plastic zone which is generated during loading. The B2
thermocouple is also an internal thermocouple which was placed at the crack plane just
below the notch. As shown in Figure 3, the surface thermocouples were placed in
analogous positions. The most important thermocouple readings are the internal
thermocouples B1 and B2. These thermocouples read within 0.3~ of each other. As
shown in the figure, the surface therrnocouples are also in very close agreement with the
internal thermocouples. The excellent agreement between the surface temperature and
the internal specimen temperature is achieved by bathing the specimen in a thermally
controlled flow of liquid nitrogen vapor.
In order to completely validate the in-situ heating and cooling technology, full size
CVN specimens were tested using the in-situ system and the results were compared with
test results obtained using the conventional liquid bath approach. The results of these
tests are shown in Figure 4. The material tested is a modified A302B reactor pressure
vessel steel referred to as plate G-8-3. This plate is a beltline plate in the Nine Mile Point
Unit 1 nuclear plant. The specimens were cut from an archive prolongation at a depth of
one quarter of the plate thickness (88T location). The orientation of the specimens tested
is transverse-longitudinal (TL).
Full Charpy transition curves were developed by Battelle and by MPM Technologies,
Inc. (MPM) using the conventional liquid bath approach. The Battelle and MPM data
were fit using the FRACTURE/FIT program [2]. The software used to fit the Charpy
data provides two alternative fitting functions: the first is a hyperbolic tangent function
with four fitting parameters; and the second is a polynomial of order two (three fitting
parameters). The hyperbolic tangent function was used for this study. In addition to
fitting the mean energy versus temperature trend, the software can also simultaneously fit
the data with a three parameter Weibull statistical distribution. This Weibull distribution
1.0
/
/ On/*~SO WITHOUT
WITH20%SI GROVES
DE
SIDEGROVES
,// o~ *J ,/A302BATROOMTEMPERATURE
.
3DFINITES,ELEMENT
ED,S.R MODEL
ACES
Z0
LOAD~GENERALYIELDLOAD
EXTENTOF NOTCHREGIONPLASTICITY
Figure 2-Plastic Zone Size Results from 3D Finite Element Simulations of Charpy
Specimens (+s are for midplane and Os are for free surfaces)
Figure 3-Plot of Specimen Temperature as a Function of ])me Obtained Using the In-situ
System. The Plot Shows that the Specimen Attains the Setpoint Temperature of
-190 ~ within 15 Minutes.
Figure 4-Comparison of CVN Data Obtained Using Conventional Liquid Bath and In-situ
Heating and Cooling System for Plate G-8-3 in the As-Received Condition
is temperature dependent and the 50% probability level of the distribution is, by
definition, the mean trend fit. A Weibull parameter of 4 was used which defines a
Gaussian distribution of the measured energies at each test temperature. Since few data
points were measured near the lower shelf, the mean lower shelf energy was set at 4 J.
The results of the fitting are given in Figure 4. The fit was performed using the liquid
bath data only and the in-situ data were plotted for comparison with the liquid bath data.
As shown in the figure, the in-situ data are within the scatter of the liquid bath data and
within the confidence intervals calculated by FRACTURE/FIT. It has been concluded
that the in-situ technology is validated and can be used routinely for CVN and MCVN
testing.
SpecimenAlignment
The in-situ heating and cooling technology significantly reduces the uncertainty
associated with thermal losses during transfer of the test specimen from the bath to the
test machine. This approach also creates the opportunity for improvement of the
aligmnent of the specimen in the test machine. The current procedure, which is widely
accepted, is to use centering tongs to center the notch relative to the plane of the
pendulums arc. ASTM E 23 requires that the notch be centered to within 0.25 mm A
centering tool has been developed which is easy to use and which is more accurate than
centering tongs.
In order to quantify the accuracy of centering using tongs and an in-situ centering
device, CVN test specimens were placed on the specimen supports 50 times each. A
closed circuit television camera was placed over the test machine supports and anvils and
used to measure the position of the root of the notch relative to the anvils. The data are
given in Figures 5 through 7. The measurements were made from the left anvil to the
center of the notch (labeled "A" in the figures) and from the right anvil to the center of
the notch (labeled "B" in the figures). These measurements were averaged and the
standard deviation for the 50 individual measurements was calculated (see Table 2).
As shown in Figure 5, the two standard deviation for centering using the tong method
is 0.11 mm, and this is within the ASTM allowable uncertainty of 0.25 mm. Figures 6
and 7 illustrate that the in-situ centering tool enables a significant improvement in the
notch centering. As shown in Table 2, the in-situ CVN centering tool improved the
specimen centering uncertainty by about a factor of 2. Figure 7 and Table 2 show that the
MCVN specimens can be centered to the same level of accuracy as for the CVNs using
the in-situ centering tool.
An important concern related to centering of the test specimens is the adequacy of the
current E 23 +0.25 mm centering range. A preliminary investigation indicates that the
current centering range may be too large, however, an in-depth 'investigation will be
needed in the future to quantify the appropriate range for the E 23 standard. The data
from the preliminary investigation is presented in Tables 3 through 6. Tables 3 and 4
present data from tests conducted using the LL-68 material from the ASTM
Instrumented/Miniaturized Round Robin Test Program, and Tables 5 and 6 present data
for the HH-71 material from the same program. Tables 3 and 5 report data for accurately
centered specimens (centered using the in-situ centering tool) and Tables 4 and 6 report
data for specimens tested with the center of the notch offset from the plane of the
pendulums arc by 0.25 mm. In the case of the LL-68 material, the trend indicates a lower
absorbed energy for the offset specimens while the opposite trend is observed for the HH-
71 material. Although the energy average is within the two sigma range, the energy
differences observed are significant. It is also interesting to note that the shear lip data
for the LL-68 material show a trend toward a fight-left (one shear lip on each specimen
half) configuration. The right-left configuration tends to result in high absorbed energies
for the centered specimens. This effect should be studied in the future using larger
sample populations.
Conclusions
The in-situ heating and cooling technology offers significant technical advantages
over the conventional bath transfer method and it offers significant cost advantages over
complicated robotic transfer systems. The impact energy results obtained from CVN
tests using the in-situ approach have been shown to be well within the scatter of energies
obtained using the liquid bath approach. The in-situ technology significantly reduces the
uncertainty associated with thermal losses during transfer of a test specimen from the
bath to the test machine. For example, conventional CVN specimen surface temperatures
increase 3.5 ~ within 5 seconds after removal from a -40~ liquid bath. Measurements
using specimens instrumented with thermocouples have demonstrated that the desired test
temperature can be reached using the in-situ system within about 10 minutes for most
Figure 5-Conventional Charpy Specimen Centering Using Tongs
Figure 6-Conventional Charpy Specimen Centering UsingIn-situ CVN Centering Tool

Figure 7-Miniature Charpy Specimen Centering Using In-situ MCVN Centering Tool
Table 2-Summary of Specimen Centering Resultsfor Conventional Tongs and In-situ

Centering Tool
Table 3-Resultsfrom Centering Study Using Centered Low Energy 4340 Test Specimens
Brittle Fracture Shear Lip
Specimen Impact Energy Load Configuration
Identification 0) ~kS) ~OSmL)
LL68-154 27.02 35.86 RL
LL68-376 28.12 37.29 RL
LL68-463 25.63 36.58 0S
LL68-531 26.25 35.53 RL
LL68-564 24.31 35.53 OS
LL6g-1175 23.00 35.23 OS
Average 25.72 36.00
StDev 1.848 0.781
Notes: Test temperature 20~ 8 mm striker radius, impact velocity 5.47 m/s.
Shear lip locations: one-sided (OS) indicates one specimen half has both shear
lips, or right-left (RL) indicates one shear lip on each specimen half.
Table 4-Resultsfrom Centering Study Using Offset Low Energy 4340 Test Specimens
Brittle Fracture Shear Lip
Specimen Impact Energy Load Configuration
Identification (0 (kS) OS/aL)
LL68-191 23.01 35.20 OS
LL68-749 23.09 37.06 RL
LL68-788 23.01 35.88 RL
LL68-860 23.71 35.43 RL
LL68-925 24.63 37.14 RL
LL68-1010 24.25 37.36 RL
Average 23.62 36.35
StDev 0.698 0.952
Notes: Testtemperature 20.9~ 8 mm striker radius, impact velocity 5.47 m/s.
Specimens offset 0.25mm from centered position. Shear lip locations: one-
sided (OS) indicates one specimen half has both shear lips, or fight-left (RL)
indicates one shear !!p on each specimen half.
Table 5-Resultsfrom Centering Study UsingHigh Energy Centered 4340 Test Specimens
General Deflection at Shear Lip
Specimen Impact Yield Load Peak Load Peak Load Configuration
Identification Ener~ (J) (~) (kN) (mm) OS/P-L)
HH71-139 123.46 18.90 25.71 2.265 RL
HH71-209 107.99 18.32 24.74 2.116 OS
HH71-473 112.25 17.63 25.20 2.042 RL
HH71-521 114.39 18.48 24.85 2.217 OS
HH71-936 114.21 18.99 25.44 1.973 RL
HH71-1027 118.03 18.70 25.51 2.115 RL
Average 115.06 18.50 25.24 2.121
StDev 5.262 0.496 0.384 0.108
Notes: Test temperature 20.8~ 8 mm striker radius, impact velocity 5.47 m/s. Shear lip locations:
one-sided (OS) indicates one specimen half has both shear lips, or right-left (RL) indicates one
shear lip on each specimen half.
Table 6-Resultsfrom Centering Study Using High Energy Offset 4340 Test Specimens
Specimen Impact General Peak Load Deflection at Shear Lip
Identification Energy (J) Yield Load (kN) Peak Load Configuration
(kN) (mm) (OS/RL)
HH71-239 113.40 18.55 24.87 2.131 OS
HH71-351 119,46 18.98 25.72 2.165 RL
HH71-479 119.65 18.90 25.49 2.061 RL
HH71-904 123.11 18.73 25.52 2.253 OS
HH71-996 126.31 18.83 25.59 2.222 RL
HH71-10$0 116.66 18.64 25.38 2.251 OS
Average 119.77 18.77 25.43 2.181
StDev 4.567 0.162 0.296 0.076
Notes: Test temperature 20.8~ 8 mm striker radius, impact velocity 5.47 m/s. Specimens offset
0.25 mm from centered position. Shear lip locations: one-sided (OS) indicates one specimen half
has both shear lips, or fight-left (RL) indicates one shear lip on each specimen half
cases of interest. The current ASTM E 23 test standard requires the specimen to be
struck within 5 seconds of removal from the bath. The in-situ system meets the ASTM
requirement because the specimen is thermally conditioned to + 1~ right up to the time
of impact. Further, the in-situ system reduces surface temperature changes to within +
I~ This provides a significant improvement over the current approach of transferring
the thermally conditioned specimen to the test machine because in the current method
large surface temperature changes occur which can influence the crack formation energy.
The in-situ approach also creates the opportunity for improvement of the alignment of
the specimen in the test machine. The current ASTM method is to use centering tongs to
center the notch relative to the center of the pendulums arc. An in-situ centering tool has
been developed which is much more accurate than centering tongs. Measurements made
to quantify the specimen centering have shown that the in-situ centering tool can reduce
the uncertainty in specimen centering using tongs by about a factor of two. Preliminary
data indicate that the current E 23 test specimen centering range of+ 0.25 mm is too large
and should be reduced. Additional research will be needed to quantify the acceptable
range for the E 23 standard.
References
[I] Manahan, M. P., Sr., Stonesifer, R. B., Soong, Y., Burger, J. M., "Miniaturized Notch
Test Specimen and Test Machine Design, "Pendulum Impact Machines:
Procedures and Specimensfor Verification, ASTM STP 1248, edited by T. A.
Siewert and A. K. Schmieder, American Society for Testing and Materials, West
Conshocken, PA, pages 39-69, 1995
[2] "FRACTURE/FIT: Weibull Based Fracture Fitting Code", MPM Technologies, Inc.,
1997
George M. W a i d 1 and Harry Zantopulos l
The Effects of OD Curvature and Sample Flattening on Transverse Charpy

V-Notch Impact Toughness of High Strength Steel Tubular Products
Reference: Waid, G. M. and Zantopulos, H., "The Effects of OD Curvature and

Sample Flattening on Transverse Charpy V-Notch Impact Toughness of High
Strength Steel Tubular Products," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. R Manahan, Sr., Eds., American Society
Abstract: With the inclusion of transverse impact toughness requirements in the

American Petroleum Institute (API) specifications as well as various companies'
specifications, the need arose for evaluating the effects of OD curvature and sample
flattening on transverse Charpy V-Notch toughness of high strength steel tubular
products. The alloy steels evaluated in this study were AISI 4130 and a modification,
4130M7. Charpy impact tests were conducted on specimens having various amounts of
OD curvature and on sample flattening to determine the degree of influence of these
effects. Tapering the specimens by as much as 75% or this much OD curvature did not
have a significant effect on the CVN energy. Flattening of the tubes before testing
reduced the CVN energy by about 4%. The factors derived from this study allow for
adjustments on the test results of non-standard specimens so that a correlation can be
made to standard Charpy V-Notch results.
Keywords: impact toughness, toughness, Charpy V-Notch test, impact testing, high
strength steel, tubular product, AISI 4130 alloy steel
The need for improved reliability in tubular products for oil and gas production
resulted in the inclusion of impact toughness, particularly transverse requirements, in the
American Petroleum Institute (API) Specification 5CT, Specification for Casing and
Tubing [ 1, 2]. During the process of including impact toughness properties in this
specification it was suggested that consideration be given to allowing part of the
transverse impact test specimens be tapered to allow for the curvature of the tubular
product 2. Although many studies have been conducted concerning non-standard Charpy
V-Notched specimens, none have addressed the influence of specimen curvature on
impact results [3-7]. The basis of this study is to demonstrate whether or not tapered or
rounded specimens be included in this standard. In addition, various individual
companies' specifications allow for use of transverse impact toughness made from
flattened sections of tubular goods to determine their transverse impact toughness
ISenior steel product specialist and bearing product specialist, respectively, The Timken
Company, Timken Research, 1835 Dueber Ave., Canton, OH 44706.
2 W B. Smith, personal communication to J. L. Peterson, 1991.
298
WAID AND ZANTOPULOS ON STEEL TUBULAR PRODUCTS 299
properties. A study was performed to evaluate the effect of specimen flattening on these
transverse Charpy impact results.
Materials and Procedures
The test specimens used in the OD curvature study were taken from the mid-wall
location of 243.1 mm (9.571") OD x 24.5 mm (0.963") wall tubing made from a modified
4130 (4130M7) alloy steel which was developed for the oil and gas industry where high
impact toughness properties and sulfide stress cracking resistance was needed for deep
wells. This tubing was austenitized, water quenched and tempered to provide four test
groups with different strength levels. The Charpy specimens were machined to various
sizes to simulate different tubing and pipe diameters. Figure 1, taken from the API 5CT
specification, shows the OD curvature allowances permitted. The tapers, machined for
this study, ranged from no taper to 1/4 the original thickness, which is a greater taper than
specified by API 5CT.
13 5 mm
Maximum Minimum Maximum
I. ' - - 28mm ~l 13.5mmLl
OI3
Curvature
Figure 1. Charpy V-Notch Impact Test

Specimens OD Curvature Allowance
The test specimens used in the sample flattening study were taken from the mid-wall
location of a 101.6 mm (4.000") OD x 12.7 mm (0.500") wall tubing made from standard
4130 alloy steel. This tubing was also austenitized, water quenched and tempered to
provide six test groups with different strength levels. The length of tubing was then
cross-sectionally separated. One piece remained as a tube. The other piece also was
sectioned lengthwise in half, then flattened. Charpy specimens were then machined to
various sizes from both the flattened pieces and the tubing.
Chemical compositions for both heats of steel and the respective chemical ranges are
given in Table 1. Both heats were calcium-treated for sulfide shape control. Listed in
Table 2 are the heat treatments that were performed on the two alloy steels.
Table 1 - Chemical Compositions

Heat
Alloy No. C Mn Si P S Cr Ni Mo Cb Ca
-- .25 .60 .20 -- .005 1.20 .25 .65 .02
4130M7 .32 .90 .3-~ max 1.5~ max .7~ .~
-- N7602 .27 ,74 .26 .009 .004 1.34 .10 .69 .030 .0024
4130 -- .28 .40 .20 -- .005 .80 -- .15 --

.33 .60 .35 max 1.10 .25
-- 97814 .30 .50 .24 .006 .004 0.96 .14 .18 -- .0026
Table 2 - Heat Treatments
4130M7 for OD Curvature Study

Austenitized at 899 ~ (1650 ~ for 1.5 hours
Water Quenched
Tube #1 - Tempered at 621 ~ (1150 ~ for 1.5 hours
4130 for Flattening Sample Study

Austenitized at 871 ~ (1600 ~ for 1.0 hour
Water Quenched
Tube #1 - Tempered at 593 ~ (1100 ~ for 1.0 hour
Test Results
OD Curvature Study
The longitudinal tensile properties o f the 4130M7 steel are presented in Table 3. The
tensile tests were performed in accordance with A S T M E8-96a, Standard Test Methods
for Tension Testing o f Metallic Materials. The yield strengths o f these heat-treated
materials ranged from 768 MPa (111 ksi) to 1007 M P a (146 ksi), which is within the
yield strength requirements for API P110 and Q125 casing and tubing grades and only
slightly above the yield strength requirements o f API T95 grade. The Charpy specimens
were tested in accordance with A S T M E23-96, Standard Test Methods for Notched Bar
Impact Testing o f Metallic Materials, at a test temperature o f 0 ~ (32 ~ generally
required per API 5CT. The Charpy V-Notched impact results are given in Table 4. All
specimens exhibited 100 percent shear. As noted in Table 4, the impact values ranged
from 19 J (14 ft-lbs) to 150 J (110 fl-lbs) depending on tubing strength and specimen
size.
Table 3 - Tensile Properties of 4130M7 Steel for OD Curvature Study a

Ultimate
Yield Tensile Reduction
Strength Strength Elongation in Area
Code (MPa) (ksi) (MPa) (ksi) (%) (%)

Tube #l 1007 146 1082 157 18 66
Tube #2 919 133 1007 146 18 66
Tube #3 837 121 919 133 18 68
Tube #4 768 111 841 122 22 68
a Average room temperature propegies
Table 4 - Charpy V-Notch Toughness Resul~ of OD Curvature Studya
Tube 1 Tube 2 Tube 3 Tube 4
Size Taper (J) (ft-lbs) (J) (~-lbs) (J) (R-lbs) (J) (R-Ibs)
Full None 116 85 136 i00 141 104 150 110
(10 ram)
Full 90 113 83 128 94 139 102 148 109

(10 ram) (7.5 ram)
Full 'A 118 87 133 98 141 104 151 111

(10 mm) (5 ram)
Full 88 124 91 133 98 147 108 154 113

(10 rnm) (2.5 mm)
90 None 82 60 94 69 98 72 109 80
(7.5 ram)
90 90 84 62 91 67 101 74 105 77
(7.5 ram) (5.62 mm)
90 89 82 60 91 67 99 73 110 81
(7.5 mm) (3.75 ram)
90 88 84 62 90 66 105 77 113 83
(7.5 ram) (1.88 mm)
89 None 50 37 54 40 57 42 60 44
(5 ram)
89 % 49 36 53 39 56 41 58 43
(5 ram) (3.75 mm)
89 % 50 37 53 39 58 43 64 47
(5 ram) (2.5 mm)
89 88 52 38 54 40 60 44 67 49
(5 mm) (1.25 ram)
88 None 19 14 20 15 22 16 22 16
(2.5 mm)
a A v e r a g e o f n i n e i n d i v i d u a l tests
Sample Flattening Study
The longitudinal tensile properties of the 4130 alloy steel are exhibited in Table 5.
These tests were also conducted in conformance with ASTM E8-96a. The yield strengths
of these quenched and tempered tubes ranged from 587 MPa (85 ksi) to 887 MPa (129
ksi), which is within the requirements for API N80, L80, C90, C95, T95 and P110 casing
and tubing grades. In most cases, the longitudinal tensile properties for the tubing were
higher than for the flattened materials for a particular heat treatment because of the
Bauschinger effect. Unfortunately, because of the wall thickness and the diameter of the
tubing and the lack of material in the case of the flattened pieces, transverse tensile
properties were not obtained. The Charpy impact toughness results are presented in
Table 6. These specimens were also tested in accordance with ASTM E23-96, and tested
at 0 ~ (32 ~ per API 5CT. All specimens exhibited 100 percent shear. These impact
energies ranged from 18 J (14 ft-lbs) to 167 J (123 ft-lbs) depending on the strength and
specimen size.
Table 5 - TensilePropertiesof 4130 Steelfor FlatteningStudy

Ultimate
Yield Tensile Reduction
Strength Strength Elongation in Area
Code (MPa) (ksi) (MPa) (ksi) (%) (%)

