STP1318 Eb.1415051 1 PDF

Copyright by ASTM Int'l (all rights reserved); Sat Jan 28 08:00:22 EST 2012
Downloaded/printed by
(PDVSA Los Teques) pursuant to License Agreement. No further reproductions authorized.
S T P 1318
Nontraditional Methods of Sensing

Stress, Strain, and Damage in
Materials and Structures
George F. Lucas and David A. Stubbs, Editors
ASTM Publication Code Number (PCN):

04-013180-30
IITI'" 100 Barr Harbor Drive

West Conshohocken, PA 19428-2959
Printed in the U.S.A.
Library of Congress Cataloging - in - Publication Data
Nontraditional methods of sensing stress, strain, and damage in materials and structures
/ George F. Lucas and David A. Stubbs, editors.
p. cm. -- (STP : 1318)
Includes bibliographical references and index.
ISBN 0-8031-2403-1
1. Strains and stresses--Measurement. 2. Detectors. I. Lucas, George F., 1951-
II. Stubbs, David A., 1958- III. Series: ASTM special technical publication ; 1318.
TA417.6.N64 1997
620.1' 123'0287--dc21 97-18165
CIP
Copyright 9 1997 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, Dan-
vers, MA 01923; Tel: (508) 750-8400; online: http:llwww.copyright.com/.
Peer Review Policy

Each paper published in this volume was evaluated by two peer reviewers and at least one
of the editors. 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 these peer reviewers. The ASTM
Committee on Publications acknowledges with appreciation their dedication and contribution
to time and effort on behalf of ASTM.
Printed in Scranton, PA
August 1997
Foreword
This publication, Nontraditional Methods of Sensing Stress, Strain, and Damage in
Materials and Structures, contains papers presented at the symposium of the same name,
held on 20 May 1996. The symposium was sponsored by ASTM Committee E-8 on Fatigue
and Fracture. George F. Lucas of MTS Systems Corporation in Eden Prairie, Minnesota and
David A. Stubbs of The University of Dayton in Dayton, Ohio presided as symposium
chairmen and are editors of the resulting publication.
Contents
O v e r v i e w - - G . F. LUCAS vii
NONTRADITIONAL EXTENSOMETERS
Use of an Optic Lever as a Short Gage Length Extensometer--G. M. HOUSER

AND D. E. WALRATH
New Applications of the Interferometric Strain/Displacement Gage--

W. N. SHARPE, JR. AND B. YUAN 16
An Extensometer for Soft Material Strain Measurement--N. SHRIVE,

E. DAMSON, C. FRANK, R. MEYER, L. MEJIA, AND G. LUCAS 29
NONTRADITIONAL CRACK MEASUREMENT TECHNIQUES
Short Fatigue Crack Characterization and Detection Using Confocal

Scanning Laser Microscopy ( C S L M ) - - A . VARVANI-FARAHANI AND
T. H. TOPPER 43
An Advanced Multiple Frequency ACPD System for Crack Detection and

Calibration--s. TIKU AND N. J. MARCHAND 56
Observation on the Growth Behavior of Small Surface Cracks Using Remote

Measurement System--s. H. NAHM, AND C. M. SUH 71
Use of Portable/In-Situ Stress-Strain Microprobe System to Measure Stress-

Strain Behavior and Damage in Metallic Materials and Structures--
F. M. HAGGAG, J. A. WANG, M. A. SOKOLOV, AND K. L. MURTY 85
OPTICAL, NoN-CONTACTING STRAIN MEASUREMENT DEVICES
BISMAP: A Non-Contact Biaxial Displacement Measurement System--

P. C. McKEIGHAN, S. B. SEIDA, E. A. FRANK, AND D. L. DAVIDSON 101
Optical Measurement of Biaxial Strain in Thin Film Polymers--J. L. RAND AND

D. A. GRANT 123
A Review of Strain Measurement Techniques Using the Grid Method--

J. F. CARDENAS-GARCIA, M. M. WU, AND J. HASHEMI 138
New Developments in Full Field Strain Measurements Using Speckles--

F. P. CHIANG, Q. WANG, AND F. LEHMAN 156
ULTRASONIC AND INFRARED TECHNIQUES
Ultrasonic Longitudinal and Surface Wave Methods for In Situ Monitoring of

Dmage in Metal Matrix and Ceramic Matrix Composites--
D. J. BUCHANAN, R. JOHN, D. A. STUBBS, D. M. BENSON, AND P. KARPUR 173
Using Ultrasonic Measurements and a Two-Phase Composite Model to Assess

Radiation Damage in Reactor Pressure Vessel Steels--J. A. WANe 187
An Array Measurement System for Thermoelastic Stress Analysis--c. WELCH,

K. E. CRAMER, J. LESNIAK, AND B. BOYCE 198
Stress Intensity Measurement via Infrared Focal Plane ArraynJ. LESNIAK,B.

BOYCE, K. E. CRAMER, AND C. WELCH 208
Thermoelastic Stress Analysis of the Human TibianK. HYODO,M. YAMADA,AND

T. TATEISH1 221
Overview
The understanding of how materials and structures behave under conditions

that stress, deform, or damage them has been of concern since the dawn of
engineering. Real advancement in these areas can not be achieved, however,
without sensing devices that can measure accurately and repeatably, the various
properties of these materials and structures. The advancement of methods that can
sense basic materials properties (such as stress and strain) as well as other "damage"
parameters (such as flaw density and type) are critical development of methods of
designing better products for the future, whether these products better airplanes,
better ground vehicles, better consumer goods, or better artificial devices for use
within the human body.
This STP is intended for use of the research and development community that
needs to generate and use better information of how materials and structures
behave. It is intended for individuals working in areas of metals and composites, for
those working in the aerospace or the biomedical areas, and for those in academic
research or those in product development functions. How do we answer such
questions as:
What is the stress actual distribution around a stress concentration during cyclic
loading?
How can crack length measurements be made in a material or structure?
How can strain be measured on a soft tissue with enough accuracy to determine
modulus?
How is bi-axial stress measured in thin tissues.
How can full-field strain measurements be made on components and complex
surfaces?
This Special Technical Publication is based on the symposium on Non-

traditional Methods of Sensing Stress, Strain , and Damage in Materials and
Structures. This symposium was held in May of 1996 in Orlando, Florida. The
symposium was preceded by two workshops on the same topic held in November
of 1993 in Fort Worth Texas and in November of 1994 in Phoenix, Arizona. The
Workshops and the Symposium were driven by the work of the task group E8.03.03
on Sensors which is a task group of the E8 committee on Fatigue and Fracture and
it's subcommittee on Advanced Apparatus. The scope of this task group is to
develop standards and encourage technology interchange concerning measurement
sensors that are used in determining fatigue and fracture characteristics of materials
and structures.
The advancement and development a applicable sensors is a critical part to any
understanding of real materials and components design characteristics. You much
vii
viii OVERVIEW
be able to measure a property before you can understand it. The entire
development of modern materials science and engineering has been tied to better
and better sensors to measure properties. For example, the development of strain
gages allowed real determination of the strain of a "spot" on a structure or materials
specimen. This has allowed a much better understanding of how components and
full structures, such as automobiles, airplanes and hip prostheses really behave. It
has allowed for models, such as various Finite Element models, to be verified, or, to
be shown where they were not correct to allow them the be modified.
Other sensors, such as modern extensometers, have allowed much more
accurate measurements of local strain to be measured, and in fact, controlled. This
has allowed strain controlled fatigue tests to be conducted. These tests better allow a
new material's tendency for crack initiation to be understood. Other examples of
sensors include eddy current devices for finding flaws below the surface, infrared
devices for seeing areas where higher strains might be present, and potential drop
techniques for looking a crack growth. All these methods have helped advance the
understanding of how materials and structures behave.
Over the past 10 to 15 years there have also been a lot of disappointments.
Various optical methods for strain measurement have been attempted with mixed
success. It was expected by many, at least 15 years ago, that optical, non contacting
strain measurements methods would replace contacting strain measurement
techniques. We had expected that these methods would allow greater accuracy,
repeatability, and be able to examine both local strains as well as full field strains.
This did not happen for many reasons, and because of this the interest has declined
in developing these devices. However, continual advances in computational
methods, and optical technology may allow these early expectations to eventually be
met.
What else may the future bring? There are many areas for potential new
sensors that can significantly improve our ability to design better products. How
can we measure the damage that occurs in a material or structure before it becomes
apparent as a change in the stress-strain parameters or surface condition? How can
we accurately "see" the full, three dimensional, micro-deformations that a
component experiences. These, as well as other advances sensors, may be possible
in the near future. Tremendous advances in computational capabilities may allow
optical methods to show, in real time, the accurate and repeatable deformations that
occur on a complex component surface. Other advances in miniaturization may
allow real time sensors to be attached to areas of a structure that, in the past were in-
accessible. All in all, there is a tremendous potential for future developments in
sensing technology related to work in fatigue and damage of materials and
structures.
The Special Technical Publication is organized into four sections, each
encompassing papers of similar interest. The first, Nontraditional Extensometers,
is for novel methods of measuring point to point deformation, or when normalized,
strain. Dave Walrath et al explains a technique of using an optical fiber based
sensor in lieu of a normal strain gage sensor. This device is not effected by any
ambient electrical or magnetic field. William Sharpe et al discusses the development
of interferometric displacement gages, which can be used for very point to point
measurement at very small scales, such as at the root of a crack. Nigel Shrive et al,
expounds
Copyright upon Int'l
by ASTM a novel arrangement
(all rights reserved); Satfor
Jan a28contacting
08:00:22 ESTextensometer.
2012 This device is
OVERVIEW ix
able to minimize contact loads to one gram, allowing direct strain measurement of
difficult materials, such as thin wire or biological tissues.
The second section, Nontraditional Crack Measurement Techniques, goes
through methods for measuring crack length, expect for the last paper by Haggag.
Haggag's paper goes through a technique of measuring the stress-strain and
"damage" parameters of the material as part of a structure in situ. As such it is
useful in determining the useful life left in a component.
Varvani-Farahani goes through a method of determining, utilizing a confocal
scanning laser microscope, the morphology of a crack as well as it's crack length.
This is a case where more information than is traditionally is made available. The
paper by Tiku et al goes through an new potential difference technique for crack
length determination that utilizes current focusing. This allows the technique to be
used with large structural applications, where traditional PD methods were limited
to certain specimen geometries. Nahra et al goes through an interesting method of
measuring surface cracks from a remote location utilizing a CCD array with
controllers, software, and a 3-D translation system. This allows crack measurement
to be made on specimens at high temperature or under other adverse conditions, as
long as a clean optical path to the specimen exists.
The third section, Optical, Non-Contacting Strain Measurement Devices, has
four excellent papers on optical methods of measuring both point to point and full
field strains. Seida et al describes a method of adapting current machine vision
technology to strain measurement by tracking the positions of features on the
surface of a specimen. Rand and Grant examine a method of measuring the planar
Biaxial strains in a thin film polymer. Their method uses a random speckle pattern
and a CCD camera. Cardenas-Garcia analyses the grid technique as a general
method of measuring biaxial stain on a simple or complex surface. Chaing et al go
though a examination of full field speckle photography as a method of measuring
strain. This paper is especially interesting as it goes looks at very small scale full
field strain measurements.
The last section, Ultrasonic and Infrared Techniques, some new and promising
techniques are described. These methods do not relay on surface features, so are
able to "see" below the surface of a specimen or component. Buchanan et al
describes a technique to measure the damage within ceramic matrix composites
(CMC) and metal matrix composites (MMC). They utilize bulk and surface
ultrasonic waves to characterize the material in situ while undergoing mechanical
simulation. Wang depicts an ultrasonic method to measure the damage caused by
radiation in a reactor steel.
Welch et al and Lesniak et al both describe different aspects of new commercial
method for thermoelastic stress measurement. This method is able to quickly
analyze measure the reversing stresses caused by temperature variations. Their
method is a evolution of the older SPATE systems. Finally, Hyodo et al describe a
similar method that measures the stresses in a biological specimen, the human tibia.
In summery, there are a great many different emerging techniques to measure
various parameters in materials and structures. Be cautioned that most of these
methods are still in the development stage. Exceptions to this include the above
mentioned thermoelastic system, the extensometer described by Shrive, and the
system described by Haggag. Some of the others will become developed further into
commercial
Copyright devices,
by ASTM others
Int'l (all rightsmay not. Sat Jan 28 08:00:22 EST 2012
reserved);
x OVERVIEW
This symposium certainly shows that the community is potentially on the verge
of a whole host of new sensing tools that may allow us to learn more about the
materials and structures that we can make the products of future from.
The symposium cochairmen would like to acknowledge the efforts of several
individuals, in addition to the individual reviewers, that made this STP possible.
George Hartmann, of UDRI, made many valuable suggestions in the planning
process, as well as chairing one of the workshops. Art Braun, chairman of
subcommittee E8.03 and Joe GaUagher, chairman of E8, both were instrumental in
encouraging this symposium and STP to exist. Finally, Shannon Wainwright of the
ASTM Staff, did a tremendous job of keeping it all together. Her endless
encouragement and gentle nudging made it a pleasure to be involved with ASTM in
editing this STP.
George F. Lucas
MTS Systems Corporation, Eden Prairie, MN, 55424
Symposium cochairman and coeditor
Nontraditional Extensometers
Guy M. Houser, ~and David E. Walrath 2
USE OF AN OPTIC LEVER AS A SHORT GAGE LENGTH EXTENSOMETER
REFERENCE: Houser, G. M. and Walrath, D. E., "Use of an Optic Lever as a Short