T1 887 129 952 138 22 68
F! 880 128 958 139 21 69
T2 837 121 910 132 21 69
F2 772 112 903 131 22 70
T3 758 110 841 122 23 72
F3 720 104 855 124 22 70
T4 720 104 793 115 25 72
F4 676 98 800 116 23 72
T5 667 97 731 106 26 72
F5 623 90 745 108 25 73
T6 596 86 703 102 27 74
F6 587 85 703 102 26 74
Table 6 - Charpy V-Notch Toughness Results of Flattening Study ~
T1 F1 T2 F2 T3 F3
Size (J) (ft-lbs) (J) (ft-lbs) (J) (ft-lbs) (J) (ft-lbs) (J) (ft-lbs) (J) (ft-lbs)
Full 132 97 117 86 136 100 132 97 146 107 141 104
(lOmm)
3/4 79 58 79 58 84 62 83 61 90 66 96 71 tO
>
(7.5 ram) z
2/3 68 5O -- -- 72 53 . . . . 84 62 79 58
(6.7 ram) N
89 49 36 48 35 52 38 52 38 52 38 61 45
(s ram) ---I
9
88 19 14 18 13 19 14 20 15 . . . . . . . . "0
c"
(2.5 ram) r
9
GO
0
T4 F4 T5 F5 T6 F6 z
Size (J) (ft-lbs) (J) (ft-lbs) (~ (R-lbs) (J) (~-lbs) (J) (~-lbs) (J) (~-lbs)
m
m
Full 147 108 156 115 162 119 158 116 178 131 167 123 r
(10 mm)
90 109 80 102 75 113 83 107 79 118 87 107 79 C
(7.5 ram) C
2/3 86 63 79 58 91 67 86 63 98 72 101 74
(6.7 mm)
89 60 44 58 43 57 42 60 44 62 46 71 52
(5 mm) O
(3
1//4 . . . . . . . . . . . . . . . . . . . . . . . . C
o
(2.5 mm)
c/~
a Average of six individual tests
GO
0
GO
Discussion
OD Curvature Study
The initial analysis of the Charpy V-Notch impact results involved determining the
absorbed energy reduction factors for the various specimen sizes and strength levels and
compares these results with the factors listed in API 5CT. This comparison is presented
in Table 7 where the data for each group is normalized to the full thickness specimen (10
mm). The results of this evaluation give significantly more conservative factors than
those listed in the API 5CT specification. One reason that these reduction factors are
lower might be that only quenched and tempered high strength products that exhibited
high impact toughness which were consistently on the upper shelf o f the Charpy
transition curve were used in this study, whereas, the API specification covers a much
wider range o f tubular products. Also, this table shows that the reduction factors remain
essentially the same regardless o f the amount o f taper that the specimen possesses. In
addition, a prediction equation was obtained using the same test results.
CVN = 2343 ~-0.871 w 1.335 (1)
R 2 = 0.9949
RMSD = 3.78%
where
CVN = Charpy impact toughness, J

o = yield strength, MPa
w = specimen width, mm
R = correlation coefficient
RMSD = root mean square deviation
R z gives the fraction of the total variation, which is explained by the regression
equation. In addition, a dispersion parameter, the root mean square deviationz(RMSD ),
was found to be 3.78%. With only the specimen width w in the regression, R = 0.9671
with a RMSD o f 9.70%. Thus, the addition o f the yield strength c~ has a significant effect
on the regression. However, it should be obvious that the most influential dependent
variable is the specimen width, w. Next, the effect o f the taper or OD curvature was
considered. This only increased R 2 by 0.0012 with a decrease in RMSD to 3.29%. This
was not considered to be statistically significant, i.e., the taper or OD curvature does not
have a significant effect on the value of CVN obtained. The relatively good fit of the
calculated to actual values of CVN using Eq. (1) is graphically illustrated in Figure 2.
Table 7 - R e d u c t i o n Factorsfor Tapered Specimens
Yield Strength
Width Taper 1007 MPa 919 MPa 837 MPa 768 MPa Avg API 5CT
10 m m None 1.00 1.00 1.00 1.00 1.00 1.00
10 m m 90 0.98 0.95 0.98 0.99 0.97 --

(7.5 ram)
10 m m 89 1.02 0.99 1.00 1.01 1.01 --
(5.0 ram)
10 m m 88 1.07 0.98 1.04 1.03 1.03 --
(2.5 ram)
10 m m Avg 1.02 0.98 1.01 1.01 1.00 --
7.5 m m None 0.71 0.69 0.70 0.72 0.71 0.80
7.5 m m 90 0.73 0.68 0.71 0.70 0.70 --

(5.62 mm)
7.5 m m 89 0.70 0.68 0.70 0.73 0.70 --
(3.75 mm)
7.5 m m 'A 0.73 0.66 0.74 0.75 0.72 --
(1.88 ram)
7.5 m m Avg 0.72 0.68 0.71 0.73 0.71 --
5 mm None 0.43 0.41 0.41 0.40 0.41 0.55
5 mm 90 0.43 0.39 0.40 0.39 0.40 --

(3.75 m m )
5 mm 89 0.43 0.39 0.41 0.43 0.42
(2.5 ram)
5 mm 1/4 0.45 0.41 0.43 .044 0.43
(1.25 ram)
5 mm Avg 0.43 0.40 0.41 0.42 0.42
2.5 m m 100 0.16 0.15 0.16 0,14 0.15 --

180
"[3
z
160
140
H.41, I
120
100
0
z
~ 80 i>
J
60
I,YS = 1007 MPa
I Y S = 919 MPa
40
kYS = 837 MPa
| Y S = 768 MPa
20 33
I
__L
20 40 60 80 100 120 140 60 180
ACTUAL CVN (J)
Figure 2 - Comparison o f Actual to Calculated Values Using Eq. (1)

Sample Flattening Study
T h e a b s o r b e d energy reduction factors for the various s p e c i m e n sizes and strength

levels in the flattening study also were d e t e r m i n e d b o t h for t u b i n g and flattened tube
sections and c o m p a r e d w i t h the A P I 5CT factors. This c o m p a r i s o n is g i v e n in T a b l e 8,
w h e r e the data for each group is nomaalized to the full thickness s p e c i m e n (10 m m ) . A s
w i t h the p r e v i o u s study, the evaluation shows these factors to b e m o r e c o n s e r v a t i v e t h a n
those listed in the A P I 5 C T standard. H o w e v e r , the factors are similar b e t w e e n the tube
a n d flattened samples. Also, prediction equations were acquired u s i n g these C h a r p y data.
F o r t u b i n g samples:
C V N = 1489 cy ~ 1.393 (2)
R 2 = 0.9942
RMSD = 4.16%
F o r flattened samples:
C V N = 2123 a -0.88i w 1.334 (3)
R 2 = 0.9939
RMSD = 4.33%
T h e s e equations give similar results, with Eq. (2) predicting, o n the average, values
a p p r o x i m a t e l y four percent higher than those o f Eq. (3). This can b e explained, since cold
w o r k i n g will decrease i m p a c t toughness [8], T h e flattened samples appear to b e
r e a s o n a b l y good indicators o f the t o u g h n e s s o f the tubing, i f the i m p a c t t o u g h n e s s o f the
material is o n the u p p e r s h e l f o f the t e m p e r a t u r e transition curve.
'Fable 8 - ReductionFactorsfor SampleFlatteningStudy
Tubing
Yield Strength
Width 887 MPa 837 MPa 758 MPa 720 MPa 667 MPa 596 MPa Avg. API 5CT
10 mm 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
7.5 mm 0.59 0.62 0.62 0.73 0.70 0.66 0.65 0.80
6.67 mm 0.51 0.53 0.58 0.58 0.56 0.55 0.55 --
5 mm 0.36 0.38 0.36 0.40 0.35 0.35 0.37 0.55
2.5 mm 0,14 0.14 . . . . . . . . 0.14 --
Flattened Tubing
Yield Strength
Width 880 MPa 772 MPa 720 MPa 676 MPa 623 MPa 587 MPa Avg. API 5CT
10 m m 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
7.5 mm 0.67 0.63 0.68 0.65 0.68 0.64 0.66 0.80
6.67 mm . . . . 0.58 0.50 0.54 0.60 0.55 --
5 mm 0.41 0.39 0.43 0.37 0.38 0.42 0.40 0.55
2.5 mm 0.15 0.16 . . . . . . . . 0.15 --
Conclusions
The OD curvature of transverse Charpy specimens essentially has no effect on the

impact toughness results of quenched and tempered high strength tubular products. The
flattening of tubular samples to obtain wider transverse Charpy specimens reduces the
impact toughness of the product by approximately 4%, but is a good predictor provided
the specimens exhibit 100 percent shear. The differences in the reduction factors between
the two studies are not significant. However, the reduction factors for both are
consistently lower than those factors given in the API 5CT specification.
References
[ 1] Shoemaker, A. K., "Application of Fracture Mechanics to Oil Country Tubular

Goods," API Standardization Conference, API Pipe Symposium, Denver, CO,
June 19, 1989.
[2] Burk, J. D., "Fracture Resistance of Casing Steels for Deep Gas Wells," Journal of
Metals, January 1985, pp. 65-70.
[3] McConnell, P., Sheckherd, J. W., Perrin, J. S., and Wullert, R. A., "Experience in
Subsized Specimen Testing," The Use of Small-Scale Specimens for Testing
Irradiated Material, ASTMSTP 888, W. R. Corwin and G. E. Lucas, Eds., American
Society for Testing and Materials, Philadelphia, 1986, pp. 353-368.
[4] Louden, B. S., Kumar, A. S., Garner, F. A., Hamilton, M. L., and Hu, W. L., "The
Influence of Specimen Size on Charpy Impact Testing ofUnirradiated HT-9,"
Journal of Nuclear Materials, 1988, pp. 662-667.
[5] Corwin, W. R. and Hougland, A. M., "Effect of Specimen Size and Material
Condition on the Charpy Impact Properties of 9Cr-1Mo-VrNb Steel," The Use of
Small-Scale Specimens for Testing Irradiated Materials, ASTM STP 888, W. R.
[6] Lucas, G. E., Odette, G. R., Sheckherd, J. W., McConnell, P., and Perrin, J.,
"Subsized Bend and Charpy V-Notch Specimens for Irradiated Testing," The Use of
Small-Scale Specimens for Testing Irradiated Material, ASTM STP 888, W. R.
[7] McNicol, R. C., "Correlation of Charpy Test Results for Standard and Non-standard
Size Specimens," Welding Journal, Welding Research Supplement, September 1965,
pp. 385s-393s.
[8] Tonaka, T., Kunekoshi, T., Ueda, M., Tsuboi, J., Yasuda, T., and Utahashi, C.,
"Developmentof High-Strength Steel with Good Toughness at Arctic Temperatures
for Large-Diameter Line Pipe," Proceedings Microalloying '75, Union Carbide
Corporation, New York, 1977, pp. 399~409.
Robert K. Hughes 1 and Brian F. Dixon 1
Electron Beam Welded Charpy Test Specimen for Greater Functionality
Reference: Hughes, R. K. and Dixon, B. F., "Electron Beam Welded Charpy Test
Specimen for Greater Functionality," Pendulum Impact Testing: A Century of
Progress, STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American Society
Abstract: Applying an electron beam weld (EBW) at the base of the notch has modified
Charpy test specimens to provide a reduced scatter, sharper transition, higher upper-shelf
energy and higher transition temperature for tough, high-strength steels. The procedure
considerably improves determination of the ductile-brittle transition temperature (DBTT)
and the reproducibility of the test.
Sectioning of the specimens after testing showed that the weld acts as a brittle crack
starter at temperatures up to the transition temperature. In other words, the weld initiates
cracking without absorbing significant energy itself. Therefore the test is measuring the
energy necessary to drive a sharp crack through the 5.8-mm-wide unwelded ligament.
At temperatures above the transition, the fracture path switches from the weld zone to
lower strength, ductile parent metal. The fracture is ductile because cracking initiates and
travels through the lower strength parent metal that is ductile at these temperatures. This
reflects the practical situation in which brittle weld zones may be present in steel that is
inherently tough at the testing temperature.
Keywords: Charpy impact testing, steels, high strength steels, modified Charpy
test specimen, electron beam weld, ductile-brittle transition, upper shelf energy,
scatter
Background
The Charpy test has been established for 100 years and is accepted worldwide as a
quality assurance test for determining the resistance of steel to brittle fracture. While the
test is incapable of measuring fundamental material properties, such as fracture
toughness, it provides a reliable, reproducible and inexpensive procedure for use in an
industrial environment.
1Research Scientist and Principal Research Scientist, respectively, DSTO Aeronautical

and Maritime Research Laboratory, PO Box 4331 Melbourne, VIC Australia 3001.
310
HUGHES AND DIXON ON ELECTRON BEAM WELDED CHARPY TEST 311
The purpose of this work was to investigate alternative designs for the Charpy test
specimen that could provide a more useful procedure while maintaining the simplicity of
the conventional technique. The conventional design is 55 mm long and 10 mm square
with a 2 mm notch across one side. The specimen is fractured by striking it on the side
opposite the notch using a hammer of known impact energy. A limitation of the test is
that the Charpy notch is blunt relative to defects such as fatigue and hydrogen cracking
that may be found in service. Therefore a significant component of the energy required
to break the Charpy specimen is absorbed in initiating the crack and this may swamp the
information of most interest, the energy required for fracture propagation.
The work described here looks at two ways of modifying the Charpy notch to
overcome problems of bluntness while maintaining the simplicity and cheapness that
makes the test popular. The first design was proposed by the authors and involves
depositing an autogenous electron beam weld (EBW) along the base of the notch. The
weld was intended to provide a potential brittle "crack starter" for the low-energy
initiation of brittle, running cracks in the specimen.
The second technique to be investigated involved Electric Discharge Machining
(EDM) to provide a sharp slit, 2.5 mm deep and 0.t5 mm wide, at the base of the
standard notch (Figure 1). This technique was proposed by Sumpter [1] as part of an
international investigation into alternative test specimen designs.
Experimental Procedure
The EBW specimens were prepared by traversing an electron beam along a stack of
Charpy specimens with notches aligned. The fusion zone profile had an arrowhead cross-
sectional shape with a depth of about 2.2 mm and a maximum width, just below the notch
root, of 0.8 mm (Figure 2). Welding parameters are provided in Table 1. The EDM
specimens were prepared in the same manner as that used by Sumpter. The depth of the
EDM notch was 2.5 mm. That is, 4.5 mm total depth or roughly the same as the total
depth of the electron beam weld.
To measure the effect on energy transition and upper shelf energy for three different
steels, full energy transition curves were generated for each of the three specimen
designs. A striking edge profile which complied with the I SO R148-1960 test method
was used for all tests.
Two criteria were used to define the ductile to brittle transition temperature (DBTT);
the temperatures at which impact energy equaled 85% and 50% of the upper shelf energy.
Table 1 - Parametersfor electron beam weld along the (?harpy notch
Machine Wentgate DW604

Voltage 60 kV
Current 9 mA
Travel Speed 16.6 mm/s (1.0 m/min)
Beam Focus Sharp focus
Figure 1 - Detail of Charpy notch showing EDM slit proposed by Sumpter [I].
Note: not to scale, all dimensions mm.
Figure 2 - Macrograph of section at notch root showing profile of electron beam weld.
Steel." 350141"[ Etchant: 2% nital.
Descriptions and compositions of the steels selected for this work are provided in
Table 2. Both the T-L and L-T specimen orientations were used for each steel.
For the X-80 steel it was not possible to obtain full (10 mm) thickness specimens
because the plate gauge was 8.6 mm. Sub-size specimens of 7.5 mm thickness were
therefore used.
Results
Energy transition curves for the specimens are provided in Figure 3 to Figure 5 and
these are described below. Estimated upper shelf energy for all tests is shown in Table 3.
The DBTT measurements for the three techniques and two orientations are shown in
Table 4.
BIS812EMA (Figure 3)
Ductile-Brittle Transition Temperature (DBTT) - For BIS812EMA steel both the

EBW and EDM DBTT were higher than standard Charpy test specimens if the DBTT
was taken to be 50% of upper shelf energy. The increase in DBTT was generally similar
for EBW and EDM specimens in the L-T orientation.
The DBTT was very sharply defined for EBW in L-T orientation. In the T-L
orientation, both the EBW and EDM had some scatter in the DBTT.
Upper Shelf Energy - The upper shelf energy was approximately the same for standard
Charpy and EBW Charpy specimens. Both had more than double the upper shelf energy
of the EDM specimens.
The upper shelf energy was 25% greater for L-T orientation compared with T-L
orientation for standard specimens and 18% greater for EBW specimens. The upper shelf
energy was approximately the same in both orientations for EDM specimens.
X-80 (Figure 4)
Ductile-Brittle Transition Temperature (DBTT) - While electron beam welding

(EBW) and electric discharge machined (EDM) raised the ductile-brittle transition
temperature (DBTT) above the value for standard (sub-size) specimens, the DBTT is
difficult to clearly identify for both standard Charpy and EDM. The DBTT is very
sharply defined in both orientations for EBW with very little scatter in the transition
region.
Upper Shelf Energy - The EBW upper shelf energy was considerably higher than
standard Charpy upper shelf energy. The EDM was significantly lower than the standard
Charpy.
The upper shelf energy for L-T orientation was 20% greater than T-L orientation in
standard Charpy specimens and 38% greater in EBW specimens. The upper shelf energy
was approximately the same in both orientations for EDM.
Table 2 - Steels chosenfor testing of Charpy specimen design
A) Properties
Steel BIS812EMA X80 350WT
Designation
Origin Australia Australia Canada
Steel Quenched and High strength thermo- High toughness
Type tempered submarine mechanically shipbuilding
pressure hull steel processed gas linepipe steel
steel
Nominal 690 MPa 550 MPa 400 MPa
minimum yield stress
Plate 35 mm 8.6 mm 12 mm
Thickness
Test Specimens 10mm 7.5mm 10mm
Thickness l
Test Specimens 1 mm below surface mid-section mid-section
Location
B) Composition
Analysis
(Wt.%)
C 0.13 0.07 0.07
Mn 0.93 1.61 1.43
Si 0.24 0.30 0.24
S 0.002 0.002 0.006
P 0.01 0.012 0.013
Ni 1.28 0.02 0.02
Cr 0.48 0.01 0.05
Mo 0.39 0.27 0.01
Cu 0.21 0.02 <0.01
V 0.02 0.0074 0.062
Nb 0.01 0.060 0.020
Ti 0.01 0.020 0.010
A1 0.07 0.035 0.037
B 0.0066 <0.0005 <0.0005
N 0.0040 0.0061 0.0076
0 <0.001 0.001 0.001
~Specimen thickness is the length along the bottom of the notch

250-~ i 7 250
IlTransverse " ~ l [
o StandardCharpy[..........y ,..~-.............,...
~
= .~../200
[
9 EBW Charpy I " ~ / .................. [................ [-150
9 EDM Charpy / ~ [] ' /
'~ I ~ .....~ .............tlOO

~o-I..........:~,,Y ........'........... ', ................ t ~o
o - ~ o
300-i
I L~ * 1
, -
iiiiiiiiiiiiiiiiiiiiii
,oo- I ............... . . . . . . . . . . . . . . . . . . . 100
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
04 , , I o
I !
Temperature(~
Figure3 - Transition curves f o r BIS 812 EMA steel.
2oo- i ..... ,. . . . . - 200
Transverse i
u
IStandardCharpyl--~r OI
~ ..................150
@ EBWCharpy I |1 I__..dl__~
o ~O..arpy , - - - ~ ~ . . . . . . . . . . . . . . . . ,oo
i . _ J l ~
I0-~
I 0 p "
s 2..' .... ~. . . . . . . . . so
o-~'- v, " , , I , o
250- i ,---. . . . . .--250
-i ',ongitu.inal Y"--- I
I ............................... ........ . :..... ..... ioo
D~II
If"
100-5................................... ..................... 10050
O-I r ' ~ , , , , ~ 0
I !
Temperature(~
Figure4- Transition curves f o r X-80 steel (7. 5 mm thickness subsize specimens).
400-] Transverse - 400

|
[] Standard Charpy
9 EBWCharpy ~f" L I 300
0 EDMCharpy [] ~ []
Iv I
......................... ....... j ...... .._,, ................
~ / 00 I
100- ................... ~ - ,- - ,-~ ~ ............... 100
~ o 400--
- ~ o - 400
el.
2-- Longitudinal
300- =---- ..... [] ...................... 300

[]
200-
i ..............................
I
200
y o 6
100- ............ 1................ 100
I
I
O_ I , I o
'r, "7,
Temperature (~
Figure 5 - Transition curves f o r 350 WT steel. (Arrows indicate that the hammer was
stopped).
3 5 0 W T (Figure 5)
Ductile-Brittle Transition Temperature (DB77") - For both the longitudinal and

transverse orientations, the EBW and EDM specimens had higher DBT'I- values than the
standard Charpy specimens. The increase in DBTT was somewhat similar for both the
EBW and EDM specimens. The DBTT was very sharply defined in both orientations for
EBW whereas the standard Charpy test specimen had some scatter.
Upper Shelf Energy - The upper shelf energy could not bc determined for EBW
specimens in the L-T orientation since these stopped the pendulum at -20~ and higher
temperatures. However it was clear that the EBW and standard Charpy results were
significantly higher in upper shelf energy than the EDM specimens.
The upper shelf Charpy energy was 25% greater for L-T orientation compared with
T-L orientation. The upper shelf energy for EDM was slightly higher in the L-T direction
than in the T-L direction.
Fracture Path (Figures 6 - 8)
Metallographic sections across representative upper shelf and lower shelf Charpy
specimens are provided in Figure 6 and Figure 7. These macrographs illustrate that
cracking occurs around the weld heat affected zone when the fracture is ductile (Figure 6)
and through the electron beam weld when the specimen behaves in a brittle manner
(Figure 7). Fracture surfaces are shown in Figure 8 for the two types of fracture.
Discussion
The Charpy test reveals the transition in steels from ductile behaviour at high
temperatures to brittle fracture at lower temperatures. It is often difficult to accurately
measure the ductile-to-brittle transition temperature because generally the transition is not
sharp and no single criterion exists to define the DBTT. The range of arbitrary
definitions includes the lowest temperature at which all results exceed a given percentage
of upper shelf energy, typically 85% or 50%, or the temperature at which all results
exceed half the energy difference between upper and lower shelves. Further possible
DBTT criteria include the lowest temperature for upper shell" fracture, the temperature for
some specific energy, or the nil ductility temperature (NDT) at which the lowcr shelf is
reached [2]. If a steel exhibits a gradual transition from upper to lower shelf in the
conventional Charpy test, the various DBTT criteria will occur over a wide range of
temperatures. Difficulty in identification of the DBTT will be further compounded if
scatter in results occurs in the transition region. This may be attributable to variations in
the properties of the steel or possible difficulty in providing perfectly reproducible
notches [3,4].
Table 3 - Estimated upper shelf energy." ('~ indicates upper shelf not achieved)
Upper shelf Increase in upper

energy shelf energy:
(J) T-L --~ L-T
(%)
T-L L-T
BISS12EMA
Standard Charpy 200 250 25
EBW Charpy 220 260 18
EDM Charpy 105 t 05 0
X80
Standard Charpy, sub-size t04 125 20
EBW Charpy, sub-size 160 220 38
EDM Charpy, sub-size 62 68 10
350WT
Standard Charpy 260 325 25
EBW Charpy 315 350 1" 11"1"
EDM Charpy 105 125 19
Table 4 - Ductile to brittle transition temperatures at 85% and 50% of upper shelf energy
DBTT DBTT Change in DBTT