Gage Length Extensometer," Nontraditional Methods of Sensing Stress, Strain, and
Damage in Materials and Structures, ASTM STP 1318, George F. Lucas and David A.
Stubbs, Eds., American Society for Testing and Materials, 1997.
ABSTRACT: A short gage length extensometer, based on the principle of an optic lever,
was designed and tested as a medium resolution, short gage-length displacement transducer.
The goal was to develop a short gage-length extensometer as a potential replacement for
foil resistance strain gages. The optic lever used step-index optical fibers to modulate
reflected light as a function of target displacement. This provides the basis for a responsive
displacement (strain) transducer that is small in size, insensitive to electrical noise, and
resistant to elevated temperature and radiation levels. Transducer design was facilitated
through use of an analytical model to predict performance. A prototype transducer was
tested and compared to foil resistance strain gage results. The prototype demonstrated 25
microstrain resolution over a 10% strain range.
KEYWORDS: fiber optic, strain measurement, optic lever
Foil resistance strain gages have been used to measure strain for many years. They
have excellent strain measurement resolution and can be made to small size, thus measuring
average strain on a small area. Foil resistance strain gages are a low cost alternative to a
variety of strain measuring systems. Foil resistance strain gages require careful installation,
and can not be removed and reused. In materials testing, where specimens are typically
destructively tested, the time and expense of mounting strain gages on each test specimen
can become significant, making reusable transducers desirable.
IDirector of Research and Development, Model and Instrument Development,

26296 Olhava Way NE, Bldg. 1, Poulsbo, WA 98370.
2Associate Professor, Mechanical Engineering Department, PO Box 3295,

University of Wyoming, Laramie, WY 82071.
3
Copyright9
(PDVSA by ASTM
Los Teques) pursuant International
to License Agreement.www.astm.org
No further reproductions authorized.
4 SENSINGSTRESS, STRAIN, AND DAMAGE IN MATERIALS
One such strain transducer design is an extensometer, an adaptation of the foil

resistance strain gage. The extensometer consists of foil resistance strain gages mounted on
a c-shaped or curved beam. As the ends of the beam are displaced (extended), the beam is
subjected to bending, producing strain which is detected by the strain gages. The
extensometer is calibrated to measure displacement AL, from which the strain s is computed
by dividing by the gage length L, the original distance between the ends of the c-shaped
beam. That is,
AL
e=L
The extensometer measures average strain in a specimen through the region encompassed
by the gage length. Strain resolution of an extensometer is enhanced by using longer gage
lengths. Typical gage lengths for commercial extensometers range upward from 12.5 mm.
These longer gage lengths are typically not a problem in materials testing where uniform
strain (stress) fields are produced in the test specimens. Shorter gage length extensometers
are more difficult to mechanically attach to a specimen, and have poorer strain resolution.
The desired strain transducer for materials testing would be one with a strain
resolution of 10"~over a gage length typical of common foil resistance strain gages, say 3-4
mm, while being easily installed and reusable. The objective of the work described in this
paper was to develop such a device, based on use of a fiber optic lever [1].
FIBER OPTIC LEVER
The fiber optic lever was described by Frank in 1966, [2] and Kissinger in 1967 [3].
A quantitative description of the principles governing fiber optic levers was presented by
Cook and Hamm in 1979 [4]. The fiber optic lever, shown in Figure 1, is a
Figure 1. Operating Principle of the Fiber Optic Lever.
HOUSER AND WALRATH ON USE OF AN OPTIC LEVER 5
device that is capable of modulating the reflected power from a translating target in such a
manner that it is a function of the target displacement. The basis of operation is that light is
transmitted through a step index optical fiber to illuminate a reflective target. Light exits
the transmitting fiber as a diverging cone. The light cone is incident on a reflective target
that translates parallel to the fiber bundle longitudinal axis, or perpendicular to the ends of
the fibers. Translation of the target modifies the amount &reflected illuminance incident on
receiving fibers surrounding the transmitting fiber. Part of the reflected illumination is
transmitted by receiving fibers to a photodetector. The output of the photodetector is a
function of target displacement. A conceptual fiber optic lever based displacement
measurement system is shown in Figure 2.
Figure 2. Conceptual Model of a Fiber Optic Lever Displacement Measurement System.
The fiber optic lever displacement transducer is a low cost device that is insensitive to
electrical noise. Original applications employed the device as a responsive, small
displacement sensor for use in vibration measurements. Sciences ranging from
biomechanics to automated manufacturing have since utilized the fiber optic lever for
commercial use, see for example [5,6].
SEVEN FIBER OPTIC LEVER MODEL
A mathematical model for calculating the illuminated receiving fiber area was greatly
simplified by assuming that dectroluminescent energy is coupled into lossless, nondispersing
ideal step index fibers. It was further assumed that the reflective target translates in a
direction parallel to the longitudinal fiber axis [4]. Extending this analysis to a seven fiber
model, from the target vantage point one would view the reflected illuminance area on the
fiber plane as shown in Figure 3. The illuminated fiber area of a single fiber can be
multiplied by the number of receiving fibers to compute the total illuminated fiber area.
Figure 3. Illuminated Receiving Fiber Core Area Due to Reflected IUuminance on the Fiber
Plane for a Seven-Fiber Optic Lever.
Figure 4 shows a cutaway view of the 3 middle fibers. The triangle (abd) shows
Figure 4. Reflected IUuminance Radius at the Fiber Plane
that the reflected illuminance area radius q minus the transmitting fiber core radius p is
twice the target distance y times the tangent of the fiber acceptance angle 0g. The reflected
illuminance area radius is
q = (2 y tan 0A) + p.
The illuminated receiving fiber area for a single fiber is the intersection of areas defined by
the reflected illuminance and the receiving fiber core, as shown in Figure 5. The area of
Figure 5. Illuminated Receiving Fiber Area (~/2 < 7d2)
the arc swept by the illuminance radius, an angle 0/2 (/bca), is AI, and by the fiber core
radius, an angle V/2 (/bac), is A2. The area Aa is the intersection of Al and A2 and the
subtraction of excess triangular area (abc). Thus
A3 = A ~ n A 2 = q2
I ( q\(O'S+P2
2J J~ - x ~
-2-L\-2-)
sin(-~)]2
where xo is related to the fiber center spacing as
2(p+ x)
X0-- - -
P
where T is the cladding thickness. By relating qJ/2 and ~b/2as,
- - = sin -~ sin
2
defining the reflected iiluminance ratio such that
8 SENSINGSTRESS, STRAIN,AND DAMAGEIN MATERIALS
_q
P
and multiplying by the number of receiving fibers, the total area illuminated by reflected
illuminance is
as long as W/2 is an acute angle. If ~P/2 is obtuse,
A, = 6p'I)c: ~ + , - sin-'0: sin~l - ~:xo sin~l,
The irradiated fiber area becomes the total fiber core area when the reflected
irradiated radius extends beyond the fiber core or when q > 3(p + x). The area ratio
response plotted as a function of target displacement for a seven-fiber optic lever is plotted
in Figure 6.
1.0
0.9
0.8
0.7
I II
Il Ill
o
,,.q
0.6 I
0.5 I
<~ 0.4
0.3
0.1
0,0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 o.s 0.9 1.0
g~ee~ve TargetDisplacement (mm )
Figure 6. Illuminated Area Ratio as a Function of Target Displacement for a Seven-Fiber

Optic Lever (99 lxm core diameter, 140 ~tm fiber diameter, and numerical
aperture = 0,29)
Three important regions are defined:
Region I - ~/2 is acute; the reflected illuminance begins to illuminate the receiving
fiber cores,
Region n - q~/2 is obtuse; the reflected iiluminance completes illumination of the
receiving fiber cores, and
Region III - q _>3(p + x), the reflected iiluminance is beyond the receiving fibers.
Peak area ratio is observed to be in Region II before the total receiving fiber area
has been illuminated. This occurs when approximately 80% of the receiving fiber core area
has been illuminated. The area ratio can never reach unity due to dead space or no-
receiving area composed of cladding area and the packing fraction area losses between
fibers. The peak area ratio is an important optic lever characteristic, dependent on the
specific optical fiber used.
Radiant power returned by the receiving fibers as a function of target standoff
distance is the integral of reflected illuminance over the illuminated area. However, this
integration is complicated by the fact that illuminance is not constant over the area of the
fiber. Many effects can degrade the ideal transmission of light in an optical fiber, including
dispersion, scattering, microbending, and flaws in fiber end-plane finish [7]. These factors
play an important role in defining the illuminance profile of the transmitted light. In the
present work, the illuminance profile was modeled as
E ~ = cos 7 y - c t s i n y
Ei
where E0 is the reflected illuminance at the fiber plane,

Ei is the transmitted illuminance at the fiber plane,
~, is the angle between the longitudinal fiber axis and a corresponding light ray, and
ct is an adjusting constant.
This procedure was suggested by similar work done by Cook and Hamm [4]. The basic
form of the illuminance profile (the exponential cosine term) was used in [4]. The ctsiny
term is included to model reflection/transmission variations at the fiber interface as
suggested in [8]. The exponent of 7 was determined by experimental measurement of an
illuminance profile. The r coefficient was empirically determined. First a trial value of et
was selected. Then by numerically integrating the reflected illuminance over the illuminated
area (as modeled by the previously described area ratio model), accounting for a variable
illuminance profile, the return illuminance was calculated. The ~ coefficient was then
adjusted to best match the modeled response to an experimentally measured response for a
specific seven-fiber optic lever configuration. A comparison of modeled response and
measured response for a specific seven-fiber optic lever is shown in Figure 7. Some
deviation does exist between measured and modeled response, but overall representation of
the response is good.
10 SENSING STRESS, STRAIN, AND DAMAGE IN MATERIALS
TRANSDUCER DESIGN
The prototype strain transducer consisted of two components, the fiber optic lever
displacement sensor, and the attachment system. Using the response model, effects on
transducer response of different optic fiber parameters, including fiber diameter, cladding
thickness, and numerical aperture, were studied. Results of these parametric studies are
summarized in [1]. Based on the parametric studies, a 230 ~tm diameter, 200 gun core
diameter step index fiber with a numerical aperture of 0.37 was selected. A 100 watt
1.0"
0.9
Mea.sta~~ I
0.8' o ~ AnalytiealResponse
0.7'
9N 0.6'
0.5"
0.4
0.3
Z 0.2
0.1
0.0
0.0 0.5 1.0 1.5 2.0 2.5
ReflectiveTargetDisplacement( m m )
Figure 7. Comparison of Modeled and Measured Fiber Optic Lever Response (99 lain core
diameter, 140 lxm fiber diameter, and numerical aperture = 0.29)
quartz tungsten halogen (QTH) lamp was used for illumination. This lamp has a significant
illuminance spectral range or 300 nm to 700 nm. The selected photo detector had a peak
response at 550 nm. The selected fiber had a low attenuation at wavelengths ranging from
500 to 1100 nm.
A conceptual view of the attachment mechanism is shown in Figure 8. This
mechanism is basically a sliding cylinder within a sleeve, designed to maintain alignment
between the optic lever fiber bundle and the target while at the same time remaining
compact and light weight. Attachment of the transducer to the material test specimen was
accomplished by either of two techniques, similar to those used for other types of
extensometers. The first method was by use of sharpened steel points attached to the
transducer housing. These points were then held against the test specimen by rubber bands.
Slippage of the points was observed when the specimen was subjected to shock or
vibration. The second method used was to bond pins to the test specimen, which could
then be inserted into corresponding holes in the transducer housing. Obviously the bonding
procedure provides a more secure attachment but makes for a much more cumbersome
installation procedure. The prototype strain transducer had an initial gage length of 5 mm
1. NylonSleeveHousing 5. FiberBundleTube
2. FiberPlane 6. FiberBundle
3. ReflectiveTarget 7. AttachmentArm
4. BondedAttachmentPoint 8. ConicalAttachmentPoint
( Optional) ( Optional)
Figure 8. Sliding Cylinder-Sleeve Fiber Optic Lever Strain Transducer, 5 mm Gage Length.
EXPERIMENTAL RESULTS
The prototype optic lever strain transducer was calibrated using a 0.5 lam resolution
micrometer to apply a controlled displacement. Output of the measurement system is
plotted in Figure 9. Note that the response plotted in Figure 9 is the initial slope or positive
slope region of the response curve previously plotted in Figure 7 or Region I of the plot
shown in Figure 6. For calibration purposes, a polynomial representation of this response
was determined using regression analysis. A cubic polynomial was of sufficient order to
represent the response.
o.400 ~
0.350
..~ 0.300
E
0.250
0
0.200
O.150
0.100 "v"-
0.0 |,0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Strain Transducer Response ( Volts )
Figure 9. Cubic Polynomial Linear Regression Representation of Fiber Optic Lever Strain
Transducer Response.
Regression analysis resulted in the cubic representation of
d = 0.11295 + 4.5191x10 "2V + 5.5324x10"3V2 + 4.5065x10"4V3
where d is the displacement and V is the output voltage. The coefficient of determination
was R 2 -- 0.9996. Uncertainty in the regression analysis, as computed by standard
deviation of the variance/mean value ofthe standoff was 0.85% [9].
The transducer was then mounted on an aluminum tensile test specimen, which was
also instrumented with a foil resistance strain gage. Strains measured by each sensor are
plotted in Figure 10 as a function of testing machine cross-head displacement. The initial
nonlinearity in strain/displacement behavior is from grip seating at the start of the test. As
can be seen from Figure 10, the results are quite comparable. Maximum deviation between
strains measured by the two different devices was 1% of the measured range.
To test transducer resolution, the same aluminum beam instrumented with a foil
resistance strain gage and the optic lever strain transducer was loaded in 4-point bending.
Figure 11 presents measured strain for each sensor plotted as a function of the computed
strain based on beam deflection. As can be seen in Figure I1, there is a friction effect
observable in the optic lever displacement transducer response. The sliding mechanism
"sticks" for a portion of the displacement, then slides. Only monotonic testing was
conducted. The effect of a reverse deformation would exhibit hysteresis in the transducer
response. Deviations between the two sensors are between 10 and 15 Its as plotted in
Figure 10.
0.30
'-"0"--- F.O. Lever jO

0.25"
I
0.20'
~ 0.15-'
0.10"
0.05
0.00 ,I. ~ , . I . . . . I . . . . I 9 ' ' ' I 9 9 o ,
0.0 0.5 1.0 1.5 2.0 2.5

Cross-head Displacement(ram)
Figure 10. Comparison of Measured Strains as Functions of Cross-head Displacement for

an Aluminum Tensile Test Specimen.
To summarize the results, an optic lever was configured in a sliding cylinder sleeve
mechanism to construct a 5 mm gage length extensometer. Such a device would be
particularly useful for materials testing to replace individual foil resistance strain gages
mounted on specimens destructively tested. The extensometer could be reused.
This prototype extensometer was tested on an aluminum rectangular cross-section bar
and its performance compared to a foil resistance strain gage mounted on the same bar.
Advantages inherent in this design include the ability to make medium (25 la~) resolution
strain measurements over a short gage length. Conventional foil resistance strain gage
based extensometer designs typically use gage length of 12 mm or longer. However, the
sliding mechanism does exhibit sticking due to friction. Attachment of the device to the test
specimen is also of concern. Two methods used for conventional extensometers were tried;
both methods have problems. Further work on attachment methods for all types of
extensometers is warranted.
70"
]~ F.O.L~ I
6o. [ , stniaOase[
C
5O
e~
4O
30
20
10
0 - - - i . . . . i - - - i . . . . i - - . - - i . . . .
0 10 20 30 4O 50 6O 70
Calculated Strain (10 4)
Figure 11. Small Strain Range Comparison of Measured Strains as Functions Computed
Strain for an Aluminum 4-Point Bending Test Specimen
CONCLUSIONS
Overall, the objectives of the development effort were attained, including
Development of a seven-fiber optic lever design model,
Determination of important optic fiber parameters including their relevance to optic

lever performance,
Construction of a prototype short gage length extensometer for strain measurement,

and
Preliminary evaluation of the prototype transducer including comparison to strain

gage performance,
The fiber optic lever based sliding cylinder-sleeve strain transducer offers a viable
alternative to foil resistance strain gages or strain gage based extensometers as a strain
measurement system. However, further work needs to be done to address friction concerns
in the sliding mechanismand attachment of the device to test specimens.
REFERENCES
[i] Houser, G. M., "Multiple Axis Strain Transducer Using Fiber Optic Levers," M.S.
Thesis, Department of Mechanical Engineering, University of Wyoming, Laramie,
WY, August 1990.
[2] Frank, W. E., "Detection and Measurement Device Having a Small Flexible
Transmission Level," U.S. Patent Number 3,273,447, September 20, 1966.
[3] Kissinger, C. D., "Fiber Optic Proximity Probe," U.S. Patent Number 3,327,584,
September 27, 1967.
[4] Cook, R. O., and Hamm, C. W., "Fiber Optic Lever Displacement Transducer,"
Applied Optics, Vol. 18, No. 19, December 1978, pp. 3230-3241.
[5] Jako, G. J., "Recording the Movement of the Human Basilar Membrane," Institute
of Applied Ear Resear~:h, April 1976.
[6] Fetheroff, C. W., Derkacs, T., and Matay, I. M., "Automated Plasma Spray
Feasibility Study," NASA Report No. CR-165418, March 1981.
[7] Miller, C. M., Optical Engineering; Optical Fiber Splices and Connectors Theory
and Methods, Dekker, New York, NY, 1986, pp. 31.
[8] Hecht, Eugene, Optics, Addison-Wesley,Reading, MA, 1988, pp. 79.
[9] Beckwith, T. G., Buck, N., and Marangoni, R. D., Mechanical Measurements,
Addison-Wesley, Reading, MA, 1982, pp. 270.
William N. Sharpe, Jr. ! and Bin Yuan 2
NEW APPLICATIONS OF THE INTERFEROMETRIC

STRAIN/DISPLACEMENT GAGE
REFERENCE: Sharpe, W. N., Jr. and Yuan, B., "New Applications of the
Interferometric Strain/Displacement Gage," Nontraditional Methods Of Sensine
Stress. Strain. And Damage in Materials and Structures. ASTM STP 1318, George F.
Lucas and David A. Stubbs, Eds., American Society for Testing And Materials, 1997.
ABSTRACT: The Interferometric Strain/Displacement Gage (ISDG) is an established

laser-based technique that measures the relative displacement between two gage points on a
specimen surface. The gage points are reflective markers illuminated with a low-power He-
Ne laser to form interference patterns that are monitored with a computer-controlled
system. Features of the technique are short gage length (50 - 300 Ilm) and no mechanical
contact. Two new applications of the ISDG are described in this paper. The first is the
measurement of strain on metal microspecimens that are 3 mm long with cross-sections
typically 0.3 mm square. Details of the test procedure and results from 52 tests conducted
as part of an ASTM-sponsored cross-comparison test are summarized. The second is strain
measurement on 3.5 ~tm thick polysilicon film specimens; this capability has enabled
measurement of the Young's Modulus and the tensile strength of this material which is
used in microelectromechanical systems (MEMS).
KEYWORDS: interferometry, lasers, microspecimens, polysilicon, thin films,

weldments, MEMS
The Interferometric Strain/Displacement Gage, or ISDG, is a non-contacting optical

technique for measuring strain or displacement between two reflective markers on a
specimen surface. The optical principle is simply Young's two-slit interference, but in
reflection rather than transmission. Applications of the ISDG over the past 25 years include
measurements of biaxial strains at notch roots (three markers are placed in an orthogonal
pattern), crack opening displacements of small cracks, dynamic strains and crack openings,
creep strains, and strains/displacements at high temperatures.
1Decker Professor and Chair, Department of Mechanical Engineering, Johns Hopkins

University, Baltimore, MD 21218
2Graduate Student, Department of Mechanical Engineering, Johns Hopkins University,
Baltimore, MD 21218
16
Copyright9
(PDVSA Los Teques)by ASTMtoInternational
pursuant License Agreement.www.astm.org
SHARPE AND YUAN ON NEW APPLICATIONSOF THE ISDG 17
A 1982 review article [1] summarized the applications of the ISDG until that time,
and a 1993 review [2.1described various methods (including the ISDG) of measuring crack
tip opening displacement. A NASA report L3_.]contains a detailed discussion of the optical
principles and the practical aspects of an ISDG measurement system. This paper is not a
review, but a description of strain measurements on two new kinds of specimens - - metal
microspecimens and polysilicon thin films. A brief discussion of the ISDG is first given
before the two different specimens, test procedures, and results are presented.
THE INTERFEROMETRIC STRAIN/DISPLACEMENT GAGE
The ISDG is an optical technique for measuring the relative displacement between
two reflective gage markers. These markers can be any surface that is reflective and small
enough to cause appreciable diffraction of the incident laser beam. In practice, the markers
are either indentations impressed into a metal surface with a Vickers microhardness tester or
gold pads or lines that are deposited onto a specimen surface. When the two markers are
illuminated with a laser, the diffracted reflections from each one overlap and interfere to
produce fringes. As the two markers move relative to each other, the fringe patterns also
move; their motion can be measured with photosensors and related to the relative
displacement change generated by strain (or displacement if the indents are across a crack).
The principle of the ISDG, which is actually equivalent to Young's two-slit

interference phenomenon, is illustrated in Figure 1. The schematic shows the two kinds of
reflective markers. When the gold lines are deposited, the edges of the lines are not vertical
and are somewhat rounded and tapered. This is enough to produce a strong reflection at
about 70 ~ from the incident laser beam, and the reflection is diffracted because of the small
dimensions of the gold lines which are only 0.5 ~tm thick. When indentations are used,
they are approximately 4 ~tm deep which means that the triangular facets cause diffraction
of the reflected beam.
Incident Laser Beam
Gold Lines as
Reflective Markers
Indentations as
Reflective Markers
Figure 1. Schematic of the ISDG.
A tensile strain causes the fringes to move toward the incident laser beam. Rigid-
body motion will also cause the fringes to move, so it is necessary to average the
movement of the two fringe patterns. The equation relating strain, e, to fringe movement is
~UI'I.1 + Am 2
E = (1)
2d 0 sin ot0
where k m l and Am2 are the relative fringe shifts of patterns 1 and 2, do is the original
distance between the indentations (the gage length), s is the wavelength of the laser, and
(zO is the angle between the incident laser beam and the fringe patterns.
When the fringes have moved one unit, i.e. each fringe moves into the location
previously occupied by its nearest neighbor, a relative displacement of 'one' has occurred.
Using typical values of L = 0.6328 p.m and (*0 = 42 ~ this corresponds to a relative
displacement between the two indentations of approximately 1 Ixm. If the initial spacing,
do, between the indents is 200 lxm, then a strain of 0.005 has occurred. Obviously better
resolution is needed.
The fringe patterns are converted to electrical signals with linear diode arrays - one
for each pattern. The arrays have 512 diodes - each with an aperture of 13 txm by 2.5 mm -
packaged in a 16-pin chip; the total size of the array is 2.5 mm by 12.6 mm. The arrays are
inserted in circuit boards containing amplifying circuitry and mounted in plastic mini-boxes
attached to translation stages which are mounted on adjustable rods. This provides the two
degrees-of-freedom necessary to position the diode array in the center of the fringe pattern.
Each diode array is accessed by a motherboard which samples each diode in turn and
provides an output signal from 0 to 5 volts.
In a typical setup, the spacing between maximum (or minimum) intensity is

approximately 50 diodes on the array. If one identifies local peaks by simply taking the
largest value, then the smallest resolvable fringe movement is 1/50. Since the two fringe
patterns are averaged, the minimum strain resolution occurs when one pattern moves one
diode spacing. Recall that the relative displacement for both fringes moving from one peak
to the next is approximately 1 ktm. The strain resolution is therefore approximately 1/2 x
1/50 x 1/200 = 50 microstrain if a 200 ~tm gage length is used. This resolution can be
increased by a factor of 100 by fitting a curve to selected intensity data points surrounding a
minimum and interpolating between the diode locations. A resolution of 0.5 microstrain is
not really meaningful; it is estimated that the resolution is approximately 5 microstrain.
The ISDG, as used for these tests, has a relative uncertainty o f + 3% and a
resolution of ~ 5 microstrain. The range depends upon the yield stress of the metal because
large plastic deformation causes the grains to distort in the neighborhood of the
indentations; however, a range of several percent strain is usually achievable. The sampling
rate is - 13 data points per second. The above is a very brief description of the ISDG; a
more complete presentation of the optical principles as well as details of typical
measurement systems are given in a review article [!] and a NASA report [3_].
Figure 2 is a schematic of the measurement system - - absent the microcomputer

which controls the test and records the data. The system is the same in general for both the
metal mierospecimens and the thin films, but there are differences. A linear air bearing is
used in each case to eliminate friction in the loading mechanism. Earlier versions of the
microspecimen system used a lubricated slot instead; the friction of that system was a small
percentage of the maximum force exerted on the specimen. A 90 N load cell is used for the
metal microspecimens and a 4.4 N one used for the polysilicon thin films. The piezoelectric
actuator has a range of only 180 Ixm which is satisfactory for the brittle thin films. A
SHARPEANDYUANONNEWAPPLICATIONSOFTHEISDG 19
motorized translation stage is used for the metal tests to allow sufficient extension of the
ductile metal. The thin film specimens are glued to the grips; whereas the metal
microspecimens fit into machined slots in the grips and are held in place with 0.86 mm
screws.
Laser
Fringe "'" Fringe Piezoelectric