85% upper 50% upper if change criteria:
shelf shelf 50% -4 85%
(~ (~ upper shelf energy
(*C)
T-L L-T T-L L-T T-L L-T
BIS812EMA
Standard Charpy -50 -65 -85 -95 35 30
EBW Charpy -70 -60 -80 -65 10 5
EDM Charpy -50 -45 -65 -60 15 15
X80
Standard Charpy, Sub-size -70 -80 -100 -140 30 60
EBW Charpy, sub-size -60 -40 -70 -60 10 20
EDM Charpy, sub-size -30 -65 -95 -130 65 65
350WT
Standard C h a r y -65 -75 -70 -80 5 5
EBW Charpy -50 -40 -50 -40 0 0
EDM Charpy -45 -55 -50 -60 5 5
Figure 6 - Section across upper-shelf specimen showing fracture through ductile metal
around the electron beam weld. Steel: 350WZ, Etchant: 2% nital.
Figure 7 -Section across lower-shelf specimen showing fracture through the electron
beam weld and thence into parent metal. Steel: BIS812EMA, Etchant: 2% nital.
Figure 8 - Fracture surfaces of EBW Charpy specimens. Steel: BIS812EMA, full size,
10 mm square section.
a: brittle type fracture with crack passing through the electron beam weld.
b: ductile type fracture with erack passing through parent metal.
There is little evidence of shear lips at the welded part of specimen a.
Electron Beam Welding
The EBW design provided a sharper, more clearly defined transition and a higher
upper shelf energy than the conventional Charpy specimens for the three steels
investigated. By providing a sharper ductile to brittle transition, the EBW technique
provides increased precision in identifying the DBTT with the various measurement
criteria spread over a much narrower band of temperatures. It is also arguable that fewer
specimens would be required to identify the more clearly defined DBTT when using this
modified specimen.
The sharper transition given by the EBW design is confirmed by the minimal change
in DBTT if the criteria is changed from 50% to 85% of upper shelf energy, as shown in
Table 4. While the ductile-brittle transition is generally shifted to higher temperatures by
the introduction of the weld, the extent is dependent on DBTT measurement criteria and
specimen orientation.
The thermal cycle produced by the electron beam induces very rapid solidification and
cooling beneath the notch. The weld region will therefore be likely to have properties
which vary markedly from the surrounding steel. In particular, the weld zone is
considerably harder and more brittle than the unaffected metal. It will consequently act
as a metallurgical notch at the base of the standard geometric notch. Sections across the
notch (Figures 6 and 7) clearly show that the electron beam weld has the effect of
enhancing both the brittle (lower shelf) and ductile (upper shelf) behavior. For metal
exhibiting upper shelf behavior (Figure 6), the high yield stress weld deposit causes
cracking to travel around the weld heat affected zone where greater energy is absorbed.
For specimens having lower shelf characteristics (Figure 7) the electron beam weld
behaves as a hard, brittle crack starter which reduces energy to initiate crack propagation
and causes rapid growth of a running crack.
Small regions of fibrous fracture appearance, known as shear lips, were visible at the
edges of the brittle type specimen (Figure 8, specimen a) where triaxial constraint was the
least. These shear lips were not apparent in the weld region below the notch indicating a
change in the components of the triaxial state of stress. This provides evidence that when
fracture propagation changes from around the weld to through the weld a higher level of
triaxial constraint is encountered contributing to the sharp transition from ductile to brittle
behavior.
EDM Notch
For the three steels, the EDM notches have caused a general shift in transition
behavior to higher temperatures, relative to the standard Charpy specimens, however the
transition is not always clear and the upper shelf energy is consistently significantly
lower than that for the standard or EBW Charpy notches. This illustrates that the sharper
notch has resulted in reduction of the energy necessary to initiate cracking and hence a
shift in transition temperature to higher values. The reduction in upper shelf energy for
EDM specimens can be explained with reference to the 2.5 mm slot at the base of the 2
mm deep Charpy notch. This means that the modified specimen has a ligament of 5.5
mm compared with 8 mm available in the standard test specimen. This reduction in
energy does not occur with the EBW specimens because the fracture path is around the
weld (Figure 6) and therefore roughly the same distance as the standard Charpy specimen
undergoing ductile fracture.
Cost of Testing
The simplicity of the conventional Charpy test, the existence of a large database of
results, and the widespread acceptance by industry has established the test for perpetuity.
Numerous attempts have been made to modify the technique to improve the quality of the
information available without substantially changing the test procedure. While the use of
instrumentation and possible fatigue pre-cracking of specimens will enable collection of
quantitative fracture mechanics information, it will also necessitate a more complex test
procedure. Both the EDM slit and EBW weld are simple modifications to the Charpy
specimen but the EBW technique is the simpler and least expensive because a large
number of specimens can be electron beam welded in one weld pass.
The use of a weld at the base of the Charpy notch also overcomes some of the
concerns raised about testing the toughness of metals in the welded condition.
Specifications for the toughness testing of weld designs for procedure qualification
generally require welded coupons to be Charpy tested at the fusion boundary, 3 mm from
the fusion boundary on the parent metal side, and the weld metal. There has been some
debate about Charpy testing of fusion boundaries and heat-affected zones (HAZs)
because it is often difficult to accurately locate the notch and the fracture path may
deviate from the region under investigation.
This EBW design actually combines the testing of weld metal, heat affected zone and
parent metal in the one test. For each test temperature, the crack travels along the path of
least resistance. If the weld is brittle, this initiates fracture. If the whole weld zone is
relatively tough, then fracture occurs around the weld HAZ. In other words, the test
actually selects the path of least resistance. For this reason it is also arguable that the
number of tests required may be reduced
Conclusion
Overall, the use of an electron beam weld at the base of a Charpy notch offers a
number of significant advantages over the conventional Charpy specimen when testing
tough high-strength steels. For upper-shelf behavior, the weld causes cracking to travel
around the weld HAZ where the path is slightly longer than in the conventional Charpy
test and the energy absorbed is higher. In other respects, the upper-shelf fracture is
equivalent to the standard Charpy test.
For lower-shelf behavior, the weld acts as a crack starter. The total path of the crack
is equivalent to that of a conventional Charpy test; however, the energy absorbed in crack
initiation is almost eliminated. Thus the DBTT temperature is shifted upward, the energy
transition curve is clearly defined and the DBTT is readily identified.
Use of an electron beam weld at the base of Charpy notches adds a small cost to the
preparation of specimens. However, it is probable that the number of specimens required
can be reduced by use of this technique because the DBTT is more easily defined and,
more particularly, be~cause many of the weld procedure qualification tests may be
eliminated.
The EDM specimens provide a considerably reduced upper-shelf energy by
comparison with the EBW and standard specimens, and a DBTT generally higher than
that of standard specimens. Determination of DBTT is not significantly improved by the
use of the EDM slit.
References
[1 ] Sumpter, J.D.G., "Fracture Safety of High Strength Steels", International

Symposium on Safety in Application of High Strength Steel, Trondheim, Norway,
July 1-2 1997.
[2] Dieter, G.E., Mechanical Metallurgy, Third Edition, McGraw-Hill Book Company,
1986 p. 477.
[3] Bailey, N., "Bimodality revisited - split behaviour of weld metal", TWI Bulletin,
September-October 1991, pp. 110-115.
[4] Bishop, T.A., Markworth, A.J. and Rosenfield, A.R., "Analyzing Statistical
Variability of Fracture Properties", Metallurgical Transactions, Volume 14A,
1983, pp. 687-693.
Fracture Toughness Assessment
from Impact Test Data
Wolfgang B6hme I and Hans-Jakob Schindler 2
Application of Single-Specimen Methods on Instrumented Charpy Tests:

Results of DVM Round-Robin Exercises
Reference: B6hme, W. and Schindler, H.-J., "Application of Single-Specimen

Methods on Instrumented Charpy Tests: Results of DVM Round-Robin Exer-
cises," Pendulum Impact Testing: A Century of Progress, STP 1380, T. A. Siewert
and M. R Manahan, Sr., Eds., American Society for Testing and Materials, West
Abstract: Based on a previous round-robin exercise of the German DVM task group on
"Instrumented Impact Testing" with about 400 instrumented Charpy-specimens being
tested, another exercise was performed to explore the possibilities of extended evaluations
such as single-specimen methods. The previously determined multi-specimen cleavage
JR-CUrVe serves as a reference. Different methods - e.g., the key-curve method - were
applied. Overall, the results are encouraging and the agreement of the calculated JR-Curve
with reference data is quite good especially for large amounts of crack extension. There
was some uncertainty left in the first part of the JR-Curve, where notch blunting and initia-
tion of crack growth take place, and where the calculated curves tend to overestimate the
crack resistance. Results of a new "blind" round-robin indicate improved results in this
initial part of the JR-CUrve.
Keywords: Charpy test, impact, instrumented impact testing, fracture toughness, crack
resistance curve, single-specimen method, key-curve method, dynamic material behavior
Introduction
For existing structures in operation there often is not enough material available to
quantify the actual quality of materials. Then testing procedures with small specimens
and preferable single-specimen procedures are recommended. The accuracy of single-
specimen methods has been investigated by the DVM task group on "Instrumented Ira-
I Research scientist, Fraunhofer-lnstitut fiir Werkstoffmechanik (IWM), W6hlerstrasse 11,

D - 79108 Freiburg, Germany.
2 Research scientist, Swiss Federal Lab. for Materials Testing and Research (EMPA),
Ueberlandstrasse 129, CH - 8600 Diibendorf, Switzerland.
327
pact Testing." These investigations are based on results of a previous round-robin exer-
cise with about 400 instrumented Charpy tests, where three different types of notches -
standard V-notch, fatigue-crack and spark-eroded slot - were considered. The effect of
the notch geometry and the accuracy of instrumented Charpy tests were the major points
of interest of this exercise. The results are reported in [1].
As a side-product, a multi-specimen JR-Curve could be determined by the evaluation
of experiments in the transition region as described below. This led to the idea to extend
the round-robin to explore the possibilities of single-specimen methods to evaluate JR-
curves when applied to instrumented Charpy tests. Six of the original seventeen partici-
pants took part in this task. The participants were free to use an evaluation method of
their choice. Two used a key-curve method, one an analytical method based on a two-
parameter fracture model to evaluate the measured force-displacement curves, and two
applied semi-empirical correlation functions to estimate the JR-Curvejust from the
Charpy energy. In addition, low-blow tests using specimens with deeper fatigue cracks
were performed. The results were encouraging, however, a reduced accuracy at small
crack extensions was observed (see [2, 3]).
Actually, the applied single-specimen methods were developed and mainly used for
the evaluation of standard precracked test specimens, not for Charpy specimens. Though
rather successfully applied to the Charpy specimens as mentioned above, this round-
robin exercise was not able to qualify or disqualify the different methods. In order to
further explore the potential of single-specimen methods, another "blind" round-robin
was initiated within the DVM group in 1998. Some preliminary results are given in this
report. They indicate that the accuracy at small crack extensions can be improved.
Multi-Specimen JR-Data as Reference Data
During the DVM round-robin [1] the accuracy of instrumented Charpy tests and the
influence of the notch shape was investigated. Three different notch shapes were consid-
ered: V-notches, crack-like spark eroded slots (notch root radius: 0.02 mm), and fatigued
cracks, all of a depth of 2 mm.
From all the experiments in the brittle-to-ductile transition regime a so-called
"cleavage R-curve" was evaluated. Therefore, the partial'energy Wiu was determined by
integrating the force-displacement curve up to the sudden drop of the force signal, which
indicates the end of ductile crack extension by initiation of cleavage fracture. Then a cor-
responding J-integral value was calculated by applying the equation:
W~. (1)
J " "~ rl B ( W _ ao )
with B being the thickness, W the width and ao the notch- or crack-depth, respectively, thus
for the Charpy specimens B -- W = 10 mm and ao= 2 mm. The factor rl was chosen to be 1.46
for ao/W = 0.2 according to [4].
In addition, the ductile crack extension up to the onset of cleavage fracture, Aa, has to
be measured on the fracture surface. These data as obtained from a series of experiments in
the transition range are forming a dynamic multi-specimen crack-resistance curve, JR =
BOHME AND SCHINDLERON SINGLE-SPECIMENMETHODS 329
Ju(Aa), as described in more detail in [1, 2]. Because of the evaluation of experiments in the
transition temperature range, this method can be called a ductile-brittle-transition(DBT)-
method with a resulting "cleavage R-curve.' . . . .
Reference - JR-Curve
3000
! i ]
2500
9
~ 2000
.z
15oo
~ * fatigued
1000
~= spark-eroded
500 + ~ ,L V-notch
i ] - - referince curve
0 1 2 3 4 5
Crack Extension &a [ mm ]
Figure 1 - "Cleavage R-Curve" data of the previous DVM round-robin [1]

and exponential fit as reference curve
For all tests in the brittle-to-ductile transition regime, the evaluated Jo-data at the
onset of cleavage fracture follow a unique crack resistance curve as documented by Fig.
1. Within the observed bandwidth of scatter, the obtained multi-specimen JR-CUrve is
practically independent of the different notch shapes. Despite usual size requirements
not being satisfied, this cleavage JR-curve can be used for an improved comparative char-
acterization of ferritic steels as discussed in [1]. An exponential curve J = C Aap fitted
through the cleavage data points (see Fig. 1) serves as a reference for comparison with
results of single-specimen methods.
Single-Specimen Methods
Key-Curve Method
A well-known single-specimen method is the so-called key-curve method accord-

ing to Ernst et al. [5, 6]. It is based on the fact, that the shape of the force-displacement
curve is directly related to the crack length, i.e., to the ductile crack extension. Therefore,
the evaluation of force-displacement curves, F(s), allows calculations of the crack exten-
sion and finally the crack resistance in terms of a JR-CUrve:
J(Aa(s)) = rl(a~
B.b
SF(s) .ds " (2)
0
where Aa is obtained by integrating the following equation:
da b ( 1 dF / F')
(3)
dsp, - 2 "[ N-sp, ~ J
with b: ligament width

N: hardening exponent of the Ramberg-Osgood-material law
Spl: plastic component of displacement s
dF / F : normalized slope of the load-displacement curve
dspt
Dahl et al. [7, 8] further developed this procedure and applied it to instrumented
Charpy tests. They used spline functions to smooth the force-displacement curves. Fi-
nally the resulting JR-data were fitted by exponential functions.
Ott and BShme [9] proposed to fit the force-displacement signals at first by polyno-
mial functions of the fourth or fifth order, which consequently enables the derivation of
analytical solutions.
Three-Parameter Analysis
Schindler [10] derived a single-specimen evaluation method based on a three-

parameter characterization of the JR-Curve (C, p, crack tip opening angle) that just re-
quires the totally absorbed energy Wt and the plastic energy at the maximum of the force
displacement curve as experimental input data. In the upper shelf the latter can be ap-
proximated by the energy at maximum load, Win. Precise definitions of Wt and Wm can
be found in the draft standard ISO/DIS 14553. Therewith the JR-Curve estimation is
given by:
J(z~) =/--~[
v-fo'~ rl(ao) .IV, v .Wml-p .Aa p for Aa < Aam (4a)
~,p) B (w- ao)'+~
2"r/'(W,--Wm) [
g(Aa)=g(Aa.)~ -~[~o_--A--~)2 9 ( A a - A a . )
(Aa--~ oAa,,)7~] f~ (4b)
with Aam Win"p" b~ 9

( w__~_.
3 . 1+
]-' ; bo=W-ao. (4c)
= 2W, ' P = 4 ~, W,)
BOHME AND SCHINDLER ON SINGLE-SPECIMEN METHODS 331
These three approaches have been applied by the corresponding participants and
the results are given in Fig. 2 in comparison with the reference curve. In addition, the
results of the low-blow tests performed by Blumenauer et al. [11] are given. Considering
the scatter of the data, the agreement of the different approaches with the experimental
results is reasonable. This accuracy is often sufficient for engineering purposes. How-
ever, for small crack extensions these methods so far may not be accurate enough for the
determination of initiation values.
Single Specimen Methods
3000
2 00
2ooo
9~ 1500 _ ~
D')
-- reference curve
"~ 1000 .~ ..-_,*., -e- IEHK

~ / ~ -A- IWM
-~- EMPA
500 ~ ] 9 Low-Blow
J
r
0 I I
0 1 2 3 4 5 6
Crack Extension da [ram]
Figure 2 - Results of single-specimen key-curveevaluations and low-blowtests
in comparisonto the reference curve
Estimation of the J. -curve from the Charpy Upper Shelf Energy, USE
Several empirical and semi-empirical correlations are available to estimate the

fracture toughness from the Charpy upper shelf energy, USE, but only few of them en-
able the complete JR-Curve to be estimated. Two of these approaches were applied during
this exercise.
At the BAM/Berlin [12] a semi-empirical formula was developed to estimate the
JR-Curve from the upper shelf energy, USE, of a series of standardized Charpy tests. This
procedure was applied by Wobst and his colleagues to one of their round-robin experi-
ments:
J(Aa)=[J,so-vL=0 I'03"(0"4"USE-C1)" ~ + X [ - ~ o (5)

+ B4ao(W_ao ) " n ~ - q r % o
with: C~=5Joule;[Jlso_v]a,=o=91N/mm.
Another estimation formula for the same purpose was obtained by Schindler [13,
14] from the analytical approach described previously (see eqs. (4)) by using the condi-
tion that at maximum force dF/da = 0. This lead to the relation Wm = n.Wt, whith n being
the hardening exponent in the material law c = Ae n. The exponent can be approximated
by the uniform fracture strain determined on a uniaxial tensile test, Ag. Based on experi-
mental results, the exponent p according to (4c) is simplified to be p = 2/3, independent
of the relation Wm/Wt. With these modifications and assumptions the equations (4a) to
(4c) result in:
J(Aa) =
3
_ ao)5/3 "USE. Ag 1/3
.Aa2/3 for Aa _<Aam (6a)
(rv
= J ( A a , , ) ~ 2.92. (1-- A_._~g).USE F(Aa- (,Xa )~ q

J(Aa) - [ for ha > Aam (6b)
B" b02 L Aa,,) 2bo J
with Aa= = Ag 9b0 ; bo = W-ao = 8 mm

3
The results are presented in Fig. 3 in comparison to the reference curve. Again the
overall agreement of the results of both approaches to the experimental multi-specimen
data is reasonably good, and might be sufficient for engineering purposes.
Charpy - Energy Correlations

3000
l j
f7
2500 ....... i
~ 2000
.Z.
~!5oo
1000
500
/~ ac - - reference curve
--n- BAM
- e - EMPA
0 1 2 3 4 5 6
Crack Extension da [mm]
Figure 3 - Results o f semi-empirical calculations based on the USE
in comparison to the reference curve
B(3HME AND SCHINDLER ON SINGLE-SPECIMEN METHODS 333
Preliminary Results of a Current "Blind" Round-Robin
Though the results are encouraging, the previous comparisons are not appropriate
to qualify the corresponding methods, since most of them were developed to evaluate
fracture toughness tests rather than Charpy tests, where the size requirements were in
general heavily violated and the initial notch sharpness is not sufficient. In order to fur-
ther evaluate the capability of the applied single-specimen methods, an additional round-
robin exercise was performed within the DVM group. The aim was to predict the JR-
curve from just one precracked specimen. It was one essential idea of this exercise that
no participant should know the material under investigation and especially the JR-CUrves
should be unknown. Two materials - a pressure vessel steel of the type A 533 B and a
structural steel Fe 510 - w e r e elected as test materials. Precracked Charpy specimens
were prepared - the A 533 B ones with side-grooves - and distributed by EMPA to six
participants, one specimen for one material each. In addition, ten specimens of each ma-
terial were sent to Otto v. Guericke University, Magdeburg, where the dynamic JR-data
were determined independently by the low-blow technique by Blumenauer and cowork-
ers [15]. The materials were not specified at that time and EMPA provided just the tech-
nical data for yield strength; UTS-point and uniform fracture strain as given in Table 1.
Table 1-Tensilepropertiesof~etest ma~r~b
Steel Rp [N/mm2] Rm [N/mm2] Ag [%]

A533 B 470 640 10.6
Fe 510 C 325 540 15.9
Since the evaluation of this round-robin is not yet finished we only show here the
results of the authors in comparison with the low-blow data as provided by Blumenauer
(Figs. 4 and 5). The curve "IWM" was obtained by the corresponding key-curve proce-
dure as outlined previously (see [9]), the curve "EMPA' was calculated by eqs. (4a) and
(4b). Additionally, the blunting line according to the draft standard ISO/DIS 12135 is
shown. It is in good agreement with the first few low-blow points and with the initial part
of the IWM curve. These results indicate that the corresponding evaluation methods are
capable of giving good JR-approximations even if there is literally just one specimen
available. The full evaluation report of this single-specimen round-robin exercise will be
published later.
Discussion
Several single-specimen evaluation procedures are described and have been applied
during different round-robin exercises of the German DVM task group on "Instrumented
Impact Testing." At first, Charpy-V tests with a relative notch/crack-depth ofao/W = 0.2
were analyzed. The results of the different evaluation procedures were in surprisingly
good agreement with a cleavage R-Curve and data of low-blow tests with precracked
800[
RPV-Steel A533 B
700 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
600
. . . . . . . Blunting-Linie
400 . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Low-Blow
300 .... EMPA
IWM
200
100
0
0 0,2 0,4 0,6 0,8 1
da [mm]
Figure 4 - Results of low-blow tests and key-curve evaluations of impactedprecracked

and sidegrooved Charpy specimens
Structural Steel Fe510

'00
700 I ~"
600 -~ / ~ -
300 ----'m/" . . . . EMPA

200 IWM
100
0
0 0,2 0,4 0,6 0,8 1
da[mm]
Figure 5 - Results of low-blow tests and key-curve evaluations of impactedprecracked

non-sidegrooved Charpy specimens
BOHME AND SCHINDLER ON SINGLE-SPECIMEN METHODS 335
Charpy specimens (ao/W = 0.5). However, the observed behavior is probably very special
for materials with a similar high ductility and toughness, where minimum size require-
ments of common standards can not be satisfied by Charpy-sized specimens. Further-
more, some input parameters like the strain hardening exponent were chosen by individ-
ual best estimate. Thus, the comparisons reported in the first part of the present paper are
not well suited to qualify or disqualify any of the applied methods. Finally, at small crack
extensions the accuracy of the results was not sufficient to determine reliable crack ini-
tiation values.
Therefore, another round-robin concentrated on precracked Charpy speci-
mens with a relative crack length aJW = 0.5 and preliminary results oflWM and EMPA
are reported. For the non-sidegrooved FE 510 specimens there are still some differences
(Fig. 5), which might be due to a loss of constraint and 3D-effects as documented e.g. by
crack front curvature. However, for sidegrooved A 533 B specimens both evaluations
agree very well and are confirmed by the results of the low-blow tests (Fig. 4). These
results demonstrate the applicability of single-specimen methods and an improved accu-
racy at small crack extensions.
Conclusion
The single-specimen methods under investigation appeared to be useful tools to

estimate the dynamic crack resistance curve from a single instrumented Charpy test. The
transferability of the results to real structures has tobe handled with care because of the
limited size of the specimen. However, single-specimen evaluation methods enable an
improved comparison and selection of materials and can be used even if only a limited
amount of material is available,
References
[ 1] B6hme, W., "Experience with Instrumented Charpy-Tests obtained by a DVM

Round-Robin and further Developments," European Structural Integrity Society,
ESIS, in: "Evaluating Material Properties by Dynamic Testing," ESIS 20, Ed.: E.
van Walle, Mech, Engng. Publ. (MEP), London, 1996, pp. 1-23.
[2] Schindler, H. J., B6hme, W., Stark-Seuken, D., Blumenauer, H., and Wobst, K.,
Erprobung von Einprobenmethoden zur bruchmechanischen Auswertung instmmen-
tierter Kerbschlagversuche, DVM-Arbeitskreis Bruchvorg~inge, 27./28. Feb. 1996,
Bremen, Tagungsband, DVM, Berlin, 1996, S. 67-76.
[3] B6hme, W. and Schindler, H. J., Single-Specimen Methods to Evaluate JR-Curves from
Instrumented Charpy Tests: Results of a Round-Robin, Proceedings of the European
Conf. on Fracture, ECF 11, Poitiers, France, September 3-6, 1996, Ed.: J. Petit,
EMAS, West Midlands, UK, pp. 1977-1982.
[4] Helms, R., et al., in: Werkstoffprtifung 1986, Bad Nauheim, Deutscher Verband flir
Materialforschung und -prtifung, DVM, Berlin, 1986.
[5] Ernst, H. A., Paris, P. C., Rossow, M., Hutchinson, J. W., Analysis of
Load-Displacement Relationship to Determine J-R Curve and Tearing Instability
Material Properties, ASTMSTP 677, 1979, pp. 581 ff.