Detector ~J Detector Translator -7
~ ~ .[/~7~r Be Load
Grips
Specimen
Figure 2. Schematic of the measurement system.
S T R A I N M E A S U R E M E N T S ON M E T A L M I C R O S P E C I M E N S
There are a number of reasons for testing microspecimens - - monitoring of

mechanical properties, studies of inhomogeneities, and testing of novel materials. The
original version of the microspecimen testing system developed at Hopkins was presented
at a 1991 ASTM Symposium [4_]whose emphasis was on the use of small specimens to
monitor the properties of steel pressure vessels subjected to nuclear radiation. Weldments
are inhomogeneous, and the properties of the base metal, weld metal, and heat-affected-
zone can differ greatly. If the weldments are large enough, microspecimens can be
extracted and tested. New materials are often made in very small quantities. Nanocrystalline
solids are typically made by depositing the material on the end of a 10 mm diameter rod in
thicknesses of only a few tens of microns. Mechanical properties can only be studied by
testing microspecimens.
Figure 3 is a schematic of the microspecimen. The 'dog-biscuit' shape was selected

because the wedge-shaped ends can fit into mating slots in the test machine grips. This
avoids any mechanical fastening; one can simply lay the specimen in the grips and run the
test. One inserts the specimen in the test machine with the aid of a low-power stereo
microscope and then moves the machine into place under the laser. In practice, each
specimen end is held down with a 0.86 mm screw, but this is simply to prevent the
specimen from falling out during the setup. The central test section of the microspecimen is
cut on a large radius to avoid failure near the grips; however, the stress concentration factor
is only about 1.03.
LI 3.1 -1
-I
Figure 3. Schematic of the microspecimen.
Specimens are cut from slices that are removed from larger samples by Electric
Discharge Machining (EDM). A small coupon, about 15 mm square, is cut from the region
of interest in a slice. The microspecimen is then cut from that coupon with a small CNC
mill using a 0,79 mm cutter. A high-speed, low-feed cut is taken in three passes with
continuous lubrication to reduce the machining residual stresses. Another approach is to
use wire EDM or ram EDM in which a hollow plunger is made in the shape of the
specimen. The thickness of the microspecimens varies because of the slicing process. Also,
the CNC mill has limited precision, and the width of the specimen may vary. Most of the
specimens tested had thicknesses varying between 0.15 - 0.3 mm and widths between 0.2 -
0.3 mm.
The coupon is polished with 240 grit wet-dry paper before the microspecimen is
cut. The machining process leaves the specimen attached to the coupon at each end. The
coupon and specimens are then polished successively to 1200 grit wet-dry paper.
Indentations are then applied and the dimensions measured. The microspecimen is cut from
the coupon with a small rotary grinder.
Forty seven microtensile tests have been conducted as part of an ASTM sponsored
Cross-Comparison Exercise on Determination of Material Properties
Through the Use of Miniature Mechanical Testing Techniques. The supplied
material is A533-B steel, and tests were conducted in the rolling, transverse, and short
transverse directions. Figure 4 is a ~ p i ~ l stress-strain curve for a microspecimen.
A15
8ooi....if....!f....if....i(
....i!....
~. 600 ..........'............ i............~............ ~............~..........
u~ 400
200
0 ....
0.0 0.5 1.0 1.5 2.0 2.5 3.0

STRAIN - percent
Figure 4. A representative stress-strain curve for a A533-B steel microspecimen.

The relative uncertainty bars associated with the stress and the strain measurements
are shown at 0,15 and 1.5 percent strain along with the 0.2 percent offset line used
to define the yield stress.
SHARPE AND YUAN ON NEW APPLICATIONS OF THE ISDG 21
Five macrotensile tests on specimens oriented in the transverse direction were also
conducted. Those specimens had the general 'dog biscuit' shape of Figure 3, but the test
cross-section was 3.2 mm square. Foil strain gages were used on both sides of the
specimens to measure strain.
Table 1 gives the mean values and standard deviations of these Cross-Comparison
tests. Most of the specimens were oriented in the rolling direction and the coefficient of
variations (standard deviation/mean) are 5.2% for the yield strength and 4.1% for the
ultimate strength. These values are close to those stated in ASTM A20 (5% for yield and
3% for ultimate strengths) which implies that the scatter in the testing of microspecimens is
not much greater than obtained in the testing of larger specimens. That speaks well for the
techniques and procedures used to test these microspecimens of very fine-grained (- 20 [.tm
diameter) steel.
Number of Yield Standard Number of Ultimate Standard'

Tests Strength Deviation Tests Strength Deviation
MPa MPa MPa MPa
Rolling 27 453 24 18 609 25
Short 10 438 19 6 597 31
Transverse 10 432 28 6 606 39
Macro 5 451 6.6 5 645 2.5
Table 1. Summary of results for A533-B steel.
There is a statistically significant difference in the yield strength in the rolling

direction as compared to the short and the transverse directions (which are nearly the
same). There is no significant difference in the ultimate strengths in the three directions; this
is an isotropic material for all practical purposes.
The microspecimens in the transverse direction show yield and ultimate strengths
that are significantly lower than measured on the macrospecimens. However, the difference
is only - 5% which means that microspecimens could be used to determine a conservative
estimate of macroscopic strength of a fine-grained material.
This microspecimen test system has been used for a study of the three regions (base
metal, weld metal, and heat affected zone) of a HY-100 steel weldment 515].Stress-strain
curves from the three regions are shown in Figure 5. One sees that there is considerably
more scatter in the larger-grained weld region, but the fact that this is an undermatched
weld is obvious. Metal in the heat affected zone (HAZ) is stronger, but there is much more
variation in the curves because this region was only about 2 mm wide and it was difficult to
be sure that a microspecimen contained only HAZ material.
Base Metal Heat Affected Zone

1500 .... ! ........ i ........ i .... 1500 . 1 . . . . . . . . .... . . . . . . . . .
} ...i... i i
i , - : j - -4 --,,~ L - - - :,-- ~ -
1200 1200 ~L~._~:..7~_...~.__.i ......
g. O.
, 900 =, ,OOr....f~ ........... i ........... ~........... ~...........
600
30O
i.oo
300
Z i,:
,'; i
i i
i ']
i !
.......... i ..........
;~:/-,-4 ........... {........... i ........... !........... ~..........
0 0 ~i,.i..i .... i ........ i .......

0.0 0.5 1.0 1.5 2,0 2.5 3.0 0,0 0.5 1.0 1.5 2.0 2.5 3,0
S T R A I N - percent S T R A I N - percent
Weld Metal
1500
1200
......... i .......... i .......... i .................................
900 ........... i .......... -i .......... ~ ..................................
r 600
9~ i - - ~ ~ i }
30O
0
0.0 0.5 1.0 1.5 2,0 2.5 3.0
STRAIN - percent
Figure 5. Stress-strain curves from three regions of a HY-100 weldment.
Microspecimens have also been tested in fully reversed fatigue [.6.]. The specimen,
shown in Figure 6, is a little larger and has holes in the grip ends; it is glued onto the grips
and also held with 0.86 mm screws. The test system is as shown in Figure 2 with the
piezoelectric actuator providing the fully reversed loading.
2-R 8,13
; ~ 0 .
I 6 1
I0 0
!~, 3.63 _!
7.62 ~_
Figure 6. Drawing of the fatigue microspecimen; the dimensions are in ram.
SHARPE AND YUAN ON NEW APPLICATIONSOF THE ISDG 23
Successful fully reversed fatigue tests have been conducted with this setup. Figure
7 shows the cyclic stress-strain curve for the first cycle and for cycle # 999. The specimen
was taken from the weld metal portion of a two-inch thick undermatched weldment of HY-
100 steel. The circles represent data taken by Dr. K. C. Wang of Concurrent Technology
Corporation of Johnstown, PA. Those specimens were larger and cycled between + I
percent strain. They showed cyclic softening of the weld metal as do the microspecimens.
To our knowledge, these are the smallest specimens that have ever been tested in fully
reversed cyclic loading.
FATA8W1
1000 . . . . . . . . . . . . . . . .
ft..
w
tr
-soo ................... i.~i ................ C y c l e #1 [

o CTC Dala
..... Cycle # 999
-1000 , , , , , , , , , , , ~ t , , , , I
-1.0 -0.5 0.0 0.5 1.0
STRAIN - percent
Figure 7. The first cycle (#1) and a stabilized (#999) stress-strain curves for
weld metal from a HY- 1O0 steel weldment.
STRAIN MEASUREMENTS ON THIN FILMS
Microelectromechanical systems (MEMS) are small motors, sensors, valves, etc.,

that offer great potential for a variety of commercial products ranging from arterial 'roto-
rooters' to carburetor air-flow controls. The same processes used to make integrated
circuits can be used to fabricate MEMS from ceramics and metals. The resulting
microdevices have planar dimensions on the order of one-tenth to several millimeters and
thicknesses on the order of one to several hundred micrometers. Engineers are now able to
design a system and predict its overall response. However, they cannot yet optimize the
design or predict the allowable load and life of a component because the material's
mechanical properties are not available.
It is obviously difficult to prepare and handle tensile specimens that are only a few
microns thick. It is also difficult to measure the strain accurately on the specimen;
crosshead displacements are not sufficient. Techniques and procedures have been
developed to measure the stress-strain curves of polysilicon film that is 3.5 ~m thick.
The material tested is phosphorus doped polysilicon deposited at the

Microelectronic Center of North Carolina as part of their ARPA-sponsored manufacturing
process. As such, it is representative of the material commonly used to form structural
elements for surface micromachined MEMS devices. The specimen is deposited on a
silicon wafer in the shape shown in Figure 8 with the tensile portion in the middle and two
support strips on each side. The silicon die is I cm square and the specimen is 3.5 I.tm thick
and 0.6 mm wide at its narrowest point. The wafer material beneath the tensile section and
in the shaded area shown in Figure 8 is removed from the back of the die by anisotropic
etching leaving the tensile polysilicon portion suspended between two grip ends that are
connected by the support strips.
24 SENSINGSTRESS,STRAIN,AND DAMAGEIN MATERIALS
. . ~ lOmm -,.
0.6~m
E
0
v- -W >----..
Grip
end
S u p p o r t strip-" ~'~'4 m m ---,-
Figure 8. Polysilicon specimen schematic.
Figure 9 is a SEM photograph showing the specimen from an oblique view. One can
see the silicon wafer in the foreground and the polysilicon layer on top. The tensile
specimen is stretched across the center of the photograph between the two grips.
Figure 9. SEM photo after the substrate is etched away. The tensile specimen
is in the middle of the picture.
Strain is measured with the ISDG using two gold lines as the reflective markers; see
Figure 1. These lines are applied by vapor deposition during the manufacturing process and
are 0.5 I.tm high, 20 lxm wide, 200 gm long, and 300 p.m apart; see Figure 10.
After the tensile specimen is released, the silicon die is placed in the grips of the
testing machine shown in Figure 2. The two large grip ends are glued in place and then the
two thin support strips are cut with a small rotary diamond saw. The tension test is then
conducted in the same manner as for the metal specimens.
Figure 10. SEM photo of two gold lines on a polysilicon specimen.
PS9
1.50 . . . . . . r . . . . . . I ' ' '
1.25 ............. ~. . . . . . . . . . . . . . . :. . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . ~. . . . . . . . . . . . .
.='o,,oo iiiiiiiiiiiiiiiilli ii ii!!ii i i i i iiii i i i i

~r 0.50 --............i .......... ~-.--..E - 163 Gt,,, ....
9
O~ : i :: Strength 9 1.31 GPa
0.25 ........
ooo ii
0.0 0.2 0.4 0.6 0.8 1.0
STRAIN - Percent
Figure 11. Representative stress-strain curve for polysilicon.
Figure 11 is a typical stress-strain curve for polysilicon. It is a brittle material with a

Young's modulus near that predicted from the properties of single crystal silicon. Ten tests
were conducted on specimens taken from the MUMPS 6 run and nine from MUMPS 8.
Table 2 gives the results. There is a statistically significant difference between the Young's
Modulus of the two runs but not between the strengths.
-- Young's Standard Tensile Strength Standard

Modulus Deviation GPa Deviation
GPa GPa GPa
MUMPS 6 163 4.8 1.23 0.20
MUMPS 8 173 6.9 1.27 0.11
All Data 168 7.6 1.25 0.16

==
Table 2. Summary of results for the polysilicon tests.
Figure 12 shows the results from tension-tension fatigue tests on a thin polysilicon
specimen. The fatigue specimen was first cycled at a low stress for 50 cycles. After no
change in response was observed, the load was increased for another 50 cycles. The last
incremental load increase occurred for cycle #300 and the specimen broke three cycles later.
The last loading response looks almost identical to a monotonic test which indicates that the
material is not damaged by the earlier cycling.
Figure 12. Fatigue testing of polysilicon.
CLOSING COMMENTS
The mechanical testing of small and/or thin specimens requires novel and specialized
procedures for both preparation and handling as well as force application. One way to
overcome these difficulties is to use geometries such as cantilever beams or membranes and
measure deflection. This indirect approach relates the measured forces and deflections to
the material properties through an equation derived for the particular geometry. These
approaches are not consistent with the definition of Young's Modulus given in ASTM
Standard E-111 which states, "The test specimen is loaded uniaxiaUy and load and strain
are measured, either incrementally or continuously."[.7.]. Further, ASTM E-111 requires
that strain be measured directly on the specimen with suitable extensometers. The specimen
preparation and loading are difficult problems, but they can be solved. Accurate strain
measurement has been a barrier to mechanical property testing of microspecimens, and this
paper has described two applications of the ISDG that remove that barrier.
There is always the question of the effect of the indentations on the measured
strains when metal specimens are tested. Comparisons have been made between the stress-
strain curves of microspecimens and larger specimens whose strain was measured with foil
gages L~]. Studies of biaxial elastoplastic strains at notch roots have been compared with
finite element computations [.8]. In both cases, one concludes that the indentations have
little effect on the measured strain. One can always make the indentations shallower and use
a more powerful laser if there is concern. However, the deposited gold reflective marker
approach may be preferable. That approach has not yet been used, but the technology
developed for Moir6 grid application [9.] can certainly be used to apply two gold lines.
Work in progress on metal microspecimens is worth mentioning; it is a

straightforward application of the techniques and procedures described above. A test
system that enables strain measurement on both sides of a metal microspecimen is now
being used. It is common practice in materials testing to measure strain on opposite sides of
a specimen and average them to eliminate the effects of bending. This new system will
enable better measurements of the elastic region of the stress-strain curve. A test program is
underway to grow larger grains in the A533-B steel to permit a study of the effect of grain
size on the stress-strain curves obtained from microspecimens. Techniques for preparing
microspecimens from prototype nanocrystalline materials have been developed and these
new materials are being tested as they are provided by colleagues.
There are no reliable measurements of Poisson's Ratio for polysilicon. It is quite

easy to extend the two-line gage to a four-line gage and measure biaxiai strains on these
thin materials. The procedures for biaxial strain recording using three indentations is fully
developed ~ and specimens are being prepared with four gold lines in a square array.
Fracture toughness of thin polysilicon films is difficult to measure, but the ISDG and the
techniques developed for tension testing can overcome this difficulty. Wider tensile
specimens of polysilicon with a thin slit in the middle are being prepared; they have the
appearance of a classic center-cracked panel. Two gold lines will be placed astride the crack
to permit opening displacement measurements. One can then record the load-COD curve
and determine the fracture properties of this thin material.
The ISDG is a specialized, although relatively easy to use, strain/displacement

measuring technique. There are certain situations in which it is the best (or only) means of
accurate strain measurement, and two of those have been presented here.
ACKNOWLEDGMENTS
The results reported come from various projects sponsored by The National Center
for Excellence in Metalworking operated by Concurrent Technology Corporation, the
National Science Foundation, and the Army Research Laboratory. David Danley, an
undergraduate at Hopkins, performed most of the A533-B microspecimen tests, and David
LaVan, a graduate student, tested the A533-B macrospecimens and assisted in the
development of the polysilicon tests procedures. Mr. Rick Edwards, Research Engineer at
the Applied Physics Laboratory of Hopkins, developed the procedures for releasing the
polysilicon tensile specimens. The support of the sponsors and the cooperation of
colleagues is greatly appreciated.
REFERENCES
[1_3 Sharpe, W. N., Jr., "Applications of the Interferometric Strain/Displacement

Gage," Optical Engineering, Vol. 21, 1982, pp 483-488.
[2] Sharpe, W. N., Jr. "Crack Tip Opening Displacement Measurement Techniques,"
Experimental Techniques in Fracture, Society for Experimental Mechanics, Inc.,
Bethel, CT, Chapter 7, 1993, pp 219-252.
3[2] Sharpe, W. N., Jr., "An Interferometric Strain/Displacement Measurement

System", NASA Technical Memorandum 101638, 1989.
Sharpe, W. N., Jr. and Fowler, R. O., "A Novel Miniature Test Machine",
ASTM STP 1204, Small Specimen Test Techniques Applied to Nuclear Reactor
Vessel Thermal Annealing and Plant Life Extension. pp 386-401, 1993.
[~] Sharpe, W. N., Jr., Zeng, H., Yuan, B., and Wallace, S., "A Technique for
Microsample Testing of Weldments", Fracture '94, The Fourth National
Conference on Fra.cture. University of the Witwatersrand, Johannesburg, 11 pgs.,
1994.
I_6_] Yuan, B. and Sharpe, W. N., Jr., "Fatigue Testing of Microspecimens",

FATIGUE 96, Berlin, May 6 - 10, 1996.
ASTM Standard E 111 - 82, "Standard Test Method for Young's Modulus,
Tangent Modulus, and Chord Modulus", 1995 Annual Book of ASTM Standards.
Sharpe, W. N., Jr., Yang, C. H., and Tregoning, R. L., "An Evaluation of the
Neuber and Glinka Relations for Monotonic Loading," Journal of Applied
Mechanics, 1992, pp $50-$56.
[9A Post, D., "Moir6 Interferometry", Handbook on Experimental Mechanics, Society

for Experimental Mechanics, Inc., Bethel, CT, Chapter 7, 1993, pp 297-364.
Nigel Shrive, 1 Erich Damson) Cy Frank, 3 Rick Meyer, 4 Lito Mejig5 and George Lucas5
AN EXTENSOMETER FOR SOFT MATERIAL STRAIN MEASUREMENT
REFERENCE: Shrive, N., Damson, E., Frank, C., Meyer, R., Mejig, L., and Lucas, G.
"An Extensometer for Soft Material Strain Measurement," Nontraditional Methods
of Sensing Stress, Strain, and Damage in Materials and Structures, ASTM STP 1318,
George F. Lucas and David A. Stubbs, Eds., American Society for Testing and
Materials, 1997.
ABSTRACT: The criteria for the design of a device for measuring strain in flexible
materials, particularly soft tissues, are described. An extensometer is described which
meets the design criteria. The resolution of the extensometer depends on its gage length.
With a gage length of 5 ram, a resolution of 20-50 microstrain can easily be obtained.
Some preliminary results are presented showing the sensitivity of the results obtained to
the contact force between the extensometer and the specimen.
KEYWORDS: strain measurement, extensometer, soft materials, biomechanics, soft

tissues
~Professor and Head, Department of Civil Engineering, The University of

Calgary, Calgary, Alberta T2N 1N4 Canada.
2Senior Research Technician, Department of Civil Engineering, The University of

Calgary, Calgary, Alberta T2N 1N4 Canada.
3Professor and Head, Division of Orthopaedics, Department of Surgery, The

University of Calgary, Calgary, Alberta T2N IN4 Canada.
4Manager, Transducer Products, MTS Systems Corp., Eden Prairie, Minnesota,

MN 55344-2290 USA.
5Senior Applications Engineer and Manager, respectively, Biomechanics

Division, MTS Systems Corp., Eden Prairie, Minnesota, MN 55344-2290 USA.
29
Copyright9
(PDVSA by ASTM
to License Agreement. www.astm.org
INTRODUCTION
The measurement of strain on soft, flexible materials has proven to be difficult.

Many techniques have been developed, and new methods are still being proposed. In the
area of biomechanics, the problem relates specifically to the measurement of strains on
soft tissues like tendons, ligaments, arteries and heart structures. These are soft tissues
which have been "replaced" with various synthetic alternatives over the years. Clearly
for the substitute materials to behave like the originals, knowledge of material and
structural behaviour is required.
A whole variety of optical and non-optical techniques has been described for
measuring strain on soft tissues. Video (1-4), and CCD (5-7) camera techniques have
been used to measure both axial and biaxial strain distributions on various tissues. Some
researchers have used the displacements measured as input to Finite Element
Formulations for obtaining estimates of the local strain field (7-10). In other
methodologies the displacements are used to calculate strains directly. The error
associated with the length or displacement measure in such techniques causes these
techniques to have a sensitivity typically in the 0.5% strain range for in-plane
deformation (11). Video methods in particular, are very sensitive to out-of-plane
displacements which can cause substantial error in the "measured strain" (11-13). Optical
methods for measuring strain on soft tissues have been assessed in detail elsewhere (14).
In the non-optical area, researchers have again been innovative, utilizing different
physical principles to develop strain measuring techniques. A commonly'used one in soft
tissues is the Liquid Mercury Strain Gage (LMSG) (15) which works on the same
principle as a foil strain gage. Fixation of the LMSG to soft tissue remains a problem, as
does definition of the gage length once mounting has been achieved. The Hall Effect
Displacement Transducer has been superseded recently by the Differential Variable
Reluctance Transducer (16) which overcomes the torsional problems associated with the
former. The latter however still suffers from the same difficulties as the LMSG, together
with alignment requirements of the two components. Despite their limitations, these two
techniques (LMSG and DVRT) do have the distinct advantage of being small and
implantable. In vivo measurements have been made with these devices (16-21),
providing invaluable insight to tissue deformation during activity.
The immense variety of the methodologies developed to measure strain on
flexible materials, including soft tissues, is indicative of the need to make such a
measurement, the difficulties involved and the lack of a generally accepted, accurate
technique. We have therefore developed a device which can measure strain accurately (•
a few microstrain, depending on the gage length) during the deformation of soft, flexible
materials. From the soft tissue perspective, the instrument can be applied to in vitro
experiments only at the moment. The device has been described in some detail
previously (22-24). Here, the design criteria are discussed briefly, and the sensitivity of
strain measurements made with the device, in terms of the force with which the device is
attached to the specimen, are examined.
SHRIVE ET AL. ON AN EXTENSOMETER 31
DESIGN CRITERIA
The criteria for designing a flexible material strain measuring device are:
1) Capability of measuring large strains: Soft tissues typically fail at 10-40% strain,
and other flexible materials can undergo deformation to much larger engineering
strains. Such strain levels are well beyond the capacity of standard foil gages.
2) Accommodation of rigid body motion of the gage length: Given the strains above,
the gage length will displace substantially from its initial position as the specimen
is deformed. Any strain measuring device must therefore travel with the gage
length and not read strain as a result of the rigid body displacement. Rigid body
displacement is not of concern with standard measuring techniques, since the
measuring device (strain gage, extensometer, Moir6 grid) is attached to and
carried by the specimen. These latter two factors however, cause problems (see
below) for the application of standard techniques on soft tissues.
3) No interference with strain being measured." Whatever device is used, the act of
measuring the strain should not interfere with the strain being measured. There
should be no chemical reaction altering the substrate material and thus its
deformational characteristics; no reinforcement of the specimen; no transfer of
load from specimen to device. Glues which stick to soft tissues have recently
become available but they react with the tissue and alter it!
4) No damage to or distortion of specimen: The spring-loaded knife edges of a
standard extensometer will cut straight through a soft-tissue and many flexible
materials. Further, the weight of an extensometer, hanging on a flexible
specimen, will distort that specimen, since typically these materials have little
flexural rigidity. The behaviour of a distorted specimen under load is unlikely to
be the same as an undistorted one. In this sense, the strain being measured is
altered (see above). These factors have ruled out the use of standard
extensometers.
5) Frequency response: Since many tests involve cyclic loading, the device must be
able to follow cyclic deformation at frequency levels which allow reasonable
times for tests involving large numbers of cycles.
6) Accommodation of lateral displacement of gage length: A particular requirement
for soft tissues is that the device be able to move laterally as well as longitudinally
in a test. Many soft tissues are curved in the undeformed state, and thus straighten
under tensile load. The straightening can involve lateral movement of the gage
length from its initial alignment.
The curvature of many soft tissues in their undeformed state also means that the
device will have to be attached to different specimens at different angles. Angular
adjustment will be required in whatever support mechanism is used for the strain
measuring device.
7) Temperature insensitivity: If the device is to be used for flexible materials other
than soft tissues, there may well be the need for strain measurement over a range
of temperatures. Ability to accommodate such a need would provide wider
applicability of the device than just soft tissues. This could be achieved with
appropriate temperature compensated gages, and making the tips of the