[6] Ernst, H. A., Paris, P. C., Landes, J. D., Estimations on J-Integral and Tearing
Modulus T from a Single-Specimen Test Record, ASTMSTP 743, 1981, pp. 476
ff.
[7] Briininghaus, K., Twickler, R., Heuser, A., Memhard, D., Dahl, W., Application of
the Key-Curve Method on Static and Dynamic JR-CUrve Determination with CT-
and SENB-Specimens, 6th Europ. Conf. Fracture (ECF 6), Amsterdam, 1986, pp.
429 ft.
[8] B~ninghaus, K., Falk, J., Twickler, M., Dahl, W., Determination of Crack Resis-
tance Curves Under Static and Dynamic Loading by Analysis of Load Displace-
ment Relationship, Engineering Fracture Mechanics, Vol. 34, No. 4, 1989, pp.
989 ff.
[9] Ott, R. and B~hme, W., Anwendung der Key-Curve Methode zur bruchmecha-
nischen Analyse von instrumentierten Schlagbiegeversuchen, Fraunhofer-IWM
Report Z 18/92, Freiburg, December 1992.
[ 10] Schindler, H. J., Estimation of the dynamic J-R-curve from a single impact bending
test, Proceedings of 1 lth European Conf. on Fracture (ECF11), Poitiers, 1996,
EMAS, pp. 2007-2012.
[ 1l] Blumenauer, H. and Eichler, B., Results of low-blow tests, a contribution to the first
round-robin on single-specimen methods of the DVM group on instrumented im-
pact testing, private communication, 1996.
[12] Aurich, D., et al., Analyse und Weiterentwicklung brnchmechanischer Ver-
sagenskonzepte, BAM-Forschungsbericht 174, Bundesanstalt ~ r Material-
forschung und Prtifung, Berlin, 1990, pp. 128 ff.
[13] Schindler, H. J., The correlation between Charpy fracture energy and fracture tough-
ness from a theoretical point of view, Proceedings of 12th European Conf. on
Fracture (ECF12), Sheffield, 1998, EMAS, pp. 2007-2012.
[ 14] Schindler, H. J., Estimation of fracture toughness from Charpy tests - basic theoreti-
cal relations, Pendulum Impact Testing: A Century of Progress, ASTM STP 1380,
T. Siewert and M.P. Manahan, Sr. Eds., American Society for Testing and Mate-
rials, West Conshohocken, PA, 1999.
[ 15] Blumenauer, H. and Eichler, B., Results of low-blow tests, a contribution to the sec-
ond round-robin on single-specimen methods of the DVM group on instrumented
impact testing, private communication, Aug. 1998.
Hans-Jakob Schindler 1
Relation Between Fracture Toughness and Charpy Fracture Energy: An

Analytical Approach
Reference: Schindler, H.-J., "Relation Between Fracture Toughness and Charpy

Fracture Energy: An Analytical Approach," Pendulum Impact Testing: A Century
of Progress, STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American Society
Abstract: Fracture toughness is often estimated from the Charpy fracture energy by
empirical correlation formulas, although the latter are known to be in general of poor
accuracy and valid only with restrictions. In the present paper the relation between these
two types of toughness parameters is considered from a theoretical point of view.
Unexpectedly a mathematical relation was found to exist. As known from experimental
data, this relation is not unique, but dependent on some material properties like tensile
strength and uniform fracture strain. The derived relation can be used in the upper-shelf
and upper ductile-to-brittle transition range. It is in good agreement with experimental
data and the current empirical correlation formulas, giving theoretical support to the
latter and confirming the present model. In the lower transition range, the use of the
ASME lower-bound curve is suggested. The temperature shift due to loading velocity
and notch sharpness requires additional experimental work.
Keywords: impact testing, fracture energy, Charpy-V-notch test, J-resistance curve,

fracture toughness, correlation, analytical relation
Nomenclature
0 relative rotational angle between crack faces

a crack length
A nondimensional constant
a0 initial crack length or notch depth, respectively
A~ uniform fracture strain
b ligament, b = W-a
B specimen thickness
b0 initial ligament width, W-a0
C nondimensional constant
1 Senior research engineer, Swiss Federal Labs. for Materials Testing and Research
(EMPA), CH-8600 Duebendorf, Switzerland.
337
CTOA crack-tip opening angle

crack-tip opening displacement
Aa stable tearing crack extension
Aam crack extension at maximum load
Fm maximum load
J J integral
J0 near-initiation J as defined in the present paper, Fig. 3
J0.2 J integral at Aa = 0.2 mm
J0.2t near-initiation J as defined in the present paper, Fig. 3
Jlc critical J integral
Jit conservative near-initiation J (either J0, J0.2, or J0.Et)
Jm J integral at maximum load
Jmp plastic part of Jm
Klc SIF at initiation of ductile tearing
Kjo SIF calculated from J0.2t obtained from the KV-correlation
KV Charpy fracture energy
m constraint factor
M bending moment
n hardening exponent
P exponent
Rm ultimate tensile strength
Rp yield stress
(~fd flow stress o'f at increased strain rate (dynamic flow stress)
SIF stress intensity factor
T temperature
W specimen width
Wm energy consumed by the test specimen up to maximum load
Wmp plastic (nonrecoverable) part of Wm
Wp plastic part of the energy consumed by the specimen
Wt total fracture energy
Wtmp fracture energy consumed beyond maximum force, Wp - Wrap
z nondimensional distance of the center of rotation from the crack tip
In practical applications of fracture mechanics, fracture toughness in terms of KIr or

Jle of the considered material is required but often not available. If it cannot be evaluated
experimentally, maybe because of a lack of testing material, time or testing budget, a
well-known possibility is to estimate it from correlations with the Charpy fracture
energy, KV. For this purpose a number of empirical relations are offered in the literature
[1-4]. However, lacking a sound theoretical basis, they are known to be of limited
accuracy and to hold only with restrictions to certain types of fracture behavior (e.g.
upper-shelf) and to the family of materials for which the correlation has been established
experimentally. Correspondingly, the estimated fracture toughness values should be used
in the analysis with appropriate caution.
SCHINDLER ON FRACTURE TOUGHNESS 339
In the present paper the physical connection between Charpy fracture energy KV and
fracture toughness JIc is considered from a theoretical point of view. At first sight, the
existence of an analytical relation between these two different types of toughness
properties seems to be unlikely, because the involved physical processes are rather
different. In KV, there are contributions from various physical processes, like crack tip
blunting, ductile crack initiation, and, mainly, tearing crack growth, shear-lip formation
and plastic deformation of the ligament, whereas JIc or Kic characterize just the resistance
to initiation of crack growth. Nevertheless, for reasons discussed below, some theoretical
relations between these parameters can still be established by using an appropriate model
of ductile fracture.
In the analysis, we consider first of all the upper-shelf range, where fracture
toughness has to be evaluated from the J-resistance (J-R) curve, as prescribed in the
current standards, such as ASTM Standard Test Method for J~c (E 813), Standard Test
Method for Measurement of Fracture Toughness (E 1820) or the draft international
Standard Unified Method of Test for the Determination of Quasistatic Fracture
Toughness (ISO/DIS 12135). The determination of the J-R curve requires recording the
force-displacement diagram and monitoring the corresponding crack prolongation. The
latter is in general not possible in the case of an impact test such as an instrumented
Charpy test. Therefore, in [5] and [6] an approximate evaluation procedure is proposed,
which allows the dynamic J-R curve to be estimated from essentially two key parameters
of the force-displacement curve, which are the energy at maximum load and the total
fracture energy. The corresponding derivation, which is based on a three-parameter
characterization of tearing crack growth, is summarized below. However, in the case of a
classical Charpy test, the only available parameter is the total fracture energy, KV. As
shown below, by introducing an additional mathematical condition for maximum load, it
is possible to eliminate the energy at maximum load from the equations, leaving the total
fracture energy as the only relevant parameter in the equations. Therewith an analytical
relation between fracture toughness and KV is obtained, which can serve as a correlation
formula to estimate Jlc or Kic from KV.
Since Charpy specimens often do not fulfill the requirements concerning size,
loading rate, crack depth and notch sharpness as given by the above mentioned testing
standards, it is necessary to account for these effects when using the estimated data in a
fracture mechanics analysis. Generally, as discussed below, these deviations result in a
temperature shift of the KIe vs. temperature curve.
Estimation of J-R Curves from a Single Force-Displacement Diagram
Before turning to the problem of estimating Jle from the Charpy fracture energy, we
consider the case of estimating the J-R curve from a force displacement diagram of a
quasi-static bending test of an edge-cracked specimen (Fig. 1). A well-known evaluation
procedure for this purpose is the key-curve method [7]. However, for an impact bending
test, where the force-displacement diagram is in general affected by significant
oscillations (Fig. 2), the accuracy of this method in the initial phase of tearing fracture is
limited. Therefore the simplified approach as proposed in [5, 6] can be advantageous.
The corresponding mathematical derivation is shown in the appendix. Briefly, the idea of
this simplified analytical approach is the following. In the range F<Fm, or Aa<Aarn (with
Fm denoting the maximum force and Aarn the corresponding crack extension), the crack
extension Aa is in general J-controlled. The J-resistance curve (J-R curve) in this phase is
assumed to be
J(Aa) = C. Aap for Aa<Aam (1)
where C and p are material-dependent constants. For the subsequent tearing crack
growth in the range Aa>Aam it is assumed that crack growth is governed by a constant
crack tip opening angle CTOA. As shown in the appendix, the latter can be obtained
from the energy consumed in the tearing phase between maximum force and complete
fracture. CTOA is mathematically related to the slope of the J-R curve in the first part of
the tearing process. From this condition one can find the unknowns C, p and Aa as given
by eqs. A.17 - A.19 of the Appendix: Taking into account that for small specimens in
the upper-shelf range the elastic part of the energy at maximum load is small compared
to the non-recoverable part required in eq. A. 16, C and p simplify to
C=(2) p r/(a~ W, e .W~p'-p (2)

a (w - ~o)'+'
p=-~. 1 W, ) (3)
The exponent p as given by eq. 3 represents a minor modification of eq. A. 19 based on

experimental data, since eq. A.19 turned out to lead to overly conservative J-R curves
[5].
Fig. 1 - Mechanical system (a) and the corresponding force-displacemant diagram of a

bending test with an edge-cracked specimen in the upper-shelf range (b).
As indicated in Fig. l(b), Wm and the total fracture energy Wt can be obtained from
the load-displacement diagram delivered from an instrumented Charpy pendulum
hammer. (Fig. 2). 1"1 is the well known rl-factor, which is about 2 for deep cracks and,
according to [8],
q=13.81.W-25.12.(W ~ for 0<a/W<0.275 (4)
Note that the J-R curve according to eqs. 1-3 is determined by only two
experimental parameters, Wm and Wt, which both are well defined and can be
determined from a load-displacement diagram with good accuracy and reproducibility
[9], leading to J-R curves that are unambiguous, i.e., essentially free from engineering
judgements. The agreement of the J-R curve estimated by eq. 1 - 3 with the experimental
one evaluated by multi-specimen techniques is sufficient for most practical applications
[5, 10].
8.0
I .i 9J ArangeI [ range
6.0 r ~,~ Jm
J0 2t
u. 4.0
,,o ,~ Jo,2
2.0 '~-~
0.(3 I
0,0 1.0 2.0 3.0 4,0 5,0 6.0 7.0 8.0 9,0 10.
0
Displacements (rnm) 0.2ram ~ar. Aa
Fig. 2 - Force-deflection diagram of an Fig. 3 - J-R curve and definition of near-

instrumented Charpy test initiation values Jo.2Bt and Jdi
Estimation of J-R Curve From Upper-Shelf Fracture Energy

In the case of a non-instrumented impact test like the classical Charpy test, the only
known experimental parameter is the total fracture energy Wt. To get rid of Wm in eq. 2
and eq. 3, an additional mathematical equation is required. For this purpose we use the
condition that at maximum load is
dF
dzla( Aa= Aa m)=O (5)
According to [11], the general functional relationship between J and F is
Leo ) (6)
where n is introduced as the hardening exponent according to the material law

o=A.e" (7)
The reference force F0 is suitably assumed to be the plastic limit load, which depends on
the actual ligament width b = b0-Aa by
Fo~ - d a ) 2, (8)
(see eq. A.6), With eqs. 1, 6 and 8 one obtains from eq. 5 in a first order approximation
(since n << 1 and p < 1):
n, p,b o
Aa,, - 2 (9)
By comparing eq. 9 with eq, A.18 one finds
Wm --W.~ = n -W, (10)
With eq. 10, Wm can be eliminated from eqs. 2 and 3, leaving the total fracture energy
Wt to be the only experimental parameter therein. The hardening exponent n is known to
be approximately equal to the uniform fracture strain Ag, i.e., the plastic strain at
maximum load of a uniaxial tensile test. Strictly speaking n is equal to the tree
(logarithmic) strain, but regarding the approximate nature of the present derivation, we
neglect, for the sake of simplicity, the minor difference between engineering and true
strain, thus
n=Ag. (ii)
With eqs. 10 and 11, one obtains from eq. 1 - 3
'7(ao) Wt "Ag I-p 9AaP (12)

B ( W ' a o ) 1+p
where p=3,O+ag)-I (13)
This equation enables one tO estimate the J-R curve just from the total fracture
energy and the uniform fracture strain. In the case of standard Charpy tests, the following
values have to be inserted: Wt = KV, B = W = 10 mm, a0 = 2 mm and, according to eq.
4, r I = 1.76. Since the exponent p according to eq. 13 is not very sensitive with respect to
Ag, eq. 13 can be simplified by taking p = 2/3. Therewith one obtains a single equation,
J(Aa)=ll.44.KV.Ag~.Aa ~3 (with J in N/mm, KV in J) (14)
Eq, 14 enables the J,R curve to be estimated just from the upper, shelf Charpy
fracture energy KV, and the uniform fracture strain of the uniaxial tensile tesL Ag, The
effect of the finite notch root radius is discussed below.
Determination of Fracture Toughness
The common practice in fracture mechanics is to determine technical near-initiation

J-values such as Jle or J0.2al from the J-R-curve, as defined in the standards for quasistatic
testing, e.g. ASTM E 1820 or ISO/DIS 12135. Analogously, the estimated dynamic J-R
curve can be used as a basis to determine corresponding dynamic fracture toughness
properties. However, due to a lack of constraints, J-R- curves evaluated from small
specimens like Charpy specimens are known to rise too steep after crack initiation, so
they tend to deliver too high values of J0.2n31 [5, 12]. For this reason, three more
conservative alternatives, J0, J0,2 and J0.2t as defined in Fig. 3, are suggested in [5, 12].
The first one is obtained by the intersection of the linear extrapolation of the J-R curve
:from the range Aa > AamtO the intersection with the blunting line, the second is the value
of J at Aa = 0,2 m m and the third is its value at a distance 0.2 mm from the intersection
with the blunting line. According to the draft international Standard ISO/DIS 12135 the
blunting line is given by
J = Sl.Aa = 3.75~Rm.Aa (15)
(where Rm denotes the ultimate tensile strength) which seems to be conservative even for
Charpy-type specimens [12]. Working out Jo, J0.2 and J0.2t as defined in Fig. 3 for the J-R
curve described by eq. 14 leads to
7;33. Ag . ~
Jo (16)
1-1.47. KV
Rm
J0.2 = 3i92. Ag~. KV (17)
J~ Ag~ KV "[( 3.05.KV

Rm ) 3 . Ag +0.2]~ :(18)
where KV has to be inserted in N.m (Joule), and the resulting J is delivered in N/mm.
Which one of these three candidates (eqs. 16 - 18) is preferable :to estimate fracture
.toughness depends on the material and the required degree of conservatism. ~ used as a
fracture toughness value in a quantitative failure assessment analysis, it is advisable for
the sake of conservatism to consider the smallest value of the three candidates, thus
Jta = Min(Jo, Jo.2,Jo.2t) (19)
Additionally, to be used as a toughness parameter in a fracture analyis o f a structural

part, JId as given by eq. 19 should fulfill the size requirement according to the standards
ASTM E 813 or ASTM E 1820, thus,
J~d-<afd.bo/25) (20)
In eq. 20 the flow stress try as defined in the above mentioned standards is replaced by
the dynamic one, trfd 9 In general, the latter is not available, so we suggest that the static
tensile strength be used as an approximation, i.e., trfa --- Rm in eq. 20.
From eq. 19 a representative (conservative) fracture toughness value Kjd can be

derived by the well-known relation
K ja = (Jla .El(l- l~ ) ) 1/2 (21)
Comparison with Experimental Data
In Fig. 4 the comparison of eqs. 16 - 18 with the empirical correlation formulas

suggested in [2 - 4] is shown for the ultimate tensile strength and the uniform fracture
strain of a typical structural steel (chosen properties Rm = 650 MPa and Ag = 0.15). The
agreement is remarkably good. The formula by Norris et al. [3] describes the same
upwards-bent trend as eqs. 16 and 18, whereas eq. 17 follows the simple linear behavior
of the other two empirical formulas. Since there is experimental justification for all the
empirical formulas, the question whether eq. 16, 17 or 18 is preferable depends on the
material and the desired degree of conservatism. Figure 5 shows the comparison with
data from the literature and the author's own experimental data for structural steels with
Rm<700 MPa. It reveals that eq. 16 represents roughly a lower-bound of the correlation,
(i.e., rather conservative estimates), where eq. 18 an upper bound. Fig. 6 shows a
comparison between fracture toughness data of relatively high strength steels, with
tensile strengths in the range of 700 MPa < Rp < 1700 MPa, and the values predicted by
eqs. 16 - 18. For these materials, eq. 17 and 18 seem to give quite realistic predictions,
whereas eq. 16 is rather conservative.
600
- - - '- JO.2
JO
..x4 Z
500
. . . . JO.2t , --," ,Z x,
X*
A Ref. [2] B d~X,A
400
X Ref. [3]
x aef. [4]
300
'~ 200
100
0
0 50 100 150 200 250
Charpy fracture energy [J]
Fig. 4 - Comparison o f eqs. 16 - 18 (for Rm = 650 M P a and Ag = 15%) with the

empirical correlation formulas given in [2 - 4].
SCHtNDLER ON FRACTURE TOUGHNESS 345
Discussion
Comparison with Experimental Data
Overall, the three presented analytical correlation formulas, eqs. 16, t7 and 18,
compare with current empirical correlation formulas (Fig. 4) and experimental data
pretty well, at least better than one could expect from such a relatively simple theoretical
analysis of the involved relatively complex physical processes. Recall that essentially no
empirical adjustments were introduced in the derivation of these formulas, except a
minor one by modifying p as given in eq. A.19 to eq. 3 and finally simplifying it to
p=2/3.
Fig. 5 - Comparison of eqs. 16 - 18 with experimental data for low- and medium strength
structural steels (Re < 700MPa)
400.0
300.0
.R ./"" 9
200.0
~./"" 9
...," 9
100.0.
~." 9 9
0.0
0 100 200 300 400
measured JIc
Fig. 6 - Comparison of estimated and measured fracture toughness values from [1]for
high strength steels (Rp > 760MPa)
Definition of Near-Initiation Fracture Toughness
The J-R curve suggested in eq. 14 is unique. Nevertheless, lacking a well-defined

criterion for ductile crack initiation in small specimens, a choice of three candidate
correlation functions with respect to initiation toughness is given above for near-
initiation toughness values (eqs. 16 - 18). In general, for the sake of conservatism, it
surely is advisable to choose the one that delivers the lowest Kjc. In most cases this
seems to be eq. 16. If realistic rather than conservative predictions are desired, then eq.
16 performs best for low and medium strength steels, whereas for higher strength steels
(Rp > 760 MPa) eq. 18 seems to be preferable. Preliminary results indicate, that eq. 18
may also be preferable for aluminum.
According to eqs. 16 - 18 the material properties Rm, Ag and E affect the relation
between KV and fracture toughness significantly. These effects are illustrated in Fig.. 7,
where eqs. 16 and 18 are shown for some typical values for Rm and Ag. Based on the
comparisons discussed in the previous section, the lower bounds are assumed to be given
by eq. 16, the upper ones by eq. 18. Thus, the areas between these two curves shown in
Fig. 7 for some families of materials represent the estimated bandwidth of correlation
formulas for these materials.
Fig. 7 - Estimated bandwidths of correlation for three classes of materials (assumed

properties for the considered materials: Ag = 0.4, 0.1, 0.05; Rm = 500, 800, 300 MPa;
E = 200,000, 210,000, 70,O00MPa, respectively)
Physical Reasons for Analytical Relations
Actually, as discussed in the introduction, there is no direct physical relation

between Charpy fracture energy and fracture toughness. Nevertheless, the analytical
equations derived above, and their good agreement with experimental data imply that
there still are some physical relations between the two parameters. The main ones appear
to be the following: In small specimens such as Charpy specimens, tearing crack growth
is initiated in general not before full plastic yielding occurs. Accordingly, the main
contribution to J results from the energy that is associated with plastic bending of the
ligament. The latter also controls the subsequent stable crack growth. Thus, a major part
of J at crack initiation has the same physical origin as J at larger amounts of Aa. By
assuming that the general law according to eq. 1 covers the initiation phase as well as the
subsequent tearing phase of the J-R curve, an extrapolation from the tearing behavior to
crack initiation is possible. Thus, the obtained relations between fracture toughness and
Charpy fracture energy are partly rooted in the physical behavior of ductile fracture, but
also in the used model of tearing crack extension.
Estimation of Fracture Toughness in Brittle-to-Ductile Transition Range
The J-R curve described by eqs. 1 - 3 is physically meaningful only in the upper-
shelf range. In the upper brittle-to-ductile transition (BDT) range, where a cleavage
fracture is initiated after some amount of.stable tearing, it is evident that the initial J-R
curve stays essentially the same as in the upper-shelf range. Nevertheless, the total
fracture energy KV in the upper BDT is smaller than the one in the upper-shelf range.
Thus, eqs. 19 and 20 obviously deliver conservative fracture toughness values in the
upper BDT range. For structural steels, this practically means that eqs. 19 and 20 can be
used as conservative estimates in the BDT range for about KV>30J (or
Kje>100MPa.m 1/2, respectively).
In the lower BDT range (for structural steels at about KV<30J), where cleavage
occurs without prior stable crack extension (Aa<0.2 mm), it can be shown that JId
according to eq. 19 is smaller than the actual J integral at instability according to eq. A1.
Thus, eqs. 19 and 20 could be in principle applied even in the lower transition range as
conservative estimates, although the theoretical basis is weak in this case. However, in
this toughness range it is preferable to use the so-called master curve according to [13] to
estimate Kjd(T). As shown in [14], the latter is essetltially equivalent with the ASME
lower-bound Kit-curve [ 15] which is widely used in nuclear engineering.
Effect of Finite Notch Root Radius
It is well known that the initiation fracture toughness depends on the notch root
radius. Therefore it is kind of suspicious that the notch root radius does not appear in the
theoretical correlation function. However, one has to keep in mind that that the derived
equations strictly hold only for the upper-shelf range. In this range, the energy consumed
until crack initiation is only very small compared with the energy required for ductile
tearing and the associated plastic bending of the ligament. Therefore the fracture energy
and the energy up to maximum load are not much affected by the notch or crack
sharpness. As the fracture energy is reduced with decreasing temperature, the effect of
the notch increases. However, as discussed in a previous section, the energy decreases in
the ductile-to-brittle transition range, so eqs. 16 - 18 become more and more
conservative, tending to compensate the increasing effect of the notch radius. In the
lower transition range, where the effect of the finite notch radius is crucial, the general
lower-bound curve is used as discussed above, which covers sharp cracks as well.
Shift of Brittle-to Ductile Transition Temperature
The main effect of the reduced notch sharpness and constraints is a shift in the
transition temperature. The latter is known to depend on the loading velocity, the notch-
root, the crack-tip constraints and the material. In general, the loading velocity of the
Charpy test and the constraint conditions of the corresponding specimen are rather
different from the ones of a quasistatic Kit-test using a standard CT-specimen. The
fracture toughness curve Kjd(T) estimated from the Charpy fracture energy as suggested
above corresponds to the loading velocity and the constraint conditions of the Charpy
test. Although the effect of an increased loading rate and lower crack-tip constraints tend
to compensate each other, a certain (material-dependent) temperature shift is in general
present[ 16].
Conclusions
This investigation revealed some unexpected theoretical relations between fracture

toughness and Charpy fracture energy. They are in surprisingly good agreement with
theoretical data and some of the well known empirical correlation formulas, giving the
lacking theoretical support to the latter. They also show, that correlation formulas
between KV and fracture toughness are not unique, but dependent on further material
properties, first of all the uniform fracture strain, the tensile strength and Young's
modulus. This agreement between theory and experiment also confirms the simplistic
analytical model that is used in the presented theoretical derivation, supporting the
single-specimen evaluation method described in the Appendix and, in more detail, in [5].
The presented relations are derived essentially without using adjustable open factors
("fudge-factors"). So there still is a potential to improve their accuracy by introducing
suitable open factors, which are to be determined by comparison with experimental data.
In this way, for certain classes or families of materials, improved semi-empirical
correlation formulas between Charpy fracture energy and fracture toughness are likely to
be obtained.
The fracture toughness values Kle or Jie delivered by correlation formulas correspond
to the loading rate and the crack-tip constraints of a Charpy test. Compared with a
standard quasistatic fracture toughness test, the former is increased and the latter are
reduced. Although the corresponding effects on fracture toughness and brittle-to-ductile
transition temperature tend to compensate each other, in general there is a (material-
dependent) shift of the brittle-to-ductile transition temperature. Therefore, when using
fracture toughness values estimated from Charpy tests in a fracture mechanics analysis o f
a structural component, the corresponding temperature shift as well as the restricted
validity criteria corresponding to the small specimen size should be taken into account,
at least qualitatively.
APPENDIX
Derivation of J-R Curve from a Single Force-Displacement Diagram by a 3-

Parameter Approach 1
Neglecting its elastic part,.which is small in cases of small specimens, the J-R curve
of an elastic-plastic three-point bend specimen (Fig. 1) is given by
J ( A a ) = rl . Wp( zia ) (A.1)

B.(bo - aa )
The general form of the assumed J-R curve is shown in Fig. 3: Within range I which is J-
controlled it is usually approximated by
J( Act ) = C. Act p for Aa< Aa m (A.2)
In this range, J is connected to the crack tip opening displacement ~ by
J = o/a .m.tS( Aa ) (A.3)
where ~fd is the local dynamic flow stress defined below (eq. A.6) and m the constraint
factor, which is about 1 for plane stress and about 1.6 for plane strain.
O
Fig. A. 1 - Model and parameters for an edge cracked beam under bending
The J-R curve in the adjacent range II can be expressed as (see eq. A.1 and Fig.
A.1):
O'Wm~ ~.W~(~)
J( da )= B . b o + B . ( - -o- , d a forAa>Aam (A.4)
1This appendix contains only the main equations. For further discussion and references
see [5].
where Wmp is the plastic part of Wm and
W,,,p(Aa) = f~(Aa,,)M(Aa).dO
o(~) = fbboO.-~a~M ( b ) .dO(b) (A.5)
is the e n e r ~ consumed in the subsequent tearing phase. In this phase a plastic hinge is
formed in the ligament, so the bending moment is given by
C
M ( b ) = ~ cryd 9B . b 2 , (A.6)
where c represents a nondimensionai factor that takes the value 1 for plane stress and
about 1.40 for pla~ne strain.
The tearing crack extension for Aa > Aam is assumed to be governed by a constant
crack tip opening angle CTOA, which is defined as
CTOA = ~ (A.7)
For crack growth in bending this leads for kinematical reasons to
CTOA
dO(b)=- z .b db (A.8)
(see Fig. A. i), where z.b denotes the location of the center of rotation. Inserting eqs. A.8
and A.6 into eq. A.5 gives
n:' c
Wtmp = "-~z CTOA . tr/d . [ 2bo " ( A c t - .da m ) - ( ,da - .da m)2 ] (A.9)
By inserting eq. A.9 in eq. A.4, then taking the derivative with respect to Aa and
comparing the resulting expression at Aa = Aam, with the derivative of eq. A.3, which is
dJ dJ 71. c d~
da - d d a - 4Z CTOA'aId = ~ "m'cr/d (A.IO)
one obtains, by using eq. A.7
~'C
Z = 4m (A.I I)
The total energy W t iS obtained by

W t = Wp(Zla=b O) = Wrap + Wtmp(Zia=b0 ) (A. 12)
Therefore one obtaines by eqs. A.9 and A.11

2rl .(W,-W,p)
CTOA = (A.13)
B.m.ala .(bo - Aa. f
With eq. A. 13, the J'R curve according to eq. A,4 results in
J( ~ )= Jmp Jr ~( z~a - z~a ) 1 for Aa > Aam (A.14)

L 2bo J
where
Jmp = ri . W,p 2. rl .(W, - W,p ) (A~15)
B.bo s2= B'(bo- ~ , , f
The three unknowns, Aam, C and p, are d e t e ~ n e d by the following matching conditions
of eqs. A.2 and A.15 at Aa = Aam:
Jl ( Zlam )= J1( Aa, )= J ~ (A.16a)

dJ 1 d J"
dda ( da . )= ~ ( d a , )= s2 (A. 16b)
deJl d2jn S2
dda2 ( Aa m)= ~ ( d a , ) = -~-0 (A. 16c)
Herein, the superscripts I and 11 indicate correspondence to range I (eq. A.2) and II (eq.
A. 15), respectively (see Fig. 3). From the conditions A. 16a- A. 16c one obtains
(ao, (A.17)
W,p . p.b o
~'- 2w, (A.18)
P = l+ W,~ -1
(A. 19)
References
[1] Server, W.L., "Static and Dynamic fibrous initiation toughness results for nine
pressure vessel materials," Elastic-Plastic Fracture, ASTM STP 668, J.D. Landes,
et al., Eds., American Society for Testing and Materials, West Conshohocken,
PA, 1979, pp. 493-514
[2] Norris, D.M., Reaugh, J.E. and Server, W.L., "A fracture toughness correlation
based on Charpy initiation energy," Fracture Mechanics: 13th Conference,
ASTM STP 743, R. Roberts, Ed., American Society for Testing and Materials,
West Conshohocken, PA, 1981, pp. 207-217.
[3] Barsom, J.M. and Rolfe, S.T., Fracture and fatigue Controle in Structures -
Application of Fracture Mechanics, 2nd ed., Prentice-Hall. Englewood Cliffs,
NJ, 1987
[4] Sailors, R.H. and Corten, H.T., "Relationship Between Material Fracture Toughness
Using Fracture Mechanics and Transition Temperature Tests, Proc. of the 1971
National Symposium on Fracture, ASTM STP 514, American Society for Testing
and Materials, West Conshohocken, PA,1972, p. 164-191
[5] Schindler, H.J., "Estimation of the dynamic J-R curve from a single impact bending
test," Proc. 11th European Conf. on Fracture, Poitiers, 1996, EMAS, London,
pp. 2007-2012
[6] Schindler, H.J. "The Use of Instrumented Impact Testing in Engineering Integrity
Assessment," Evaluating Material Properties by Dynamic Testing, ESIS 20,
Ed. E. Van Walle, Mech. Eng. Publ., London, 1996, pp. 45-58
[7] Ernst, H.A., Paris, P.C. and Landes, J.D., "Estimations on J integral and Tearing
Modulus from a Single Specimen Test Record," Fracture Mechanics: 13th
Conference, ASTM STP 743, R. Roberts, Ed., American Society for Testing and
Materials, West Conshohocken, PA, 1981, pp 476 ff.
[8] Nevalainen, M. and Wallin, K., "The effect of crack depth and absolute thickness on
fracture toughness of 3PB specimens," Proc. lOth Europ. Conf. Fracture,
Berlin, EMAS, 1994,pp. 997-1006
[9] B6hme, W., "Experience with Instrumented Charpy test Obtained by a DVM Round
Robin. Evaluating Material Properties by Dynamic Testing," ESIS 20, Ed. E.
Van Walle, Mech. Eng. Publ., London, 1996, pp. 1-44
[10]B6hme, W. and Schindler, H.J. "Application of Single Specimen Methods on

Instrumented Charpy Tests: Results of DVM Round Robin Exercises,"
Pendulum Impact Testing: A Century od Progress, ASTM STP 1380, T. Siewert
and M.P. Manahan, Eds., American Society for Testing and Materials, West
Conshocken, 1999
[ 11] Electric Power Research Institute (EPRI), "An Engineering Approach to Fracture,"
EPRI NP 1931, Palo Alto, July 1981
[12]Schindler, H.J., Bond, P. and Prantl, G., "The Effect of Crack Length on Fracture
Mechanics Properties Estimated from Instrumented Charpy-Type Tests," Proc.
12th European Conf. on Fracture, Sheffield, 1998, pp. 1279 - 1284
[13]Wallin, K, "Recommendation for Application of Fracture Toughness Data for

Structural Integrity Analysis," Proc. CSNI/IAEA Specialists' Meeting, Oak
Ridge, TN, 1992
[14]Nagel, G., Blauel, J.-G. "Evaluation of the Standard Master Curve for Fracture
Toughness Determination," Nuclear Eng. and Design, 190,1999, 159 - 169
[15] American Soc. Mechanical Engineers (ASME), ASME-Code, Sec. XI, A-5300 1995
[16]Barsom, J.M., "Structural Problems in Search of Fracture Mechanics Solutions,"

Fracture Mechanics: 23 ra Symposium, ASTM STP 1189, R. Chona, Ed.,
American Society for Testing and Materials, West Conshohocken, PA, 1993,
pp. 5-34
Hans-W. Viehrig,l Juergen Boehmert,2 Holger Richter,3 and Matti Valo 4
Use of Instrumented Charpy Test for Determination of Crack Initiation Toughness
Reference: Viehrig, H.-W., Boehmert, J., Richter, H., and Valo, M., "Use of
Instrumented Charpy Test for Determination of Crack Initiation Toughness,"
Pendulum Impact Testing: A Century of Progress, STP 1380, T. A. Siewert and M. E
Manahan, Sr., Eds., American Society for Testing and Materials, West Conshohocken,
PA, 2000.
Abstract: The instrumented Charpy test provides the opportunity to determine dynamic
crack initiation (Jia) toughness. Under the condition of the impact bending test, the critical
step for evaluating Jid is the identification of the ductile crack initiation point on the impact
load versus displacement curve. Therefore, acoustic emission (AE) was monitored and
used as a tool for ductile crack initiation detection. With a tup-integrated AE sensor,
different types of AE events could be identified during loading ofa Charpy size specimen.
One type of AE signal appeared in the temporal vicinity of crack initiation and was
applied to define Jid. A verification of these results was performed by simultaneous optical
measurement of COD with an inverted impact pendulum. Additionally "Cleavage R-
Curves" (Jdt~a) were evaluated and physical crack initiation toughness (Jidszw) values were
determined by measurement of the stretch zone width (SZW).
Keywords: dynamic fracture toughness, ductile crack initiation, single specimen method,
J-integral, acoustic emission, instrumented impact testing
Introduction
Dynamic fracture testing is of interest because many structural components are

subjected to high loading rates in normal service or must withstand dynamic loading
during accident or unscheduled conditions. To assess the structural integrity of such
1 Senior Scientist, 2 Head of Department, 3 Research Scientist, Department Material

Behaviour and Component Safety, Institute for Safety Research, Forschungszentrum
Rossendorf e.V., P.O.Box 510119, D-01314 Dresden, Germany
4 Group Manager, VTT Manufacturing Technology, P.O.Box 17042, FIN-02044 VTT,
Finland
354
VIEHRIG ET AL. ON CRACK INITIATION TOUGHNESS 355
components, appropriate dynamic fracture mechanics parameters are needed. Basically, it

is well known that instrumentation of the Charpy impact test for measuring the impact
load allows the determination of these parameters [1].
In the range of lower shelf and the lower ductile-brittle transition, J-integral-based
fracture toughness values (Kjcd) can be determined at the onset of cleavage crack
initiation. The onset of cleavage fracture appears on the measured load time track as a load
drop. These Kjcd data have the drawback to be limited by the maximum Kjc measuring
capacity of the standard Charpy size specimen.
The approach is not applicable for materials that exhibit large amounts of stable crack
growth to failure. In this case, one of the commonly used methods is based on the
determination of the crack resistance curve R curve) by means of multiple specimen
technique. This technique is based on loading of specimens up to different pre-defined
displacement levels and measuring the resulting crack extension. Under the conditions of
impact testing the displacement can be restricted by "low blow"and "stop block"
techniques. For different displacements the J integral is calculated according to standard
procedures and the R curve can be constructed using the achieved crack extension. The J
integral at the point of ductile crack initiation (Ji) is considered to be independent of
specimen type and size and, thus, is a material parameter to characterize the fracture
toughness for elastic-plastic material behaviour [2]. It can be determined on the R curve
by measuring additional microfractographic parameters namely the stretch zone width
(SZW). Within the transition region the so-called ductile-to-brittle-transition (DBT)
method [3-5] can also be applied using the phenomenon that small variations of the
temperature significantly change the amount of ductile crack growth. The onset of
cleavage fracture is marked on the load time track. The load at cleavage initiation and the
achieved crack extension measured after the test define a point on the resulting "cleavage
R curve". Nevertheless, in every case the techniques need many tests and are, thus, time
consuming, expensive and inconvenient to use.
To avoid these drawbacks, Jid should be determined with methods based on single
specimen test technique. The key point is the detection of ductile crack initiation in the
load versus displacement curve by impact testing. Unfortunately, there are many
suggestions but no satisfactory solution for this. Apart from methods, which recognise
crack initiation from the sophisticated analysis as the load-displacement curve proceeds,
most of them measure physical signals or properties which are related to crack initiation
or crack growth.
This paper presents experiments to determine Jid with a single specimen method using
the emission of acoustic waves (AE). For validation, experiments are performed at a
modified measuring system based on the inverted impact pendulum ofVTT (VTT-Tester).
It allows the simultaneous measurement of AE, impact load and the crack opening
displacement (COD).
Detection of Ductile Crack Initiation During Impact Loading
From the list [3, 4] of methods proposed to detect the crack initiation point, there are
three methods which have been applied successfully to the instrumented Charpy impact
test till now:
crack opening displacement (COD) extensometer [1],
acoustic emission (AE) [6-8] and
magnetic emission (ME) [9, 10].
The COD method was developed by Rintamaa et al. [1]. The equipment uses an
inverted impact testing arrangement (rigid hammer with attached specimen, mobile
support), which allows the crack opening to be measured and reduces the load oscillations
due to the impact impulse.
Measurements of the acoustic emission during a Charpy impact test were already
carried out by Crostack et al. [6] with a AE probe fixed at the specimen. Richter et al. [7,
8] used a AE probe installed in the tup of the impact hammer. It could be proved that
characteristic AE signals appear in the temporal and spatial vicinity of the crack initiation.
Applications of ME for ductile crack initiation detection were demonstrated by Winkler
[9] and Lenkey et al. [10].
Experimental Procedure
Materials and Specimens
Two types of steels were used: (1) a heat resistant 10CrMo9 10 steel according to
DIN 17243 and (2) a ASTM A533 grade B class 1 reactor pressure vessel, designated JRQ
and used as IAEA reference material. The chemical composition is given in Table 1.
Table 1 - Chemical composition (weight-%) of the materials used
Material C Si Mn P S Mo Cr Ni Cu
10 CrMo 9 10 0.10 0.32 0.55 0.01 0.019 0.96 2.32 0.10 0.15
A533B1 (JRQ) 0.18 0.24 1.41 0.02 0.007 0.50 0.13 0.87 0.15
The commercially fabricated forging plate from 10CrMo9 10 has a cross sectional
dimension of 70 ram" 120 mm. To have a relatively wide range of toughness and strength
the plate was austenitized at 950~ oil quenched and tempered for 2 hours at 600~
(code D), 640~ (code E), 720~ (code F) and 760~ (code G). Different levels of
toughness were, furthermore, reached by different orientation of the specimen (L-T, T-L
according to ASTM Standard Test Method for Plane-Strain Fracture Toughness of
Metallic Materials (E 399)). Table 2 lists the types and mechanical properties of the
different materials.
Altogether, materials with the yield strength range of 390 MPa to 750 MPa, the upper
shelf energy of 55 J to > 300 J and a transition temperature of-105 ~ to -13 ~ were
available.
Charpy size specimens were used with 20% side grooving and precracking up to a
crack length-to-width-ratio adW = 0.5.
Table 2 - Mechanical properties of the tested materials
Material Rpo,2 R m Ag A5 Z USE T6s j Red Rind

MPa MPa % % % J ~ MPa MPa
A.S.TM.S33...B!i.JR..Q....(..L...-.T).....4Z7.......6.3..O....................... 26 ..........76 ....... .750....... q,3........653 .......7O6....
10 CrMo 9 10: D (L-T) 739 846 7 18 73 191 -47 885 1119
D (T-L)
752 850 7 16 64 55 870 995
E (L-T) 630 723 8 21 76 206 -99 776 973
F (L-T) 457 569 10 26 80 260 -96 635 875
F (T-L) 465 588 10 22 69 100 -20 625 835
G (L-T) 393 513 17 35 81 >300 -105 550 783
R~.2 0.2 % yield strength; USE uppershelfenergy;
Rm tensile strength; TT6s J transition temperature related to 68 J impact energy;
Ag uniform elongation; Red dynamic stress at general yield load;
A5 total elongation; Rind dynamic stress at maximum load
Z reduction of area;
Instrumented Impact Testing
The block diagram in Fig. 1 shows an overview of the experimental arrangement.