extensometer from a low conductivity ceramic.
ASSESSMENT OF CRITERIA AND DEVICE DESIGN
The issue in criterion 6 above ruled out optical techniques because of their
sensitivity to out-of-plane movements (12, 14). Moir~ fringe grids can not be glued on to
soft tissues without changing the properties of the tissue, and foil gages have insufficient
capacity. Suturing a device onto the tissue was discarded, as the linkage between the
device and the tissue will have highly variable rigidity, and the gage length can not be
defined accurately. The only established strain measuring methodology which appeared
to be useable was extensometry.
With an extensometer, the problem of the weight of the instrument distorting the
specimen when the instrument is attached to the tissue, is easily solved by not hanging
the insmgnent on the tissue, but supporting its weight on an external frame. This solution
to the one problem however, simply creates another: how are lateral and large
longitudinal movements of the gage length accommodated by the measuring device? The
device must follow the gage length, yet the device must not apply extraneous load to the
specimen. This problem was solved through the use of a flexible support ring between
the measuring arms of the extensometer and the main supporting frame (Figure 1).
~ flexible support ring
specimen
// ~ ~ 1 ~ ' g a u g e length
~ ' ~ '1 ~ ~ settingclip
. ~ \ \ \" ~ ~ cantilever arm
~ ~ with serrated tip
support system to ~\
frame adjustable in \ \ /
the rotational and ~k,~ ///
linfreTodegrftt~?f
plane of the ring I ~.~
9O ~
FIG. 1--Schematic drawing ofextensometer for soft flexible materials. The width of the
flexible ring is 5 ram.
The flexible ring supports the weight of the working part of the device (made as
light as possible), and is squashed lightly in compression as the extensometer arms are
pushed up against the specimen. If the gage length moves toward the frame when load is
applied, the ring merely deforms more. If the gage length moves away from the frame,
then the ring decompresses a little. Clearly the amount of lateral movement must be
determined roughly in pilot tests, so that at maximum lateral movement of the gage
length there is still sufficient force on the extensometer arms to avoid slip. The ring has
to be sufficiently flexible that there is little change in the contact force as the lateral
movement occurs.
Longitudinal movement of the gage length is accommodated by the flexible ring
rolling and deforming in shear. The high flexibility of the ring ensures that longitudinal
forces induced by the movement are minimal. The strain gage bridge circuit on the
extensometer arms is wired to nullify the effects of such movement on output voltage.
However, this circuit arrangement can be negated if the vertical contact force is too high
(see use of device).
The extensometer arms themselves are simply cantilevers, strain-gaged at their
roots in a full bridge arrangement, temperature compensated by the manufacturer. To
avoid damage to the specimen, the contacting ends are not sharp knife-edges but serrated
with the ends of the serrations being relatively blunt. Initially the serrations were bent
alternately up and down as in a saw-blade, but this has proven unnecessary. Sufficient
grip can be obtained with straight points.
The two cantilever arms set the gage length. A little clip can be used to hold the
two arms a set distance apart as they are brought onto the specimen (Figure 1). Initial
devices had a gage length of 5 mm simply because that was an appropriate length for the
tissue for which the device was initially devised - the rabbit medial collateral ligament
(MCL). The cantilever arms deform as the specimen strains and the output voltage is
converted to strain given the gage length and the deformation measured. Larger or
smaller gage lengths are feasible. With larger gage lengths, greater strain sensitivity can
be expected. With the 5 mm gage length, the force required to open the arms 2 mm (to
give 40% strain) is 0.05 N. This force should not interfere with the strain being measured
unless the material is exceptionally flexible. Strain resolution depends on the stability
and sensitivity of the conditioning and amplifying circuits. In the 50,000 lae range, with
a 10 volt excitation voltage, the ability to read 1 mvolt suggests a resolution of 5 tie. In
practice, with noise and the current 5 mm gage length, the resolution is more in the region
of 20-50 tae. With the original instnunent, a strain capacity of 40% was obtained: the
commercial device has a safety stop limiting strain capacity to 25% (Figure 2). The
device is patented (24) and is obtainable only as MTS Systems Corp. Model 632.32
Extensometer.
USE OF DEVICE
Results obtained on the rabbit MCL with the new device have been compared
with those obtained simultaneously with a Video Dimension Analysis System (22). The
new device showed much greater resolution. Results obtained from the extensometer, a
VDA and a DVRT from t~ts on human patellar ligament stretched at different extension
rates again showed the improved resolution and accuracy of the extensometer (23). The
versatility of the device was demonstrated by using it to determine longitudinal and

circumferential strains on pressurized artificial aorta specimens (23).
0.76
0.2
0.00 6.00 10.00 16.00 20.00 26.00
STmUN (%)
FIG. 2--Stroke vs strain measured in a direct "calibration" test with one arm of the
extensometer held to the base of the test machine and the other attached t6 the actuator.
A stroke of 1.25 mm on a 5 mm gage length should give 25% strain.
Experience in the above tests indicated that good results would only be obtained if
care was taken to bring the extensometer arms onto the specimen as close to
perpendicular as possible (+ 2.5 ~ to perpendicular). Magnification of the contact zone as
contact is made is essential in this task: a video camera has been used, allowing an
enlarged view on the monitor, but a magnifying glass is easier to manipulate. Failure to
adjust the angle of the arms such that they are brought onto the specimen together can
easily cause one or other arm to slip during a test. Figure 3 shows the result of one arm
slipping repeatedly in a cyclic creep test compared with correct contact.
Recent experience indicates that the contact force is a second essential to get fight.
The current device with a 5 m m gage length can have contact forces up to 0.06 N ( 6
grams force) per arm. Experience indicates that at least 0.02 N (2 grams per arm) is
needed for good results. In Figure 4 we show the effect of too much (> 0.06 N per arm)
and too little load (0.01 N per arm) on cyclic tests on an elastic band. If the initial contact
load is too light, in the test configuration in Figure 5, then the lower arm gets dragged up
toward the upper arm (slipping) as the test begins. A large negative strain thus appears to
occur as a base shift in subsequent cycles. If the initial contact is too heavy the flexible
ring flexes rather than deforming in shear and a much smaller slip occurs as the upper
arm is unloaded. The result is therefore similar to the too light case, an apparent negative
strain as a permanent offset in subsequent cycles.
SHRIVEET AL. ON AN EXTENSOMETER 35
LENGTH ST~N (X)

(m)
2.8
51 LENGTH
'l::::
,- iil
,.2 ii:: "2
o.e
!
o ,~
0
I I I I
50 100 150 200 250
TIME
LENGTH STUN (X)
1.6
1.4 ~
1.2 ~
I m
0.5
0,8 ~
0.6 ~ 0
0.4 ~
-0.5
.1
~ -i,iiiiiilillllllllllllllll L
-I.5
-0.2"
.0.4 I I I I I -2
100 150 300 ;!50 300
TIME
FIG. 3--(a) A normal test result above for a cyclic tensile creep test on a rabbit medial
collateral ligament. Midsubstance strain measured with the extensometer is shown with
the solid line. The dotted line shows the stroke (length) of the actuator. The shift of the
midsubstance strain from being under the stroke curve initially to going above it, suggests
strain redistribution along the length of the ligament during the test. (b) These curves
show the effect of too small a contact force. The length increases as before, but the strain
shows a rapid movement towards compression before setting down to a repetitious result
still involving minor slip of one arm.
36 S E N S I N G S T R E S S , S T R A I N , A N D D A M A G E IN M A T E R I A L S
EXTENSOMETER ...... TOO LIGHT mlcrostrain J

- - - TOO P,ARD m e r o ~ n I
30000.
20000-
10000
0
Z
t
*10000 ~, t , / I \,, ! ,
:8
-20000
-30000 f" :': /". :',, /",,

o
9"
; ,,', l
, ,,", .
+. ",
"o :
40000 ; : ,o ", : ",
",.: ,, ? \ : ".
-50000 . . . . . . . . . "" . . . . :i '. .................
0 2 4 6 8 10 12
TIME (sec~
FIG. 4--Results for different contact loads for a cyclic tensile creep test on a normal
elastic band.
FIG. 5--The test configuration giving the results in Figures 3 and 4. The extensometer,
the support ring and adjustments can be seen on the left and a closer view of the
extensometer mounted on the elastic band is on the right.
Similar results have been obtained in cyclic creep tests on soft tissues. The
difficulty this produces is that what may be the right contact load for one specimen may
be too light for the next. It therefore becomes a matter of user skill and experience to
mount the extensometer at the correct angle, with the right contact force for each
specimen. For more prismatic and less variable specimens than soft tissues, this should
not prove such a problem.
Another issue we have run into is that of condensation. Soft tissues are typically
tested while immersed in a physiological solution because their properties change quite
rapidly as they dry in air. However, the solution in which the tissue is tested affects the
result obtained (25). Therefore we developed a new test system in which the atmosphere
surrounding the specimen is kept at body temperature and 95-100% R.H(26). With this
system the tissue retains its moisture content and its chemical constituents. However, the
tips of the extensometer inside the chamber act as condensers. Over time water builds up
at the tips and the weight of this water can cause exlra deformation in the supporting
flexible ring, affecting results, as shown in Figure 6.
STATIC LOAD RELAXATION I Force 1 N I

L. . . . . . STRAIN1%j
4.5 5
4 ~-- -::-=~~. 44"5
3.5
93,5
3
2.5
92.5 ~"
1.5
9 1,5
1
.l
0.~ 90.5
0 0
TIME (sec)
FIG. 6--Test result for a static creep test on a rabbit medial collateral ligament. Water
drops build up through condensation and fall off the arms periodicaUy, causing an
oscillation in the strain during a stress relaxation test.
Therefore we are now assessing the ramification of testing tissues horizontally,

and supporting the extensometer vertically above the specimens. Water condensing on
the extensometer arms in this situation would run down onto the specimen.
specimens consistently for good results. Once this is achieved for inert materials, then
the operator can apply the device to soft tissues with confidence that results are a
manifestation of tissue behaviour rather than arm slippage on the specimen.
CONCLUSION
An accurate strain measuring device for flexible materials including soft tissues
has been developed. Strain resolution varies with the exact construction of the
extensometer but is on the order of a few microstrain. For accurate results, the device
must be brought square onto the specimen, and the contact force should be great enough
to avoid slip of the arms, but not be great as to restrict the desired deformation mode of
the flexible support ring.
ACKNOWLEDGEMENTS
Much of the development work of this device was supported by grants from the
Medical Research Council of Canada: greatly appreciated. The efforts of Gail Leask
(technician) and Susan Anand (secretary) are of inestimable value and gratefully
acknowledged.
REFERENCES
(1) Woo, S-Y., Akeson, W.H. and Jemmott, G.F., "Measurements of"
Nonhomogeneous Directional Mechanical Properties of Articular Cartilage in
Tension," Joomal ofBiomeehanics, Vol. 9, 1976, pp. 785-791.
(2) Lanir, Y. and Fung, Y.C. "Two Dimensional Mechanical Properties of Rabbit
Skin - Experimental System", Journal of Biome~hanics, Vol. 7, 1974, pp. 29-34.
(3) Vawter, D.L., Fung, Y.C. and West, J.B., "Elasticity of Excised Dog Lung
Parenchyma," Journal of Applied Physiology, Vol. 45, 1978, pp. 261-269.
(4) Lendon, C.L., Davies, M.J., Richardson, P.D. and Bom, G.V.1L, "Testing of
Small Connective Tissue Specimens for the Determination of the Mechanical
Behaviour of Atherosclerotic Plaques," Journal of Biomedical Engineerine, Vol.
15, 1993, pp. 27-33.
(5) Vito, R.P., "The Mechanical Properties of Soft Tissues-I: A Mechanical System
for Bi-axial Testing," Jgurnal ofBiomechanics, Vol. 13, 1980, pp. 947-950.
(5) Vito, R.P., "The Mechanical Properties of Soft Tissues-I: A Mechanical System
for Bi-axial Testing," Journal of Biomeehanics, Vol. 13, 1980, pp. 947-950.
(6) Stewart, S.F.C., Lyman, D.J., and Benner, R., "Solid State Line Scan Cameras for
Measuring Strain in Soft Tissues and Implants," Annals of Biomedical
Kagiag.c2:iag,Vol. 17, 1989, pp. 671-684.
(7) George, D.T. and Bogen, D.K., "A Low-Cost Fiber-Optic Strain Gauge System
for Biological Applications", IEEE Transactions on Biomedical Engineering, Vol.
38, No. 9, 1991, pp. 919-924.
(8) Hoffman, A.H. and Grigg, P., "A Method for Measuring Strains in Soft Tissue,"
Journal ofBiomechanics, Vol. 17, No. 10, 1984, pp. 795-800.
(9) Oomens, C.W.J., Ratingen, M.R., Janssen, J.D., Kok, J.J. and Hendriks, M.A.N.,
"A Numerical-Experimental Method for a Mechanical Characterization of
Biological Materials," Journal of Biomechanics, Vol. 26 No. 4-5, J993, pp. 617-
621.
(lO) Hashima, A.R., Young, A.A., McCulloch, A.D. and Waldman, L.K.,
"Nonhomogeneous Analysis of Epicardial Strain Distributions during Acute
Myocardial Ischemia in the Dog," Journal of Biomechanics, Vol. 26, No. I,
1993, pp. 19-35.
(11) Lain, T.C., Frank, C.B. and Shrive, N.G. "Calibration Characteristics of a Video
Dimension Analyser (VDA) System," Journal of Biomechanies, Vol. 25, No. 10,
1992, pp. 1227-1231.
(12) Derwin, K.A., Soslowsky, L.J., Green, W.D.K. and Elder, S.H., "A New Optical
System for the Determination of Deformations and Strains: Calibration
Characteristics and Experimental Results", Journal of Biomechanic~,Vol. 27, No.
10, 1994, pp. 1277-1285.
(13) Amadio, P.C., Berglund, L.J. and An, K-N., "Biochemically Discrete Zones of
Canine Flexor Tendon: Evaluation of Properties with a New Photographic
Method",Joumal of Orthooaedic Research, Vol. I0, No. 2, 1992, pp. 198-204.
(14) Shrive, N.G., "Soft Tissue Strain Measurement," in Ot3tical Measurement