Two arrangements were applied. The first arrangement uses the standardized version of a
pendulum impact tester (named FZR-Tester). It has a maximum initial energy of 300 J
and is instrumented to measure the impact force according to the ISO/DIN Steel-Charpy
V Pendulum Impact Test- Instrumented Test Method (14556) standard. In order to
measure AE during loading a piezoelectric broadband AE sensor was designed and
installed within a drilled hole at the impact tup. The distance between sensor and crack
front is 29 mm at the beginning of loading.
The second arrangement is an inverted and laser instrumented pendulum
equipment (named VTT-Tester) [1]. This arrangement measures COD by means of laser
beam which is scattered at the crack surfaces. For assuring that the results of both
equipments are comparable, the impact tup of the VTT-Tester was modified and equipped
with the FZR-AE transducer in the same manner as the tup of the FZR-Tester. Fig. 1
provides some characteristic parameters of the measuring system.
The specimens were impact-loaded by 3-point-bending with an impact velocity of 2,8
m s j and an impact energy of 79 J (FZR) and 161 J (VTT). These conditions remained
constant for all tests. The test temperature interval o f - 5 0 to +150 ~ covered the lower
transition range and reached the upper shelf energy range. After testing the specimens
were microfractographically evaluated and both ductile crack growth and stretch zone
width (SZW) were measured.
Fpreamolifier
AE (t) | 1 4 0 dB" amplifier
Fd (t) Lf,IIr~Filtering_
sookHz 0...60 dB ,
iiiii i | iiiiiiiiii--fI - plifier I
i D.jgital Storage
Optical Q) ~ (-~ QfLcilloscop~eDSO
Oh1 Oh2 Ch3 200000 pt8
COD
__ device _
t 12bit
9/'1 IdH7
I I
C O D (t)
~ Moving hammer ofVTT tester~ Moving Instrumented tup (AE, load) ~i Specimen
(=~) or fixed anvil of FZR, of FZR (,b) pendulum or fixed tup of
respectively V'I-I" tester, respectively
Figure 1 - Experimental arrangement usedfor conventional (FZR-Tester) and inverted

(VTT-Tester) impact testing equipments.
Results
A typical load curve of a impact loaded specimen at the ductile-to-brittle region and
the simultaneously recorded AE which were obtained with the FZR-Tester at room
temperature are shown in Fig. 2. From the load vs. time curve three different phases earl
be recognized. They are also connected with typical acoustic features. Phase I shows
strong oscillations of the load and burst-like AE-signals. Small load oscillations and large
plastic deformation characterize phase II. It is accompanied by a lower acoustic intensity.
The characteristics of acoustic emission clearly differs from phase I. The Phase III begins
with a load drop caused by a discrete unstable crack extension (pop-in). The cleavage
fracture event releases a high energy rate and is additionally connected with the loss of
contact between the specimen and the striker. Both events generate the strong AE signal
seen in Fig. 2.
The initiation of stable crack growth cannot be identified from the load trace, but an
analysis of AE regarding the signal amplitude and duration renders it possible. In
addition, low blow tests with varying blow angle and the Cleavage R-Curve method were
chosen to find the point of stable crack initiation. Both methods provide approximately the
same initiation deflection d~* after extrapolation to zero crack growth. The time ti*
pertaining to initiation deflection is marked in Fig. 2.
Figure 2 - Characteristic load (Fd) and acoustic emission (AE) curves for the
conventional FZR arrangement in the TT region of toughness.
The following different AE types are observed near the initiation time t~':
- Type I: bursts characterized by a duration o f < 7 ~ts and a signal amplitude o f > • V
- Type II:bursts characterized by a duration o f > 10 ~ts and a signal amplitude o f < +0.5 V.
Signals of type I are caused by mechanical (inertia) effects during initial impact loading
(phase I). However, signals of type II usually occur very close to the estimated crack
initiation time t~'. In agreement with the results mentioned in references [6-8] the
occurrence of signal type II seems to be associated with the beginning of ductile crack
extension.
Until now, a convincing physically based model which describes the relation
between material processes during crack initiation and the AE has not yet been available.
Thus, the method cannot be verified directly. It is, however, possible to compare the
obtained values with results of other multiple or single specimen methods. For this reason
experiments with the modified VTT-Tester were performed. In this way crack initiation
could be determined at the same specimen by two independent methods. The use of
material of different strength-toughness-relations generalizes the comparison.
In Fig. 3 characteristic signal sequences (load, AE and COD) obtained from tests with
the VTT-Tester are depicted. The detection points for crack initiation are marked. The
load and AE signals are comparable with the signals measured by the FZR-tester. The
most characteristic phenomena could be recognized by the AE- and COD-method. Only
in the lower transition range neither the COD method nor the AE method provide
evidence of crack initiation. This is not surprising for the COD method because the
evaluation procedure requires indispensably a sufficiently large amount of ductile crack
growth. The AE-method probably fails as the ductile crack initiation already starts before
the impact-induced AE signals faded~ These impact-induced AE signals are not reduced by
the inverted arrangement in contrast to the inertia and load oscillations, which are weaker.
~nstable crack onset ,

Force
0
i
AE o 1 ,
M rllr I rrr
. . . . . . . . . . .
'" ' ......
-5 ~q__~ra~ iniiia~o~/.
./
0 0.2 0.4 0.6 0.8 110

Time [ms]
Figure 3 - Characteristic load (Fa), acoustic emission (AE) and laser COD curves for
modified inverted VTT arrangement in the TT region of toughness.
As far as the characteristic phenomena for crack initiation could be detected, the crack
initiation times were determined by AE or COD respectively. The concerned load versus
displacement curves were used to calculate the absorbed deformation energy and the J-
integral J~d in the same way as recommended by the ESIS Procedure for Determining the
Fracture Behaviour of Materials (P2-92). This J-integral is defined as dynamic crack
initiation toughness J~dc~ or JidA~. The upper index refers to the method used.
A comparison of dynamic toughness results is given for 10CrMo910 steel (heat
treatment D) in Fig. 4 as a transition curve. In addition to the crack initiation toughness
measured by COD and AE, the J-integral at unstable crack growth Jdna and the J-integral
at the maximum load Jdm are presented there, too. Their courses provide the typical
temperature dependence of the toughness behaviour offerritic steels.
The
Figure 4 - Dynamic J-integrals in dependence on the temperature for steel

10 CrMo 9 10, state D.
The crack initiation toughness Jid AE is only a little higher than the cleavage crack
initiation J ~ in the lower shelf region and depends hardly on the temperature. The COD
method indicates crack initiation clearly later and this results in higher values of JidcOD.
These values weakly decline with increasing temperature, similar to the temperature
dependence of the J-integral at maximum load J~. The same tendencies are found for the
other materials or material conditions. The weak temperature dependence was neglected
for further evaluation and mean values were calculated from all results for each set of
specimens. These values are given in Table 3 together with the crack initiation toughness
J~dszw and the J integral at crack extension of 0.2 mm (Jd0.2)which were determined by the
cleavage-R-curve method.
The physical crack initiation toughness Jid szw w a s determined by measuring the stretch
zone width (SZW). Then the physical crack initiation toughness is equal to the J-integral
at a crack growth Aa~= SZW. A disadvantage of the method is that crack extension cannot
be predestinated. Thus, the positions of measuring points are stochastically distributed on
the R curve. In these experiments only few points lay at crack extensions ~ 0.2 mm.
Therefore, the curve has to be extrapolated from the range of higher crack growth and has
a limited accuracy in the range of crack initiation.
Dynamic crack resistance curves determined in this way are shown in Fig. 5 for two
orientations of condition D of 10 CrMo 9 10 steel. One can see the lower crack resistance
and, thus, the lower toughness of the T-L orientation.
Table 3 - Dynam& ductile crack initiation toughness parameter (,lid)determined by

single specimen methods in comparison with the crack initiation
toughness Jidszw and Jdo.2
Ductile crack initiation toughness parameters [N/mm]

Material S Z W [p.m]
Jid SZW *) Jid AE Jid COD Ja 0,2 *)
A S T M A533B1
J~..Q...~L.-.!.)"...............
.8...5...~..1,8.
..............
!.~.9..~..~6 .................
1.9L~..2..3"................
!.~3.~..2..z.................
..2.25.......
10 CrMo 9 10
D(L-T) 44• 4 83• 5 83• 148• 237
D(T-L) 24• 1 46• 1 76• 109+30 133
E (L-T) 42 • 7 79 + 10 74 • 24 144 • 26 262
F(T-L) 29+ 1 60+ 2 73• 116• 191
G (L-T) 9 2 • 14 1 4 2 + 16 62• 115• 252
*) curvefit accordingto power law: Jd = A Aa~
Fig. 5 also shows the crack initiation toughness determined by the different methods.
It is evident that the COD method defines crack initiation afterwards the crack was already
extended a little. The initiation parameter JidcOD is situated above J~dszw values and is
250
I 0 CrMo 9 10 i i /
heat treatment D Ii i : L-T
200 I.J.Z._I IIIIZ~is
i
.................................................................. i
. . . . . . . . . .... ....
150
0~
.o lOO
:" - J ~ . . . . d-'- . . . .
50 ~ ~ . ~ ................................
r
0.05 0.1 0.15 0.2
Stable crack growth a a [rnm]
Figure 5 - Dynamic Ja-~a- crack resistance curve with scatter band for ductile crack
initiation obtained by different methods.
smaller than the crack initiation values related to 0.2 mm crack extension (Jd0.2). AE shows
crack initiation near the SZW-defined physical parameter Jidszw obtained with specimens
in L-T orientation. In other cases the Jid^E values clearly differ from Jidszw.
Assuming that the SZW method provides the correct (physical) crack initiation
toughness, Fig. 6 illustrates a summarizing evaluation. In Fig. 6 the distinctly defined
ductile crack initiation parameters are correlated to the physical crack initiation Jidszw.
A good correlation to the 1: 1 line shows the Jd0.2trend. This trend line is shifted to
about 120 N/mm higher values in comparison to the 1:1 line. That can be expected since
the parameter is determined from the same R-curve but at larger crack extension. The Jm
parameter also shows a correlation to Jidszw. However, the slope of the trend line is steeper
than the 1:1 line. At the maximum load a distinct crack extension occurred. The Jdm
parameter is influenced by the constraint and, thus, not only by specimen size but also by
preceding plastic deformation. COD indicates crack initiation after an amount of
macroscopic ductile Crack growth which is small but cannot be ignored. The correlation to
Jid szw depends on the toughness. For high toughness steels the correlation is unsatisfying.
If the two measuring points of high toughness are neglected there is a 1:1 correlation
between Jidc~ and J~dszw although shifted to higher Jid parameters. This stimulates two
conclusions. First, the COD method responds to crack initiation with delay and is suitable
to determine an engineering approach of the ductile crack initiation toughness. Second, the
COD method is not applicable without restrictions. Large plastic deformation before crack
initiation additionally influences the COD measurement and affects the results.
~, .- [~10CrMo9 10 A533B1
,00tl:oi o 9
•
400
~ ,..- ~" O .......... If .................. ~ - - - -
200
<~7.~' - ......... ~_ ] j coo trend ]
0 0
40 60 80 100 120 140 160

j~zwiN/ram]
Figure 6 - Comparison of different dynamic ductile initiation J-Integrals with the
physical crack initiation toughness J~dszW.
The Jid AE parameter is not only very close to the 1:1 trend line but also near the physical
crack initiation toughness Jidszw. The condition G of the 10 CrMo9 10 steel is not
coincident with the 1:1 line. In this condition the 10 CrMo9 10 steel shows pronounced
Lueders Strain. The microstructure contains a high fraction of ferrite and relatively coarse
carbides. This suggest that the correlation is not generally valid but it depends on the type
of steel. Further, it should be reminded that in some cases no typical AE appears near the
crack initiation, thus, the results can better be understood based on the assumption that AE
is not really a consequence of the onset of ductile crack growth but it is caused by material
processes which occur in the spatial and temporal vicinity of ductile crack initiation.
Previous studies [5, 6] support this assumption.
A single-specimen test methods for the determination of dynamic fracture toughness

parameters using the instrumented Charpy impact test was developed. The method is
based on the measurement of the impact energy up to the onset of ductile crack extension
and on the calculation of the corresponding J integral. To detect the ductile crack
initiation, emitted acoustic (AE) signals were used and the result was compared with the
laser optical measurements of crack opening displacement (COD).
Both the COD and AE methods are suitable to provide a fracture mechanics
parameter to estimate the onset of ductile crack growth, at least if the ductile crack growth.
The use of these methods provides a better approximation of the critical load for crack
initiation than the use of the maximum load point as often proposed. However, the
methods do not detect the same material processes. Whereas the COD method responds
to ductile crack growth and also depends on the intensity of the deformation processes in
the ligament, the AE method is sensitive to the ductile crack initiation and provides values
near to the physical crack initiation toughness as determined by multiple specimen
techniques and measurement of stretch zone width. Thus, the AE measurement is an
appropriate method to determine a dynamic parameter of elastic-plastic fracture
mechanics. The measurement can be integrated in the instrumented impact testing without
additional instrumentation or preparation of specimens. Because the method is relatively
simple and convenient, it is particularly suitable for the testing of irradiated specimens
under hot cell conditions. On the other hand the method is not applicable for any steel,
through.
Acknowledgments
The authors would like to express their appreciation to German "Deutsche

Forschungsgemeinschaft (DFG)" for financial support. This work was partly performed at
VTT Manufacturing Technology of the Research Centre of Finland.
References
[1] Rintamaa, R., "Single Specimen Fracture Toughness Determination Procedure Using
Instrumented Impact Test", VTT Publication 140, Technical Research Centre of
Finland, Espoo, 1993
[2] Roos, E., and Eisele U., "Determination of Material Characteristic Values in
Elastic- plastic Fracture Mechanics by Means of J-integral Crack Resistance
Curves", Journal of Testing and Evaluation, JTEVA, Vol. 16, No. 1, Jan, 1988, pp.
1-11
[3] MacGillivray, H. J. and C. E. Turner 1989, "A Comparison of Dynamic R-curve
Methods", Int. Phys. Conf. Set. No 102: Session 1, Paper presented at Int. Conf.
Mech. Prop. Materials at High Rates of Strain, Oxford, 1989
[4] Demler, T., "Untersuchungen zur RifSinitiierung unter schlagartiger Belasttmg",
Forschtmgsbericht MPA Stuttgart, 1990.
[5] B6hme, W., "Experience with Instrumented Charpy Tests Obtained by a DVM
Round-Robin and Further Developments", Evaluating Material Properties by
Dynamic Testing, ESIS 20, (Ed. E. van Walle), Mechanical Engineering
Publications, London, 1996, pp. 1-23
[6] Crostack, H.-A., Steffens, H.-D., and Engelhardt, A. K., "Measurement of Dynamic
Fracture Toughness Applying Acoustic Emission Techniques", Proc. of an lnternat.
Conf., Analytical and Experimental Fracture Mechanics, G.C. Sih, M. Mirabile,
Eds., Alphen an den Rijn (Netherlands), 1981, pp. 507
[7] Richter, H., B6hmert, J., and Viehrig, H.-W., "Determination of Crack Initiation
Toughness by Impact and Dynamic Testing", Proceedings of the 11th European
Conference on Fracture - ECF 11, Vol, III, J. Petit, Ed., Poitiers-Futuroscope,
France, 1996, pp. 2001-2006
[8] Richter, H., "Untersuchungen zum Ril]einleitungsverhalten von Stahl unter
schlagartiger Beanspruchung mittels Schallemission", Statusberichte zur
Entwicklung und Anwendung der Schallemissionsanalyse, 11. Kolloquium
Schallemission, DGZfG-Berichtsband 58, Jena, Germany, 1997, pp. 77-88
[9] Winkler, S., "Magnetic Emision Detection of Crack Initiation", Fracture
Mechanics." Twenty-First National Symposium, ASTM STP 1074, J. P. Gudas, J. A.
Joyce, E. M. Machell, Eds., Philadelphia, American Society for Testing and
Materials, 1988, pp. 178-192
[10] Lenkey, G., Winkler, S., Major, Z., and L6vay, I., "Applicability of Magnetic and
Electric Emission Techniques for Detecting Crack Initiation in Impact Tests",
Proceedings of the 11th European Conference on Fracture - ECF 11, Poitiers-
Fnturoscope, France, 1996, pp. t 989-2000
Gy6ngyv6r B. Lenkey1
On the Determination of Dynamic Fracture Toughness Properties by Instrumented

Impact Testing
Reference: Lenkey, G. B., " O n the Determination of Dynamic Fracture

Toughness Properties by Instrumented Impact Testing," Pendulum Impact
Testing: A Century of Progress, STP 1380, T. A. Siewert and M. R Manahan, Sr.,
Eds., American Society for Testing and Materials, West Conshohocken, PA, 2000.
Abstract: In the engineering practice it is of importance to know the effect of loading

rate on the material behaviour, e.g., on the fracture mechanics properties. One of the most
widely used testing techniques for determining these properties is the instrumented
impact test. Depending on the material behaviour under a given loading condition,
different fracture mechanics parameters should be determined. Most of the evaluation
procedures of these parameters are still not standardised for higher loading rates. So the
aim of the present work was to investigate the applicability of different methods for
determining dynamic fracture toughness properties. Instrumented impact experiments
combining with magnetic emission measurement have been performed on pre-cracked
Charpy-V specimens of E420 C and QStE 690 TM high strength steels. Critical values of
stress intensity factor (Kid, Kji) related to the initiation of the fracture at different
temperatures have been determined using different methods depending on the type of the
fracture.
Keywords: instrumented impact testing, dynamic fracture toughness, magnetic emission

measurement, impact response curve method, high strength steel
Nomenclature
ao Initial crack length of pre-cracked specimen, mm

B Specimen thickness, mm
cM Machine compliance, m/N
f Displacement, mm
f~ Displacement at initiation, mm
F, Force, N
J02d Critical J-integral value at 0.2 mm stable crack extension, kJ/m 2
Head of department, Department for Structural Integrity, Bay Zolt~n Applied Research
Foundation Institute for Logistics and Production System, H-3519 Miskolctapolca,
Igl6i u. 2., Hungary.
366
LENKEY ON DYNAMIC FRACTURE TOUGHNESS 367
Jid Critical value of dynamic J-integral at ductile crack initiation, kJ/m 2

Kid Dynamic fracture toughness, MPa~/m
Kjc Critical stress intensity factor calculated from J-integral at brittle crack initiation,
MPax/m
Kji Critical stress intensity factor calculated from Jid, MPax/m
m Mass of the hammer, kg
ME Magnetic emission signal, V
MF Magnetic field, Vs
k, R Constants
T Temperature, ~
t Time, s
tF Time to fracture, s
Ui Energy absorbed up to crack initiation, J
v Velocity of the hammer, m/s
v0 Impact velocity, m/s
W Specimen width, mm
x Time, s
One very important condition for the reliable operation of structures is that the danger
of existing cracks or crack-like defects has to be judged. This can be done by using
fracture mechanics approaches. Usually the dynamic loading is more critical for a
structure, thus the determination of dynamic fracture mechanics properties of materials is
of importance. One of the most widely used techniques for this purpose is the
instrumented impact testing of pre-cracked specimens.
Depending on the fracture behaviour of the specimen, different dynamic fracture
mechanics parameters should be determined and different evaluation procedures should
be used. In the case of a ferritic steel material the fracture toughness shows a temperature
dependence (Figure 1 [I]).
In the lower shelf region (I) the cleavage fracture dominates, and in the upper shelf
region (III) the ductile dimple failure is typical. Mixed mode fracture is usually observed
in region II There are standardised testing and evaluation procedures for determining
fracture mechanics characteristics for quasi-static loading in the different regions, but the
procedures for dynamic fracture toughness determination are still not standardised. So the
aim of the present work is to analyse the methods and applicability of the different
evaluation procedures for determination of dynamic fracture mechanics characteristics.
Experiments
Instrumented Charpy impact experiments have been performed on pre-cracked
samples of two high strength steel materials. The mechanical properties of the
investigated materials are shown in Table 1. E420C is micro-alloyed ferritic steel, and
QStE 690 TM is thermo-mechanically treated high strength steel of bainitic-martensitic
microstructure. The instrumented impact experiments were combined with magnetic
emission measurement [2] using a 300 J instrumented impact pendulum (Figure 2).
368 PENDULUMIMPACTTESTING:A CENTURYOF PROGRESS
I
Kid Ductile/ I Ductile
KI t Cleavage I
Kj i Cleavage Kid I Kji
Kid K~
I H II ]II
[
[
i.mpact response [ I 1 Dynamic j_R eurve
curve m e t h o d l ~ l " I . . . . . . ] I or erackinitiation
[ detection
[ I
I I
I i
Temperature
Figure 1 - Schematic Presentation of a Ductile-to-Brittle Transition Curve
for Fracture Toughness and the Evaluation Methods
o b'D
13
P
9 i
f{
OoOoO110
'l /
1 Angle measurement 8 Voltage supply and amplifiers
2 PSD 300/150 impact pendulum 9 Clock display
3 Instrumented tup 10 Digital memory scope
4 Strain gage 11 Computer
5 Magnetic and electric probes 12 GPIB card
6 Optical trigger device 13 Printer
7 Triggering flags
Figure 2 -Instrumented Impact Pendulum and the Data Acquisition System
Table 1 - Mechanical Properties of the Investigated Steel Materials
Material Yield strength, Ultimate tensile strength,

MPa MPa
E420 C 444 593
QStE 690 TM 768 854
The force measurement is carried out by strain gages glued on both sides of the tup
(3). The magnetic emission probe (5) is located in a metal box close to the notch root of
the specimen. An external optical trigger device (6) is used to start the data acquisition.
When the two pins flag on the hammer (7) goes through the optical trigger device, two
pulses are produced and the time interval between these pulses is measured by a clock
(9). From this time the impact velocity can be determined. The strain gages and the
emission probe are connected to the voltage supplies and the amplifiers (8), their signals
are recorded by a tour channels digital oscilloscope. The data could then be transferred to
a PC and the data evaluation procedure is usually done with spreadsheet programs.
In the case of E420C steel two different impact velocities were applied: v0 = 2.75 m/s
and v0 = 5.5 rn/s to investigate the effect of loading rate as well. Only a few pre-cracked
specimens of QStE 690 TM steel have been tested at different temperatures with 5.5 m/s.
The pre-cracking was done on an Amstler resonance machine producing a/W .~ 0.5
relative crack length. The impact experiments were performed at different temperatures
between 20~ and -60~ The fracture surfaces were studied by SEM.
Evaluation Procedures and Results
Upper Shelf Region
In the case of E420C steel fully ductile behaviour was observed at room temperature.
In this upper shelf region there are two possibilities to characterise the material resistance
to crack propagation:
1. to determine J0.2d on the basis of the dynamic J-R curve, or
2. to determine Jid at ductile crack initiation.
In the latter case additional measurement techniques should be applied to find the onset
of stable crack extension, e.g. strain gages, stretch zone measurement, COD
measurement, etc. The magnetic emission measurement has been successfully applied for
this purpose [3], so the magnetic emission signals were used for determining the onset of
ductile crack extension. Stable crack initiation can usually not be determined directly
from ME signals. This is demonstrated by an experimental result of a ductile behaving
specimen (Figure 3a). For these applications a method was developed by Lenkey and
Winkler [3] which uses the integrated ME signal, i.e., the magnetic field history:
MF(t) = J'ME(x) dx (1)

x--0
It was observed that ME signals originated by crack propagation can be distinguished

from those originated by Barkhausen noise in the field curve by a change of the slope.
This is demonstrated in Figure 3b. The rupture event is indicated by a discontinuity in the
slope of the field curve.
! Fi i
' '1I"~*~TI~It,.wlVq, "" ......