Methods in Biomechanics, J. Shelton and J. Orr, Eds., Chapman and Hall, in
press, 1996.
(15) Brown, T.D., Sigal, L., Njus, G.O., Njus, N.M., Singerman, R.J. and Brand, R.A.,
"Dynamic Performance Characteristics of the Liquid Metal Strain Gage,", Journal
~ , Vol. 19, No. 2, 1986, pp. 165-173.
(16) Tohyama, H., Beynnon, B.D., Flemming, B.C,, Johnson, R.J. and Pope, M.H.,
"Ultimate Failure Strength is Not the Only Criterion for Evaluation of Tendon
Graft Fixation," Transaction 40th Qt~opaedie Research Society.,Vol. 19, 1994, p.
641.
(17) Lochner, F.K., Milne, D.W., Mills, E.J. and Groom, J.J., "In Vivo and in Vitro
Measurement of Tendon Strain in the Horse," Journal of Veterinary.Research,
Vol. 4t,No. 12,, pp. 929-1937.
(18) Riemersma, D.J., van den Bogert, A.J., Schamhardt, H.C., and Hartrnan, W.,
"Kinetics and Kinematics of the Equine Hind Limb: In Vivo Tendon Strain and
Jint Kinematics," Journal of Veterinary Research, Vol. 49, No. 8, pp. 1353-1359.
(19) Colville, M.R., Marder, R.A., Boyle, J.J. and Zarins, B., "Strain Measurement in
Lateral Ankle Ligaments," The American Journal of Sports Medicine, Vol. 18,
No. 2, 1990, pp. 196-200.
(20) Howe, J.G., Wertheimer, C., Johnson, R.J., Nichols, C.E., Pope, M.H. and
Beynnon, B.D,, "Arthroscopie Strain Gage Measurement of the Normal Anterior
Cruciate Ligament," Journal of Arthroscooic and Related Surgery_,Vol. 6, No, 3,
1990, pp. 198-204.
(21) Terry, G.C., Hammon, D., France, P. and Norwood, L.A., "The Stabilizing
Function of Passive Shoulder Restraints," The American Journal of Sports
Medicine. Vol. 19, No. 1, 1991, pp. 26-34.
(22) Shrive, N.G., Damson, E., and Frank, C.B., "Technology Transfer Regarding the
Measurement of Strain on Flexible Materials," in Experimental Mechanics, 1992,
E.G. Little, Ed., Clinical Aspects of Biomedicine, Vol. 3, Elsevier, pp. 121-130.
(23) Shrive, N.G., Damson, E., Iverslie, S., Moore, K., Frank, C., Mejia, L. and Lueas,
G., "A New Technique for Measuring Strain on Flexible Materials such as Soft
Tissues," Proceedings of the XIII IMEKO (International Measurement
~Olffederation~Conference. Vol. 2, 1994, pp. 1552-1556.
(24) Shrive, N.G., Damson, E., Meyer, R.A. and Iverslie, S.P., U.S. Patent No.
5,463,902, 7 November, 1995.
(25) Chimich, D., Shrive, N., Frank, C., Marchuk, L. and Bray, R., "Water Content
Alters Viscoelastic Behaviour of the Normal Adolescent Rabbit Medial Collateral
Ligament," Jgumal of Biomechanics, Vol. 25, No. 8, 1992, pp, 831-8337.
(26) Wilson, A., Shrive, N., Damson, E., Frank, C., Leask, G., Mikalson, I. and Kraus,
V., "An Environment Chamber for Soft Tissue Testing," Proceedings of the 1995
l~i9engineering Conference, ASME, BED-Vol. 29, pp. 201-204.
Nontraditional Crack Measurement
Techniques
CopyrightbyASTMIntl'(allrightsreserved);SatJan2808:00:22EST2012
Downloaded/printedby
(PDVSALosTeques)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
A. Varvani-Farahani and T. H. Topper
Short Fatigue Crack Characterization and Detection Using Confocai Scanning

Laser Microscopy (CSLM)
REFERENCE: Varvani-Farahani, A. and Topper, T. H., "Short Fatigue Crack Char-

acterization and Detection Using Confocal Scanning Laser Microscopy (CSLM),"
Nontraditional Methods of Sensing Stress, Strain, and Damage in Materials and Struc-
tures, ASTM STP 1318, George F. Lucas and David A. Stubbs, Eds., American Society
for Testing and Materials, 1997.
ABSTRACT- This paper presents a new technique for studying the growth and
morphology of fatigue cracks. The technique allows short fatigue crack growth, crack
depth, aspect ratio (crack depth/half crack length), and crack front configuration to be
measured using a Confocal Scanning Laser Microscope (CSLM). CSLM measurements
of the initial stage of crack growth in A1 2024-T351 revealed that microstructurally
short fatigue cracks grew initially along a plane inclined to the applied stress. The angle
of the inclined plane (Stage I crack growth) was found to be about 45 degrees to the
axis of the applied tensile load. Aspect ratio and the angle of maximum shear plane
(Mode II), obtained using the CSLM technique, showed a good agreement with those
obtained using a Surface Removal (SR) technique. The aspect ratios obtained using the
CSLM technique were found to remain constant with increasing crack length in A1
2024-T351 and SAE 1045 Steel at 0.83 and 0.80, respectively.
Optical sectioning along the length of a crack revealed that the crack front in the
interior of the materials has a semi-elliptical shape. These results are in good agreement
with results obtained using the SR technique.
The CSLM technique was employed to characterize the fracture surface of fatigue
cracks in an SAE 1045 Steel. CSLM image processing of the fracture surface near the
crack tip constructed a three dimensional profile of fracture surface asperities. The
heights of asperities were obtained from this profile.
Optical sectioning from a post-image-processed crack provided crack depth and
crack mouth width at every point along the crack length for each load level. The crack
opening stress was taken as the stress level at which the crack depth stopped increasing
with increases in applied stress.
KEYWORDS: short fatigue cracks, aspect ratio, crack depth profile, stage I (Mode
II), optical sectioning, fracture surface asperity, crack opening stress
The most common methods used to determine the size of small fatigue cracks are
optical or scanning electron microscopy of the crack or its replicated surface. Normally,
information is provided only about the intersection of the crack with the surface. In
some cases, crack depth has been measured by progressively removing surface layers.
For example, in their paper on the propagation of short fatigue cracks Wagner et al [1]
reported crack depths after successively electropolishing a Ti-8.6 AI alloy. In other
instances electrical potential (EP) techniques have been used to study the formation and
43
Copyright9
(PDVSA by ASTM
growth of small cracks. Gangloff [2] continuously monitored crack depth via DC
electrical potential measurements, with the calibration model accounting for shape as
well as for variations in depth of the elliptical surface cracks. The accuracy of crack
depth measurement using this technique has been reported to be 0.5-1mm and
_+0.15mm in nickel-based alloys [2] and X65 steel [31, respectively. London et al [4]
utilized a method that monitored surface acoustic waves to quantify the depth and
closure characteristics of microscopic surface fatigue cracks 70 to 250~tm deep in AISI
4140 steel compact-tension specimens. Baxter et al [5] reported an electro-chemical
method for detecting small fatigue cracks in both steel and aluminum alloys through the
identification of locations where cracks ruptured the oxide film on the surface. These
rupture sites were imaged using photoelectron microscopy and fatigue cracks as short
as 10~tm in length were frequently detected.
This paper introduces the technique and results obtained when applying Confocal
Scanning Laser Microscopy (CSLM) to determine the crack depth profile of Stage I
microstructurally short fatigue cracks. This technique (CSLM) provides a quantitative
study of the relationship of fracture surface asperity characteristics, under constant
amplitude and periodic compressive overloading conditions, to crack growth rate and
crack closure stress for short fatigue cracks in SAE 1045 Steel.
CSLM allows optical sectioning of the crack to construct accurate three
dimensional images.
METHOD AND MEASUREMENTS
M~tCri~l. Svecimen Design and Preparation
A12024-T351 Alloy
This alloy was precipitation hardened by solution heat treatment at 495~ water
quenched, then stretched to about 1.5 percent strain and naturally aged. This alloy has a
chemical composition (W%) of 0.50 C, 0.50 Fe, 4.35 Cu, 0.60 Mn, 0.50 Si, 1.5 Mg,
0.10 Cr, 0.25 Zn, 0.15 Ti, and the remainder A1. The yield strength of the alloy was 350
MPa. The grain dimensions were 415 x 158 x 341xm in the longitudinal (-L), transverse
(-T), and short transverse (-S) directions respectively. Smooth flat fatigue specimens
were machined to the dimensions given in Fig. la.
$AE 1045 Steel

This material is a medium carbon heat treatable steel which is widely used in the
automotive industry. The chemical composition (W%) of this alloy is: 0.46 C, 0.17 Si,
0.81 Mn, 0.027 P, 0.023 S, and the remainder Fe. The microstructure of the SAE 1045
steel after final polishing showed pearlitic-ferritic features containing up to 30 ~tm long
sulfide inclusions in the extrusion direction. The modulus of elasticity is 206 GPa, the
cyclic yield stress is 448 MPa, and the fatigue limit stress (stress ratio of R=-I) is 300
MPa. Notched flat and smooth tubular 1045 steel specimens are shown in Fig. lb and
Fig. I c.
The specimens were mechanically polished to give a very smooth surface (final
polish with 0.25~tm diamond paste).
Flat aluminum specimens were tested under load control at constant amplitude tension-
compression (R=-1) stresses between +140 MPa. Notched steel specimens were loaded
under load control at two loading conditions of i) tension-compression stresses of
+138MPa, and ii) a loading history having a periodic compressive overload (-300,-360,
and-430MPa) followed by n=50 small tensile-compressive cycles with stresses of +138
MPa. Tubular steel specimens were tested under constant biaxial loading (180 ~ out-of-
phase) when the axial and hoop strains were controlled between +015% and -0.15%.
VARVANI-FARAHANI AND TOPPER ON CSLM 45
80 mm
2mrr
I
9.7ram (a)
E
E R=19mm
p,J,J,J,r,J, (b)
r=0.5mm
101.5mm ~ ,
A
i
(c)
!V
Gauge Length
89ram
Fig.i--a) A1 2024-T351, b) and c) SAE 1045 steel specimens.
Crack Depth Meastlr~ment
a) Confocal Scanning Laser Microscopy fC.SLM)
Fig.2 illustrates the reflected-light optical path in a CSLM assembly. Initially, the
laser beam passes through a spatial filter, beam expander, and beam splitter and is then
deflected by a scanning mirror SM1 rotating about the Z-axis. Lenses L1 and L2
comprise a unitary telescope, which focuses the rasting beam on the centre of a second
scanning mirror SM2, rotating about the X-axis. A second unitary telescope, containing
lenses L3 and L4 focuses the scanning beam back to the axis at the entrance pupil of the
objective lens, which in turn focuses the incoming beam on the specimen. Light
reflected from this spot is collected by the microscope objective and passes back
through the scan system, where it is descanned. The returning beam is partially
reflected by the beam splitter into the detection arm of the microscope, where it is
focused on a 50l.tm diameter pinhole by the detector lens. Light passing through the
pinhole is detected by a 28mm diameter detector which is a photomultiplier tube
providing peak spectral sensitivity at a wavelength of 460nm. The detector gain was
chosen to prevent saturation of the detector with light and to provide full use of the
dynamic range of the frame grabber (0-255 pixels). The scan system is computer
controlled to carry out a raster scan of the focused spot across the specimen, and the
reflected-light image is collected on a pixel-by-pixel basis as the scan proceeds. Since
the pinhole is placed at the focal point of detector lens (10cm focal length), only light
that forms a parallel beam before entering detector lens will pass through the pinhole
and be detected. Light reflected from specimen planes that are closer to the microscope
objective than its plane will form a diverging beam that would focus behind the
pinhole, so most of that light will strike its edges and will not be detected. Similarly,
light from any plane in the specimen below the focal plane wilt be focused in front of
the pinhole and will strike its edges as a diverging beam. Thus the pinhole, which is
confocal with the focus spot of the laser beam at the specimen, rejects light from above
or below the plane of focus in the specimen. It is this action that allows the confocal
microscope to perform optical sectioning. This is particularly important when imaging
through a scattering medium, because it allows the microscope to detect light reflected
from the focal plane, while rejecting light scattered by the medium above the focal
plane and it also enables the microscope to measure depth. When reflected light from
the focal plane reaches the pinhole, the detector sends a clear image to a monitor. On
the other hand reflected light from outside the focal plane does not reach the pinhole
and the image produced by the detector is dark.
Using an Argon-Krypton laser beam source (wavelength 488nm), power of
0.25mW and an objective numerical aperture 0.9, resolutions of 0.25ktrn and 0.50ktm in
the X-Y plane and depth (Z direction) respectively were obtained. The accuracy of the
CSLM system depends on the scanning mirrors (SM1 and SM2) which can be
positioned to within +-10ktradians. Errors occur due to vibration and mirror instability
which in this system do not exceed +(0.25-0.50~tm). Lens aberration was minimized by
using the double lenses LltL2 and L3/L4.
In the present work, a fatigued AI 2024-T351 specimen was initially loaded
gradually to 300MPa (less than the yield stress) using a simple tensile machine
equipped with a Ioadcell and indicator. This opened the fatigue crack. The laser beam
was centred on the cracked area by direct observation through the attached optical
microscope. The specimen was tilted to a known angle with respect to the laser beam
using a rotating level and then scanned by the beam. Images from different crack depths
at a constant angle were obtained by changing the specimen height using a piezo-
electric monitored stage.
Laser
Detector Beam
Expander
Pinhole(PH)
Beam Splitter
Detector X-Scan
Lens
Z-S~
ing Mirror SM2
(SM1)
Microscope
Objective
Cracked Specimen
Fig. 2--Schematic presentation of CSLM assembly.
A difference of 15 millivolts was equivalent to one micrometer in crack depth. A