. . . . i . . . . i . . . . I . . . .
0 0.5 1 1.5 2
t, ms
Figure 3a - Force and ME Signals o f Pre-Cracked Sample

(E420 C Steel, vo = 5.5 m/s, T = 20~
Z 5 ........ : ~.
~. 0 ,, inltlatlon 5
.,Oo to! ,o
Figure 3b - Force and MF Signals of Pre-Cracked Sample

(E420 C Steel, vo = 5.5 m/s, T = 20~
Utilising this field method the initiation point can be determined. The critical value of
dynamic J-integral related to the initiation can then be derived with eq. 2 [4]:
2.U i
Jid = B 9(W - a0) (2)
where Ui was calculated by integrating the force-displacement curve using eq. 3:
fi 9
U i = IF(f)df (3)
f=0
The displacement was determined from the measured force-time curve using eq. 4 and 5:
f(t) = Iv(~)dx (4)

X=0
1 t
v(t) : v 0 - m [F(x)dx=,~ (5)
The mass of the hammer was m = 20 kg, the v0 impact velocity was measured.
Then the relevant Kid value can be determined for plain stress condition using eq. 6:
Kji = ~ E . Jid (6)
where E = 210000 MPa was the Young's modulus of the investigated steel materials.
Transition Region
In the transition region usually brittle crack initiation occurs after significant plastic
deformation, which was observed at -20~ and -40~ in the case of E420 C steel, and at
room temperature in the case of the other steel material. The QStE 690 TM steel showed
very interesting fracture behaviour at room temperature. Unstable crack propagation
cannot be seen from the force-time curve (Figure 4a) and also not from the ME and MF
signals (Figure 4b). But studying the fracture surface by SEM, cleavage fracture was
found near the pre-crack front following the stretch zone and a very narrow strip of
dimple fracture (Figure 5). Jia and Kji values at crack initiation were determined using eq.
(2)-(6) for these cases.
15 500
lO- 400
5 300
z o 200
uZ -5 100 ,~"
-lO 0
-15 -I00
-20 -200
o 0.5 1 1.5
t, ms

(QStE 690 TM Steel, vo = 5.5 m/s, T = 20~
15
lO
Z ,,.,1
u." It.
I[
0
-5
0.5 1
t, m s
Figure 4b - Force and MF Signals o f Pre-Cracked Sample

(QStE 690 TM Steel, vo = 5.5 m/s, T = 20~
Figure 5 - S E M Picture o f Pre-Cracked Sample

(QStE 690 TM SteeI, vo = 5.5 m/s, T = 20~
In the case of some E420 C steel samples in the transition region ductile crack
extension preceded the brittle fracture. One example is shown in Figure 6. This was
proved by SEM investigation of the fracture surface (Figure 7). In these cases Jid and Kji
values were calculated using eq. (2)-(6).
...... i ...i......F ..... i..........i..........i.. i.

z
i i ME I ~ i i
:T:r' i : :, iI,
0 I 2 3 4 5 6 7
t, ms

(E420 C Steel, vo = 2. 75 re~s, T = -40~
10 10
I
5
u2
~-=i initiation
-5
MF
.10 , : ; ,--~, ; :~: : I : : : : ,, 0

0 0.5 1 1.5 2
t, m s
Figure 6b - Force and M F Signals o f Pre-Cracked Sample

(E420 C Steel, vo = 2. 75 m/s, T = -40~
Figure 7 - S E M Picture o f Pre,Cracked Sample

(E420 C Steel, vo = 2. 75 m/s, T = -40~
Lower Shelf Region
In the lower shelf region and sometimes in the transition region brittle crack initiation
occurs that is not preccded by any macroscopic plastic deformation. One example is
shown for E420 C steel in Figure 8. In these cases the "3x" criteria - proposed by Ireland
[5] - was usually not fulfilled, i. e. brittle fracture occurred before 3 load oscillations.
Therefore the quasi-static equations cannot be used any more for determining KId. For
these cases the impact response curve method - proposed by Kalthoff [6] - was applied.
For this method the time to fracture (tF) is to be determined.
Unstable crack propagation is indicated by a force drop and usually is accompanied
by a sharp peak of the magnetic signal according to a rapid crack jurnp. But due to the
strong oscillation of the force signals sometimes it was difficult to determine the instant
of the brittle fracture directly from the force-time curves. In these cases the magnetic
emission signals or sometimes the magnetic field signals had to be used for determining
the time to fracture (Figure 8b).
177 !lTi'i:' i
0 0.5 1.5 2
t, ms
Figure 8 a - Force and ME Signals o f Pre-Craeked Sample

(E420 C Steel , vo = 5.5 m/s, T = -40~
1o~ ~ , lO
F..[..t..~
Zot'............... ......=
......~
...... i
-o, ,11t
(E420 C Steel, vo = 5.5 m/s, T = -40~
The QStE 690 TM steel showed a relatively high ductility at 0~ since a long stable
crack propagation was observed on the force curve (Figure 9a). But the crack initiation
was brittle and occurred very soon after the impact, which could be seen also on the
fracture surface, where only a narrow stretch zone formation preceded the cleavage
fracture (Figure 10). Then at lower temperatures no stretch zone was found. Since the
crack initiation occurred before three oscillations of the force signal, in these cases also
the impact response curve method had to be applied.
15 500
10 400
5 300
== o 2oo
-5 loo
-lO o
-15 -1 oo
-20 -200
0 0.5 1 1.5
t, ms

(QStE 690 TM Steel, vo = 5.5 re~s, T = O~
15 20
10 15
5
Z "5
~ o
-5 0
-10 -5
0 0.5 1
t, ms

(QStE 690 TMSteel, vo = 5.5 m/s, T = O~
Figure 10 - SEM Picture o f Pre-Cracked Sample

(QStE 690 TM SteeI, Vo = 5.5 m/s, T = O~
According to the impact response curve method, with the measured time-to-fracture
the dynamic fracture toughness can then be determined with eq. 7 [6]:
Kid = R. v 0 9t" (7)
where t ''= fit') is given in the tables of Reference [6]
and
t'=g(t) tF . {1-- 0.62(~ -- 0.5) + 4.8(~ -- 0.5) 2} (8)
where R = 301 GN/m s/2 for a machine with CM= 8.1 xl 0 "9 mfN machine compliance. If
the machine compliance differs from this value, a first-order correction factor should be
used in calculating R, that is:
1.276
k = (1 + 0.276.8.1CM0_9) (9)
The machine compliance was CM= 2.335x10 8 m/N which was determined from elastic
low-blow experiments of unnotched high strength steel sample.
Summary of the Experimental Results
The experimental results for E420C steel are summarised in Tables 2, 3 and Figure
11.
As can be seen from Figure 11, K~d and Kji values in the upper shelf and in the lower
shelf region do not differ significantly for the two different loading rates. But the brittle
initiation appeared at a higher temperature (-20~ in case of the higher impact velocity,
while in the case of the lower impact velocity brittle initiation was observed at T = -40~
preceded by significant plastic deformation; the linear elastic behaviour has been reached
only at T = -60~
Table 4 summarises the results for QStE 690 TM steel. As was discussed earlier
crack initiation occurred at the very beginning of the loading process even at room
temperature. At room temperature this material showed significant plasticity before
initiation followed by microscopically brittle fracture, but at lower temperatures more
brittle behaviour was observed without any macroscopic plastic deformation before
initiation. The crack arrest capability of this material is very high, since significant stable
crack propagation was observed after the brittle initiation even at -40~
Table 2 - Specimen Dimensions and Evaluated Parameters for E420 C Steel

(vo = 5.5 re~s)
Specimen T, a0, W, B, Ui, Jid, Kji, Kid,

No. ~ mm mm mm J kJ/m2 MPa4m
1 20 4.86 9.65 10.05 4.2 174 191
2 20 4.97 9.7 10 4.09 173 191
3 20 4.98 9.7 10 4,38 186 197
4 0 4.98 9.7 10 3.66 155 180
5 0 4.74 9.65 10 4.55 185 197
6 0 4.91 9.7 10 3.6 150 178
7 -20 4.82 9.6 9.975 3.97 167 187
8 -20 538 9.65 10 - - 92
9 -20 4.77 9.55 10 4.7 197 203
10 -40 4.82 9.65 1 0 - 80
11 -40 4.8 9.7 10.05 - 54
12 -40 4.86 9.725 10 - - 53
13 -60 4.96 9.7 10.05 - 54
14 -60 4.81 9.7 10 - - 72
15 -60 5.09 9.725 10 - - 55
L E N K E Y ON D Y N A M I C FRACTURE T O U G H N E S S 379
Table 3 - Specimen Dimensions and Evaluated Parameters for E420 C Steel

(vo = 2:75 re~s)
Specimer T, ~, W, B, Ui, Jid, Kji, Kid,

No. ~ mm mm mm J kJ/m2 MPaqm
16 20 5.07 9.625 9.95 3.89 172 190
17 20 4.94 9.6 9.95 4.04 174 191
18 20 4.93 9.675 9.95 3.96 168 188
19 0 4.83 9.7 10 4.3 177 193
20 0 4.69 9.96 10 4.5 171 189
21 0 4.92 9.6 10 4.1 175 192
22 -20 4.83 9.55 10 4.37 185 197
23 -20 4.81 9.5 10 3.56 152 178
24 -20 4.65 9.675 9.975 3.68 147 176
25 -40 4.79 9.675 9.95 3.63 149 177
26 -40 4.64 9.65 9.95 3.55 142 173
27 -40 4.76 9.65 9.975 3.52 144 174
28 -60 4.67 9.65 10 - - 69.1
29 -60 4.61 9.55 10 - - 84.1
30 -60 4.71 9.65 10.1 - - 69.7
250
15o / ,-'"'"
9"-" 100 ,
.-
so .........
o I
-80 -60 -40 -20 0 20 4O
T, ~
o E420 C (v0=5.5 m/s) 9 E420 C (v0=2.75 m/s)
......
~K
Mean values
QStE 690 TM (v0=5.5 m/s)
~ Mean values
1
Figure 11 - Dynamic Fracture Toughness vs. Temperature
for E420 C and QStE 690 TM Steel Materials
Table 4 - Specimen Dimensions and Evaluated Parameters for QStE 690 TM Steel
(vo = 5. 5 re~s)
Specimen T, ao, B, W, Ui, Jid, Kji, Kid,

No. ~ mm mm mm J kJ/m2 MPa4m
1 20 5.5 10 10 5.82 259 233
2 0 5.27 10 l0 - - 117.6
3 -20 5.47 10 10 81.2
4 -40 5.23 10 10 - 54.1
The standard validity requirements for J1 and KI [ASTM Test Method for Plane-Strain
Fracture Toughness of Metallic Materials (E 399) and ASTM Test Method for JIc, A
Measure of Fracture Toughness (E 813)] are usually not fulfilled for pre-cracked Charpy-
V specimens. This is the situation now, since the validity limit for Kl is 28 MPax/m for
E420C steel and 48 MPa~/m for QStE 690 TM steel for this specimen geometry: And all
Kji and Kid values exceed these validity limits. The validity limit for Ji (59 kJ/m2 for
E420C steel and 102 kJ/m 2 for QStE 690 TM steel) is also lower then any of the
calculated Jid values. But the obtained dynamic fracture toughness parameters can be
used for comparison.
Conclusions
The initiation toughness of two high strength steel materials has been determined
with instrumented impact testing at different temperatures and loading rates. From the
results the following can be concluded:
1. The magnetic emission measurement applying the "field method" could be used for
determining the instant of stable crack propagation.
2. In the case of brittle initiation due to the strong oscillation of the force signals,
sometimes it was difficult to determine the instant of the brittle fracture directly from the
force-time curves. In these cases the magnetic emission signals or sometimes the
magnetic field signals had to be used for determining the time-to-fracture, and the
impact response curve method could be applied for determining Kid.
3. The loading rate in the order of 1 m/s has significant effect on the brittle-to-ductile
fracture transition behaviour of the E420 C high strength micro-alloyed steel. The
position of the KId(T) curve has been shifted by approximately 20~ to higher
temperatures due to increasing the impact velocity from 2.5 m/s to 5.5 m/s.
4. In the case of QStE 690 TM steel, crack initiation occurred at the very beginning of
the loading process even at room temperature. Proof of brittle initiation was found on the
fracture surface by SEM investigation. This brittle initiation was quickly followed by
crack arrest and long stable crack propagation even at -40~
5. QStE 690 TM steel has higher initiation toughness at room temperature then E420 C
steel, but the transition region is at higher temperatures.
Acknowledgement
Financial support of the OTKA T 030057 project is gratefully acknowledged.
References
[1] Rintamaa, R., "Single specimen fracture toughness determination procedure using
instrumented impact test," Evaluating Material Properties by Dynamic Testing,
ESIS 20, Mechanical Engineering Publication, London, 1996., pp. 97-123.
[2] Winkler, S. R., "Magnetic Emission Detection of Crack Initiation," ASTM STP
1074, Philadelphia, 1990, pp~ 178-192.
[3] Lenkey, Gy. B., Winkler, S., "On the Applicability of the Magnetic Emission
Technique for the Determination of Ductile Crack Initiation in Impact Tests,"
Fatigue and Fracture of Engineering Materials and Structures, Vol. 20, No. 2,
1997, pp. 143-150.
[4] Blumenauer, H., Push, G., Technische Bruchmechanik, Deutscher Verlag fOr
Grundstoffindustrie, Leipzig, 1982.
[5] Ireland, D. R., "Critical Review of Instrumented Impact Testing," Proceedings,

International Conference on Dynamic Fracture Toughness, London, 1976, pp. 47-
57.
[6] Kalthoff, J. F., "Concept of Impact Response Curves," ASMHandbook, Volume 8,

ASM, 1995, pp. 269-271.
Mikhail A. SokolovI and John G. Merklet
Estimation of NDT and Crack-Arrest Toughness from

Charpy Force-Displacement Traces
Reference: Sokolov, M. A. and Merkle, J. G., "Estimation of NDT and Crack-Arrest

Toughness from Charpy Force-Displacement Traces," Pendulum Impact Testing: A
Century of Progress, STP 1380, T. A. Siewert and M. P. Manahan, Sr., Eds., American
Abstract: A force-displacement trace of a Charpy impact test of a reactor pressure vessel

(RPV) steel in the transition range has a characteristic point, the so-called "force at the
end of unstable crack propagation," Fa. A two-parameter Weibull probability function is
used to model the distribution of the F, in Charpy tests performed at Oak Ridge National
Laboratory (ORNL) on different RPV steels in the unirradiated and irradiated conditions,
These data have a good replication at a given test temperature, thus, the statistical analysis
was applicable. It is shown that when temperature is normalized to T~T , using
(T - TsDT), or to T100,using (T - T~0o.),the median F, values of different RPV steels have
a tendency to form the same shape of temperature dependence. Depending on
normalization temperature, T~roTor T100~,the results suggest a universal shape of the
temperature dependence ofF, for different RPV steels. The best fits for these temperature
dependencies are presented. These dependencies are suggested for use in estimation of
nil-ductility transition (NDT) or T100.from randomly generated Charpy impact tests. The
linear least squares and maximum likelihood methods are used to derive equations to
estimate TsoT and T~00,from randomly generated Charpy impact tests.
Keywords: Charpy, crack-arrest, nil-ductility transition, instrumented trace, Weibull

statistics.
Introduction
A force-displacement trace for a Charpy impact test of a reactor pressure vessel

(RPV) steel in the transition range has a characteristic point, the so-called "force at the
end of unstable crack propagation," Fa. The end of unstable crack propagation in a
Charpy impact test is an event that is somewhat common in nature with events of interest
in the drop-weight nil-ductility transition (NDT) and the crack-arrest fracture toughness
tests. This similarity was used to find correlations between Charpy Fa and NDT
temperature (T~ror) and a reference crack-arrest fracture toughness temperature (T100,).
1Metals and Ceramics Division, Oak Ridge National Laboratory, P.O. Box 2008,
Oak Ridge, TN 37831-6151.
382
SOKOLOV AND MERKLE ON CHARPY FORCE-DISPLACEMENT TRACES 383
T100, is defined as the temperature at median crack-arrest fracture toughness equal to

100 MPa,/'m when the crack-arrest fracture toughness data are analyzed by the master
curve procedure. For purposes of this study, it was assumed that the temperature
dependence of crack-arrest toughness could be described by the master cui-ve equation
and the maximum likelihood estimator from Ref. [1] could be used to determine T100a.
This assumption received a strong empirical support in the recent work [2] published by
K. Wallin.
Search for a correlation between Charpy crack-arrest force and NDT or crack-arrest
toughness has been under way for many years, see for example [3-8]. The main idea of all
these studies was to establish a characteristic crack arrest force from the instrumented
Charpy trace at a given level of force, for example 0 kN, or 3 kN, or 4 kN, etc., and, then
to determine a correlation between temperature at this characteristic force and the
nil-ductility transition temperature (or crack-arrest toughness transition temperature if
such data were available). In the present work, an attempt is made to eliminate the
intermediate step of determining a characteristic force, and to normalize the temperature
dependence of Charpy crack-arrest force directly to the T~T or T100,by establishing a
universal shape of the Charpy crack-arrest temperature dependence.
Characteristic Forces of the Charpy Force-Displacement Traces
A typical force-displacement trace from an instrumented Charpy test in the transition

region is presented in Fig. 1. The force at the point of general yield, Fgy,is a characteristic
value of the onset of fuUy plastic deformation across the entire uncracked ligament and it
is determined as the force at the intersection of the linear rising portion of the force-
displacement trace and the fitted curve through the oscillations of the force-displacement
trace following the onset of the measurable plastic strain to maximum force. The
maximum force, Fro, is the largest force in the course of the force-displacement trace and it
is determined as the maximum value on the fitted curve through the oscillations of the
force-displacement trace in the area of the maximum force. The force at the initiation of
unstable crack propagation, Flu, characterizes the starting point of the unstable crack
propagation and it is determined as the force at the intersection of the fitted curve through
the oscillations and the rapidly dropping part of the force-displacement curve. The force
at the end of unstable crack propagation, F,, characterizes the point of crack arrest and it
is determined as the force at the intersection of the rapid drop of the force-displacement
curve and the fitted curve through the oscillations of the subsequent part of the force-
displacement curve. Figure 2 illustrates Flu and Fa determinations from the corresponding
parts of the force-displacement trace from Fig. 1.
Analysis Procedure and Results
The Weibull probability function proves to be a powerful tool to model the scatter in
mechanical properties, and, especially, the scatter in the transition fracture toughness. In
fact, recent progress in application of the Weibull statistics to the static transition range
fracture toughness resulted in approval of the ASTM Standard E1921 and publishing the
technical supporting document for it [9]. The master curve methodology inspired the
384 PENDULUMIMPACTTESTING: A CENTURYOF PROGRESS
20 i i i i
I specimen:RW-195
temperature: 2~ I
16
12
Z
v
ID
~ 8
i1o
~ see the nextfigure
I I I I I
0 2 4 6 8 10
Displacement (mm)
Fig. 1-Force-displacement trace showing the definition of the various force points.
20 , ,
j Fiu
16 Q
q,
z 12
o
u. 8 F/
I I
3 4
Displacement (mm)
Fig. 2-Part of the force-displacement trace showing the determination of the
F;u and Faforces
authors to apply a similar approach to describe the characteristic forces of the Charpy
traces.
The two-parameter Weibull probability function is suggested to model the scatter of
the F, in Charpy tests:
P[F.
, - - ~F~=
~, 1 -exp - i~ (1)
where P is the cumulative probability that the force at the end of unstable crack
propagation (Fu) is equal to or less than the F~ value of interest: F o and b are the scale and
the slope parameters of the Weibull distribution function, respectively. However for a
typical series of Charpy tests, often just one test result is available at any given
temperature. Fortunately, the fracture mechanics group at ORNL has been involved in
several projects which resulted in extensive Charpy characterization of several materials.
These data have a good replication at a given test temperature, thus, the statistical analysis
could be applied.
Forty-four specimens o f a Linde 80 weld were tested at 1.7~ (35 ~ as part of the
ASTM reconstitution round-robin [10]. Having such a large replicated data set, Charpy
force-displacement traces were evaluated in terms of Fa values and the Weibull function,
Eq. (1), was used to model the scatter in these values. Figure 3 presents the Weibull
distribution function fitting to the Fa data. The scale parameter, Fo, is illustrated in Fig. 3
as an F~ value at ln{ln[1/(1-P)]}= 0. Knowing parameters Fo and b allows one to
determine the median F a value, Fa (m~d),for this data population as an F, at P = 0.5.
One of the interesting observations is that the slope of the regression line, parameter b in
Eq. (1), is determined to be equal to four. The same value of coefficient b was determined
for large sets of static initiation fracture toughness data [9]. Based on these results, the
value of the Weibull slope in Eq. (1) was fixed at b = 4.
Universal Shape of the Temperature Dependence of the Force at Crack Arrest
The two-parameter Weibull probability function with b = 4 was applied to model the
distribution of the F a in Charpy tests performed at ORNL on some RPV welds in the
unirradiated and irradiated conditions [11-12]. These data have a good replication at a
given test temperature. Thus, the temperature dependence ofF~ can be established.
Table 1 summarizes the Charpy crack-arrest force scale parameter, Fo, for the data
analyzed. Figures 4 and 5 present Fo values versus temperature, T, normalized to TI,~T
and Tloo, temperatures, namely T'TNBT and T-T100,. It is shown that F o values for different
RPV welds have a tendency to follow the same general temperature dependence. This
result suggests a universal shape of the temperature dependence o f F o for different RPV
steels:
2
LINDE 80 WELD
Ttest = 35~
_
Facmed}= 8 kN
!
T'-
v
c-
-2 --
r
O
-3-
P = 1-exp[-(Fa/Fo )b]
b=4
Fa = 8.8 kN
-4-
-5 I I I
0 I 2 3 4
I n ( F a)
Fig. 3-Weibull plot o f F data from Linde 80 weld tested at 1.7~ (35~
SOKOLOV AND MERKLE ON CHARPY FORCE-DISPLACEMENTTRACES 387
Table 1 - - Summary of the force at crack-arrest scale

parameter values, F~ for materials analyzed
Ma~al T - TNDT T Tlooa

-
Number Fo
(~ (~ of (~)
specimens
analyzed
HSSI Weld 72W, unirradiated 44 34 10.9

19 9 4.1
-1 -11 4.7
-9 -19 2.2
-25 -35 1.2
HSSI Weld 72W, 1.5 x 1019 n ] c m 2 -18 -44 2.4
HSSI Weld 73W, unirradiated 37 17 9.0

10 -10 5.5
-8 -28 2.3
-34 -54 0.9
HSSI Weld 73W, 1.5 x 1019 n / c m z 29 5 9.4
-18 -42 7.6
-4 -20 3.6
Midland beltline weld, unirradiated 70 41 13 14.0

45 16 20 10.6
20 -9 16 4.2
-5 -34 11 1.7
Midland nozzle course weld, 62 N/A* 12.7
unirradiated 40 N/A 9.1
15 N/A 4.2
Linde 80 weld, unirradiated. 31 N/A 44 8.8
*N/A - not available.
Fig. 4-Force at arrest scale parameter, Fe versus normalized temperature, T- TNDz
Fig. 5-Force at arrest scale parametr, Fe versus normalized temperature, T- T~ooa

and
Fo= [2.46+ 0 . 0 2 7 . ( T - T100~)]2 (3)
where the scale parameter, Fo, is in kN and T, T~T , and T100a are in ~ According to
Eq. (2), Fo equals 3.5 kN at T = TNm and 6.0 kN at T = T100,. Consequently, at
temperature T = T~roTthe median force at crack arrest is equal to 3.2 kN and at T = Tlo0a
the median force at crack arrest is equal to 5.5 kN.
A fiend for the forces at crack arrest to form a universal temperature dependence
shape in the form of Eqs. (2) and (3) was initially reported by one of the authors in
Ref. [13] and, later, was independently confirmed by Fabry [7]. In the latter case, an
exponential function was used for fitting of the temperature dependence. In the present
case, however, a parabolic function provided a better least-squares fit than an exponential
function. It is interesting to note that later Fabry [8] also applied the Weibull statistic and
came up with about the same median values of the force at crack arrest.
Estimation of T~T and Tlooa from Randomly Generated Charpy Curve Tests
One can perform replicated Charpy tests at a single temperature to estimate T~T
and/or T100aby combining Eqs. (1), (2), and/or (3). However, in practical situations, only
limited numbers of Charpy specimens are randomly tested in the transition region. In such
cases, the linear least squares method can be applied to estimate transition temperatures of
interest. Equations (2) and (3) can be simplified and linearized as:
A+ (4)
where A and B are fitting coefficients and Tx is the transition temperature of interest.
Setting the partial derivative of the sum of the squares of the estimating errors for the
measured data, with respect to the temperature of interest, equal to zero produces an
equation for the parameter being sought, namely:
=o (5)
from which the following expression is derived:

N N
A i=l i=1
(6)
TX=-B ~ N NB
Letting
N N
i=l /=1
(7)
N -Ta" ' NB B
and using coefficients A and B from Eqs. (2) and (3) provides the following expressions
for T~x:
and Tloo.:
T~00a = 91+ T ~ - 3 7 ( ~ a ) ~ (9)
where T~ is the average of the test temperatures (~ and (r is the average of the
square roots of the forces at the end of unstable crack propagation (kN). Thus, Eqs. (8)
and (9) provide a simple solution for estimating Tr~r and Tlooafrom any randomly
generated set of Charpy data. Note that Eqs. (8) and (9) clearly indicate the following
relationship between TI~DTand Tlooa:
TN T = T,00o- 21, (~ (lO)
In addition to the linear least squares method, the method of maximum likelihood can be
applied to estimate transition temperatures of interest in the same way as in the master
curve methodology [1,9]. This method provides a more complicated solution than
Eq. (6)i however, this method proved to be more suitable for the cases when censoring is
involved because this method does not require the cumulative probability distribution
itself. Instead, it uses the first derivative, namely, the probability density function:
dP bFob-'-(F~
9 e ~F~) (1 1)
f(Fa)- dFa - Fob
The product of all the discrete probability densities for the measured Values is a joint
probability function and is called the Likelihood Function [9], L;
L= 1-~ (bF~b-1)Fo-be- (12)

i=1
The logarithm of the Likelihood Function is a sum of logarithms, and setting its partial
derivative, with respect to the parameter of interest, equal to zero produces an equation
for the parameter being sought. In the present case, that is:
81nL 8 Fo
- 0 (13)
Thus, by using Eq. (2), (3), (12), and taking into account that parameter b is fixed,
Eq. (13) becomes
N
= 0 (14)
i=l
from which Tx, representing TNDTand T~00a, can be solved iteratively.
Summary
It is shown, that when temperature is normalized to T~a~T,using (T - T~T), or to

Tlo0a, using (T - T100,),then the median Fa values for different RPV steels have a tendency
to form the same shape of temperature dependence. Depending on the normalization
temperature, T~,~Tor T100a,this result suggests a universal shape of the temperature
dependence ofF a for different RPV steels. The best fits for these temperature
dependencies are presented. An equation is derived using the linear least squares method
for the estimation of T~T or T~00afrom randomly generated Charpy impact test. It is
shown that the maximum likelihood method can also provide the estimation of T~T or
T100a.
Acknowledgments
This research is sponsored by the Division of Engineering Technology, Office of

Nuclear Regulatory Research, U.S. Nuclear Regulatory Commission, under Interagency
Agreement DOE 1886-N695-3W with the U.S. Department of Energy under contract
DE-AC05-96OR22464 with Lockheed Martin Energy Research Corporation.
References
[1] WaUin, K., "Validity of Small Specimen Fracture Toughness Estimates Neglecting
Constraint Corrections," Constraint Effects in Fracture: Theory and Applications,
ASTMSTP 1244, M. Kirk and A. Bakker, Eds., American Society for Testing and
Materials, 1995, pp. 519-537.
[2] Wallin, K., "Application of The Master Curve Method to Crack Initiation and Crack
Arrest," Proceedings of the 1999 ASME Pressure Vessels and Piping Conference,
PVP-Vol. 393, American Society of Mechanical Engineers, New York, 1999,
pp. 3-9.
[3] Vakoukis, G., "Fraktographie und Analyse der Kerbschlagbiegeversuchs" (Doctoral
Dissertation), Universitat Stuttgart, 1971 (in German).
[4] Schoch, F.-W., "Eigenschaften formgeschweisster Grossgauteile:
Werkstoffuntersuchungen an einem 72 t Versuchskorper aus Schweissgut
10MnMoNi55" (Doctoral Dissertation), Universitat Stuttgart, 1984 (in German).
[5] Ahlf, J., Bellmann, D, Fohl, J., Hebenbrock, H., Schmitt, F. J., and Spalthoff, W.,
"Irradiation Behavior of Reactor Pressure Vessel Steels from the Research Program
on the Integrity of Components," Radiation Embrittlement of Nuclear Reactor
Pressure Vessel Steels: An International Review (Second Volume), ASTM STP 909,
L. E. Steele, Ed., American Society for Testing and Materials, Philadelphia, 1986,
pp. 34-51.
[6] Wallin, K., "Descriptive Potential of Charpy-V Fracture Arrest Parameter with
Respect to Crack Arrest Ku," VTT-MET B-221, Metals Laboratory, Technical
Research Centre of Finland, Espoo, Finland, January 1993.
[7] Fabry, A., "Characterization by Notched and Precracked Charpy Tests of the
In-Service Degradation of Reactor Pressure Vessel Steel Fracture Toughness,"
SmallSpecimen Test Techniques, ASTMSTP 1329, W. R. Corwin, S. T. Rosinski,
and E. van Walle, Eds., American Society for Testing and Materials,
West Conshohocken, Pa., 1998, pp. 274-297.
[8] Iskander, S. K., Nanstad, R. K., Sokolov, M. A., McCabe, D. E., Hutton, J. T., and
Thomas, D. L., "Use of Forces from Instrumented Charpy V-Notch Testing to
Determine Crack-Arrest Toughness," Effects of Radiation on Materials: 18th
International Symposium, ASTMSTP 1325, R. K. Nanstad, M. L. Hamilton,
F. A. Garner, and A. S. Kumar, Eds., American Society for Testing and Materials,
West Conshohocken, Pa., 1999, pp. 204-222.
[9] Merkle, J. G., Wallin, K., and McCabe, D. E., "Technical Basis for an ASTM
Standard on Determining the Reference Temperature, To, for Ferritic Steels in the
Transition Range," NUREG/CR-5504, ORNL/TM-13631, Oak Ridge National
Laboratory, Oak Ridge, Tenn., November 1998.
[10] Onizawa, K., van Walle, E., Nanstad, R. K., Sokolov, M. A., and Pavinich, W.,
"Critical Analysis of Results from the ASTM Round-Robin on Reconstitution,"
Small Specimen Test Techniques, ASTM STP 1329, W. R. Corwin, S. T. Rosinski,
and E. van Walle, Eds., American Society for Testing and Materials,
West Conshohocken, Pa., 1998.
[11] Nanstad, R. K., Haggag, F. M., McCabe, D E., Iskander, S. K., Bowman, K. O.,
and Menke, B. H., "Irradiation Effects on Fracture Toughness of Two High-Copper
Submerged-Arc Welds, HSSI Series 5," NUREG/CR-5913, ORNL/TM- 12156/V1,
Vol. 1, Oak Ridge National Laboratory, Oak Ridge, Tenn., 1992.
[12] Nanstad, R. K., McCabe, D. E., Swain, R, L., and Miller, M. K., "Chemical
Composition and RT~qDTDetermination for Midland Weld WF-70,"
NUREG/CR-5914, ORNL-6740, Oak Ridge National Laboratory, Oak Ridge,
Tenn., 1992.
[13] Corwin, W. R. "Heavy-Section Steel Irradiation Program: Semiannual Progress
Report for October 1995 to March 1996," NUREG/CR-5591,
ORNL/TM-11568/V7&N1, Oak Ridge National Laboratory, Oak Ridge, Tenn.,
1997.
[14] Fabry, A., "Nuclear Reactor Pressure Vessel Integrity Insurance by Crack
Arrestability Evaluation Using Loads from Instrumented CVN Tests," Report
BLG-750, SCK-CEN Mol, Belgium, October 1997.
STP1380-EB/May 2000
Author Index
B M
Boehmert, J., 354 Manahan, M. P., 3, 181, 223, 286

B6hme, W., 327 Marsh, F. J., 3
Martin, F. J., 223
C McCowan, C. N., 3, 73, 134, 210
Merkle, J. G., 382
Chai, G., 100 Morita, S., 198
Chaouadi, R., 146 Mougin, D., 109
Charpy, G., 46
N
D
Nakano, H., 73, 164
Dixon, B. F., 310
P
F
Pauwels, J., 73, 90
Fabry, A., 146 Puzzolante, J.-L., 146
G
R
Galban, G., 109
Gyppaz, D., 90 Revise, G., 73, 109
Richter, H., 354
H Russell, S. B., 17
Ruth, E. A., 3
Holt, J. M., 3
Hughes, R, K., 310
S
I
Schindler, H.-J., 327, 337
Ingelbrecht, C., 90 Shibuya, T., 253
Inoue, N., 198, 253 Siewert, T. A., 3, 134, 210
Sokolov, M. A., 382
K Stonesifer, R. B., 181, 223
Sundqvist, M., 100
Kishimoto, K., 253
Kobayashi, T., 198
T
L Takagi, S., 164
Toda, H., 198
Laporte, S., 109
LefratNois, S., 109 V
Lenkey, G. B., 366
Loibnegger, F., 267 Valo, M., 354
Lucon, E., 146, 242 Van Walle, E., 146
395
Copyright9 by ASTMlntcrnational www.astm.org
Y
Varga, T., 267
Varma, R., 90
Viehrig, H.-W., 354 Yamaguchi, Y., 164
Vigliotti, D. P., 134, 210
W Z
Waid, G. M., 298 Zantopulos, H., 298
STP1380-EB/May 2000
Subject Index
A Charpy specimen certification,

73, 90
Charpy testing, 286, 354, 382
Absorbed energy, 3, 134, 164, Charpy type specimen, 267
181, 286 Charpy v-notch, 100, 134, 298
Accreditation procedures, 109 specimens, 109, 242, 267,
Acoustic emission, 354 286, 366
Acrylonitrile-butadiene-styrene specimens, miniaturized, 223
plastic, 210 specimens, subsize, 146
Aging, 210 testing machines, 134
stability, 100 Clamping pressure, 210
American Petroleum Institute, Cleavage, 327
298 COFRAC, 109
Anvil configurations, 164 Computer Aided Instrumented
Anvils, worn, 134 Impact Testing System,
ASTM Committee E28 on 198
Mechanical Testing Crack extension, 327
E28.07.07 Crack growth initiation, 327
ASME lower-bound curve, 337 Crack initiation, ductile, 354
ASTM standards, 3 Crack resistance curve, 327
E 23, 223 Crack tip opening displacement,
267
B C-type impact machine, 100
Curvature, specimen, 298
Bend test, metals impact, 46
Bridges, impact test, 3 D
Brittle fracture load, 181
Bureau Communautaire de Deconvolution, 253
Reference, 90 Dual energy determination, 181
Butadiene Ductile-brittle transition, 310
acrylonitrile-butadiene-styrene Ductile crack initiation, 354
plastic, 210 Ductile-to-brittle transition
range, 337
C Ductility, 181
Calibration, 100
characterization of calibrated
batches, 109 Elastic resilience, 17
dynamic force, 253 Electron beam weld, 310
load, 198 Energy, absorbed, 3, 134, 164,
Casing, American Petroleum 181,286
Institute Specification for Energy curve shift, 146
Casing and Tubing, 298 Energy divergence, 100
Charpy fracture energy, 337 Energy/lateral expansion/shear
Charpy impact testing, 210, 310 fracture levels, 146
history, 3, 17, 46 Energy levels, 164
instrumented, 198, 253, 327 Energy, variation, 73
397
European standards, 90, 109, 242

European Structural Integrity
Society, 242 Japanese V-notch reference test
specimens, 164
J-integral, 354
JR curve, 327
Failure, metal resistance to, 46 J-resistance curve, 327
Finite element analysis, 198
Force-deflection-diagram, 267 K
Force displacement trace, 382
Fracture path, 310 Key-curve method, 327
Fracture toughness, 198, 267,
327, 337 L
dynamic, 354, 366
impact, 253 Laboratorie National D'Essais,
French Accreditation 73
Committee, 109 Linear least square, 382
French standards NFA 03-508, Liquid bath transfer approach,
109 286
French Steel Federation, 109 Load, 17, 253
calibration, load signal, 198
H load vs. displacement curve,
354
Heat loss, 286 Loading rates, 267, 366
History, impact test, 3, 17 Loading velocity, 337
M
Magnetic emission
measurement, 366
Impact response curve method, Metals impact bend test, 46
366 Microstructure analysis, 100
Institute for Reference Materials Miniaturized Charpy testing,
and Measurements, 73, 90 223
Instrumented impact testing, Misalignment, 134
223, 242, 366
correlation procedures, 146 N
fracture toughness, 253
history, 3, 17 National Institute of Standards
load signal, 198 and Technology, 73, 223
single specimen methods, 327 Nickel-based alloy, 100
Instrumented/Miniaturized Nil-ductility transition, 382
Charpy Round Robin Normalization factors, 146
Test Program, 223 Notched bars, 46
Interlaboratory comparison, 210, Notch sharpness, 337
223, 242, 327 NRLM, 73
for certification, 90
international, 73 O
International Organization for
Standardization Oak Ridge National Laboratory,
striker, 100 223, 382
Izod impact testing, 210 Optical encoder, 181
INDEX 399
Optical measurement, T
simultaneous, 354
Tensile strength, 337
P ultimate, 242
Thermal conditioning, 286
Peak load, 181 Thermocouples, 286
Plastic Toughness, 90, 209
acrylonitrile-butadiene- fracture, 198
styrene, 210 Transition behavior, 267
Plastic impact specimens, 210 Transition region, 223, 337
PMMA, 253 Transition temperature, 267, 310
Polymethyl methacrylate, 253 Tubes, high strength steel, 298
Pressure vessels, impact tests, 3
R U
Railway bridges, 17
Reactor pressure vessel, 146, Uncertainty propagation, 164
223, 382 Upper shelf energy, 146, 223, 310
Reference materials, 73, 90
Reference test pieces, 109, 164
Resilience, elastic, 17 V
S Velocity, impact, 267

Verification testing, 90, 134
Scatter, 310 interlaboratory comparison,
SCK.CEN, 146 73, 242, 327
Stability analysis, 100 pendulum impact testing
Steel, 146, 267, 366 machines, 109
4340, 73, 90, 298 plastic impact, 210
A533B, 223 programs, 73, 223, 242
Frertch steel accreditation, specimens, plastic impact, 210
109 Vibration, impact testing, 134
high strength, 310 effect removal, 253
reactor pressure vessel, 146,
382
Strain localization, 198 W
Strain rates, 242
Stress, change of, 17
Stress intensity factor, 366 Wear, impact testing, 134
dynamic, 253 Weibull statistics, 382
Stretch zone width, 354 Weld, 310
Strikers, 90 W-Nr 1.4563, 100
instrumented, 164, 181, 198, Worn anvils, 134
223, 286, 298
Structural performance,
correlation to impact Y
testing, 3
Styrene
acrylonitrile-butadiene- Yield load, 181
styrene plastic, 210 Yield strength, 242
H
VJ
W
Z
D
!
0~
D
W
~J
!
nJ
C~
DI
!
n

Michael P. Monahan (Editor) Thomas A. Siewert (Editor) - Pendulum Impact Testing - A Century of Progress (ASTM Special Technical Publication, 1380) (2000)

Uploaded by

Copyright:

Available Formats

Michael P. Monahan (Editor) Thomas A. Siewert (Editor) - Pendulum Impact Testing - A Century of Progress (ASTM Special Technical Publication, 1380) (2000)

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

Michael P. Monahan (Editor) Thomas A. Siewert (Editor) - Pendulum Impact Testing - A Century of Progress (ASTM Special Technical Publication, 1380) (2000)

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What are some topics covered in the book?

What are some topics covered in the book?

STP 1380

Pendulum Impact Testing:

Thomas A. Siewert and Michael P. Manahan, editors

ASTM Stock Number: STPI380

Printed in the U.S.A.

Pendulum impact testing : a century of progress / Thomas A. Siewert and Michael R

Authorization to photocopy items for internal, personal, or educational classroom use, or

Peer Review Policy

The History and Importance of Impact TestingmT. A. SIEWERT,

Experiments with a New Machine for Testing Materials by I m p a c t -

REFERENCE ENERGIES, MACHINE STABILITY, AND CALIBRATION

International Comparison of Impact Verification Programs---c. N. McCOWAN,

European Certification of Charpy Specimens: Reasoning and Observationsm

Stability of a C-type Impact Machine Between Calibrations--M. SUNDQVIST

Indirect Verification of Pendulum Impact Test Machines: The French

Maintaining the Accuracy of Charpy Impact Machines--D. e. VIGLIOTTI,

Characterizing Material Properties by the Use of Full-Size and Subsize

The Difference Between Total Absorbed Energy Measured Using an

On the Accuracy of Measurement and Calibration of Load Signal in the

Evaluation of ABS Plastic Impact Verification Speeimensnc. N. McCOWAN,

Results of the ASTM Instrumented/Miniaturized Round Robin Test

European Activity on Instrumented Impact Testing of Subsize Charpy

Dynamic Force Calibration for Measuring Impact Fracture Toughness using

Low Striking Velocity Testing of Precracked Charpy-type Specimens--

In-Situ Heating and Cooling of Charpy Test SpeeimensnM. P. MANAHAN,SR. 286

The Effects of OD Curvature and Sample Flattening on Transverse Charpy

Electron Beam Welded Charpy Test Specimen for Greater Functionality--

Application of Single-Specimen Methods on Instrumented Charpy Tests:

Relation Between Fracture Toughness and Charpy Fracture Energy:

Use of Instrumented Charpy Test for Determination of Crack Initiation

Estimation of NDT a n d Crack-Arrest Toughness from C h a r p y Force-

The future of pendulum impact testing appears bright, as it continues to be specified in

Michael P. Manahan, Sr.

The History and Importance of Impact Testing*

Reference: Siewert, T. A., Manahan, M. R, McCowan, C. N., Holt, J. M., Marsh,

1 Supervisory Metallurgist, Materials Reliability Division, National Institute of

Impact Testing: 1824 to 1895

- limits for the velocity of the striker,

1922 to 1933: The Beginning of ASTM Method E 23

ASTM Committee E-1 on Methods for Testing sponsored a Symposium in 1922 on

9 .... : : ' ' _ , .... ,

Since it is impractical to measure the fracture toughness of large specimens throughout

0 t ,~r-[ I~_L,~ I, I, I, I, '0 0

Figure 3 - Distribution of low-energy verification data. Data for

Figure 4 - Distribution of high energy verification data. Data for

150 ' I ' I ' I i

0 '~J-~ ~ L ' 0.0

Figure 5 -Distribution of the super-high energy verification data.

Continuing Standardization Efforts

The ASTM E 23 standard is a document that continues to improve as our technical

[1] Tredgold, T., Strength of Castlron, 1824, pp. 245-268.

[10] Whittemore, H. L., "Resume of Impact Testing of Materials, with Bibliography,"

[11] Warwick, C. L., "Resume on Notched Bar Tests of Metals," Proceedings of

[13] Williams, M. L. and Ellinger, G. A., Investigation of Fractured SteelPlates

Experiments with a New Machine for Testing