set of 30 confocal images was acquired at progressive depths of l~tm. Post-image
processing was later used to combine all the images and create a complete profile of the
crack. In this manner depths at different positions along the crack were measured by
taking several optical slices along the crack length [6].
This procedure was repeated at different tilt angles. The longest (correct) crack
depths were recorded when the image processing was performed on the specimen
positioned at 40-50 degrees to the beam. It is not surprising that this occurred since
Stage I cracks are most likely to be found at this angle.
b) Surface Laver Removal Tr162
The crack lengths were first measured and then the edge of specimen was indexed
using an optical microscope (as a reference point for measurements). The initial
thickness of the specimen was measured using an Olympus type SA105A microscope
equipped with fiber optics with an accuracy of l~tm. Metal from the specimen surface
was removed by polishing the surface using 0.25~tm diamond paste. The specimen was
then cleaned with acetone and the new crack length and specimen thickness were
recorded. This process was continued until no surface crack could be observed. The
difference between the initial specimen thickness and that at which no crack could be
detected was taken as the crack depth. During this procedure the slight shift of crack
position relative to the marked area indicated that crack growth into the interior of the
aluminum alloy occurred on a plane inclined to the surface.
A~pr162Ratio and Crack Denth Profile
The ratio of crack depth to half crack length (a/c) was obtained in A1 2024-T351
and SAE 1045 Steel for a few short fatigue cracks with lengths of 10-100 ~tm using
CSLM. Aspect ratios (a/c) of cracks which experienced fatigue loading just above the
fatigue limit stress were found to be approximately 0.83 and 0.80, respectively. Using
the CSLM technique, the crack depth profile (in Stage I crack growth) was found to be
semi-elliptical in shape.
Fract~lre $~lrface Asperity Height Measurements
The fracture surface and the variation in the height of asperities on the fracture surface
were observed using a confocal scanning laser microscope. First, the laser beam was
centered on the area of the fracture surface adjoining the crack tip by direct observation
through an attached optical microscope. The fracture surface of an area of lmm 2 at the
crack tip was scanned by the laser beam and reflected to the detector. In order to make
a three dimensional profile of fracture surface asperities, a piezo-electric stage, was
used. The piezo-electic stage controls the distance between specimen and microscope.
This provides successive images of level contours of the asperities from their peaks to
their valleys. All images were combined to create an image of the configuration of the
fracture surface profile. Taking different slices through this profile and determining an
average value of fracture surface asperity height in each slice, revealed that the asperity
height is dramatically influenced by the magnitude of the compressive overload.
Crack Opening Stress Measurements

Using a pressurizing device attached to the CSLM system, the shear crack mouth
on a tubular specimen was opened and image processing was carded out for a variety of
stress levels. Optical sectioning from a post-image-processed crack provided crack
depth and crack mouth width at every point along the crack length for each load level.
The crack opening stress was taken as the stress level at which the crack depth stopped
increasing with increases in applied stress.
RESULTS AND DISCUSSION
Optical Sectionine for 3D Construction of Crack Depth and Asoect Ratio
Fig.3 represents the crack depths corresponding to four optical sections along a
crack length of 16 pm in A1 2024-T351. Slices shown in this figure correspond to depth
profiles at different points along the crack length. Slices taken along the crack length
are shown by LI, L2, L3 ..... and L10. These optical slices can be considered to be
//•rack depthend
"X-Zplane
10
2
N L1 =0p.m 1_2=2p,m L3=6pm L4=7,u.m L5=9.5p, m
0
10 .... r,,,
8
o
6
L) 4
2
L6=1 1.5p,m L7=1 3p,m L8=1 41J,m : L9=15.5pm L1 0=1 6lLtm
0
o 1 2 30 1 2 3 1 2 30 1 2 30 1 2 3
X(mn)
Fig.3--Optical sections obtained in a fatigued A1 2024-T351 specimen

at a series of locations along the crack length.
imaginary planes parallel to the X-Z plane (see Fig.3). Every imaginary X-Z plane
shows the crack depth profile at a point along the crack length (Y-direction). The
difference in height between the peak and the valley in the X-Z plot gives the crack
depth at that slice. The peak value in each slice was considered to be the average of
peaks at the crack mouth. The depths measured along the crack length (using CSLM) in
Fig.3 show that slices taken in the middle of the crack (Y-direction) correspond to the
largest crack depths. Using images obtained at the successive static load levels of the
pulling/pressurizing device, provides crack mouth displacement measurements in the
X-direction.
Fig.4 is a plot of the crack shape of two short fatigue cracks showing that the
cracks developed an approximately semi-elliptical front. Each datum point on this
semi-ellipse corresponds to an optical slice along the crack length obtained from 30
images at that point.
14 . ~ i, . ,'"'i i"", ~"'I . , , i ~ ~ . I , , , i . ~ , i i''',
Flat A12024-T351 specimen

9 Crack 1
9~ I0 [] Crack 2
Nr
8
2 J
e-- , , I , , . I , , ~ I , , , I , , , I . , , I , , , I , , , u
0 4 8 12 16
Distance along crack -Y axis (gm)
Fig.4--Profile of crack depth.

The crack length and depth were measured for several short fatigue cracks. The
results shown in Fig.5 exhibit a linear relationship between crack depth (a) and half
crack length (c).
16 ~' "' ' I ~'"' ~ i ' ' ' I ' ~- 'j I ' ' '
Flat AI 2024-T351 specimen /

.~ 12 O Experimental d a t a / O
/
r.)
4
a--0.83c
0 4 8 12 16 20
Crack half length c (~tm)
Fig.5-- Short crack depth versus surface crack length.

Crack lengths and depths obtained using surface layer removal by mechanical
polishing are given in Table 1. The a/c ratios varied between 0.74 and 0.95
(average=0.85) which is in good agreement with average aspect ratio of 0.83 obtained
using CSLM (see Fig.5). Crack depths measured using the two techniques (SR and
CSLM) showed good agreement (within 90-95%).
Monitoring the gradual shift of crack position with the removal of successive
layers from the specimen surface indicated that early crack growth (Stage I) occurred
on a plane inclined at about 45 degrees to the specimen axis. This angle is in agreement
with the inclination angle of 40-50 degrees to the axis of specimen (in Stage I)
measured using CSLM.
Table 1. Crack depth usin~ surface layer removal polishin~ method.

Crack No. 1 2 3 4 5 6
Crack length (2c) Bm 35 32 35 65 54 30
Projected crack depth (a*) ktm 10 10 10 22 14 10
Crack depth (a=a*/sin 45) I.tm 14 14 14 31 20 14
Aspect ratio (a/c) 0.80 0.87 0.80 0.95 0.74 0.93
Fracture Surface Asoeritv under Different Loadine Histories

A fractographic examination of the fracture surface of short fatigue cracks was carded
52 SENSING STRESS, STRAIN, AND D A M A G E IN M A T E R I A L S
out after breaking the notched 1045 Steel specimens in liquid nitrogen. An area of
lmm 2 located at the crack front (in the middle of specimen thickness) was scanned by
the CSLM. All cracks were of the same length when the specimens were broken and
had the same stress range for the small cycles in the load history but the stress range of
the overload cycles varied from specimen to specimen. A SEM photograph of the
fracture surface at the tip of a crack is shown in figure 6.
Fig,6--Fracture surface roughness at the tip of crack

(small arrows show the location of the fatigue crack front).
The height of plastically crushed asperities after the application of compressive

overloads was determined for each magnitude of overload. The average height of
asperities in notched 1045 steel specimens for constant amplitude loading and for
various levels of periodic compressive overloading is plotted versus the compressive
overload stress divided by the yield stress in Fig.7. The figure shows a trend of
decreasing asperity height as the magnitude of the compressive overload increases (-
300, -360, and -430MPa).
'' '1' ' ''1' ' ''1 ''' ' I''' '1'' '' I ''''/
1"2 f Notched 1045 steel specimen, n=50 "!
1,0
a:.~ 0.8
.~:~
q
< 0.4
O-
0.2 ,,,I,,,,I,,,,I,~ ,I~,~ ,I,,,,I,~,
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Compressive overload/cyclic yield stress
Fig.7-- Average asperity height divided by initial asperity height versus
compressive overloads.
VARVANI-FARAHANIAND TOPPER ON CSLM 53
The height of fracture surface irregularities near the crack tip decreased (from 28,
to 18, 13, and 8 pan, respectively) as compressive overload stresses increased (from
constant amplitude loading to -300, -360, and -430 MPa, respectively).
Figure 8 shows crack length versus number of cycle plots for cracks for various
magnitude of compressive overloads (-300,-360, -430 MPa). The overloads result in
faster crack growth and a shorter fatigue life.
1500 l I I Illll i I I I I IIII i I I I III
C.O.L=-360 MPa
C rest. Ampl.
.C C.O.L=-300 MPa
1000
P
500
} ~ 1 Notched
1045 Steel
specimen
0 I'i . , . .....i ' ' ' ' '~'' ' ' I I III
4 5 6
10 10 10
Log of Number of cycles
Fig.8-- Crack growth behavior under constant and periodic

compressive overloads of -300, -360, and -430 MPa.
Fracture surface profiles obtained using CSLM showed that a surface crack
initiates and propagates into the specimen a distance of 175-320 gm on a plane inclined
to the plane perpendicular to the applied stress. The plane of crack growth then shifts to
the plane perpendicular to the direction of the applied stress. It is noteworthy that the
height of the asperities (obtained by CSLM) during inclined crack growth (Stage I) is
less than for horizontal crack growth (Stage II). This may partly explain why short
fatigue cracks (Stage I) have a low closure stress. A careful observation of the fracture
surface of the specimen using a SEM verified that the fracture surface roughness during
Stage I fatigue crack growth was much less than that in Stage II crack growth.
Effective Crack Deoth and Crack Onenin~ Stress Measurements
A shear fatigue crack test was conducted to measure the crack closure stress and
effective crack depth of a tubular 1045 steel specimen using image processing and
optical sectioning of the crack being opened. A fatigued tubular specimen was
pressurized internally using a hydraulic device and examined using the CSLM system.
Hoop and longitudinal stresses due to internal pressure opened the crack. Images from
different levels of crack depth were obtained by changing the specimen height using a
piezo-electric stage. A set of confocal image slices at depth steps of lktm were
acquired. Post image processing were later used to combine all the images and create a
configuration of the crack depth.
A progressively increasing internal pressure gave rise to an increase in crack

opening displacement and in crack depth. A series of depths was measured as the
intemal pressure and displacement changed. The changes of crack depth versus the
applied internal pressure (hoop stress) is shown in Fig.9. Each datum point on this
figure corresponds to the deepest point along the crack length, obtained from a set of
confocal images, at an applied pressure (stress) level. The crack opening stress was
taken as the stress level at which the crack depth stopped increasing with increases in
applied pressure.
250 i i ! ! I i i t i I i I ! i I ! i I ! I | ""W ! A
- ~ ] - /"-/~ S oe =240 MPa I"

200
Shear strain amplitude--0.3% -
150 ' Shear crack length=80btm /" I-
9
N=25000cycles J I"
t2~
O 100 " aeff = ao+ ac~ Y i'
50
." ......
a~ k
l }/ acontact
---- ~ ~ ~
!'i
0~,, ,, ...... ,...TT....,..Tt

0 5 10 15 20 25
Maximum Crack Depth (btm)
Fig.9-- applied stress versus crack depth measurements.
CONCLUSIONS
1. Confocal scanning laser microscopy (CSLM) has been used to measure crack
depth and crack depth profile of short fatigue cracks of about 10-100btm in length using
optical sectioning of images obtained.
2. Both CSLM and SR techniques showed that the first stage of fatigue crack
growth occurs along an inclined shear plane at an angle of 40-50 degrees with the
specimen axis.
3. Confocal scanning laser microscopy measurements revealed that as the
magnitude of a periodic compressive overload increases, the height of surface
irregularities is reduced due to plastic crushing of asperities near the crack tip. This
resulted in faster crack growth and a shorter fatigue life.
4. Crack opening stress, taken as the stress level at which the crack depth
(effective crack depth) stopped increasing with increasing applied internal pressure,
was measured.
ACKNOWLEDGMENT-The first author wishes to thank the Ministry of Culture and

Higher Education of Iran for a scholarship.
REFERENCES
[1] Wagner, L., Gregory, J. K., Gysler, A., and Lutjering, G., "Propagation Behavior
of short Cracks in a Ti-8.6AI alloy," Small Fatieue Cracks. R. O. Ritchie and J.
Lankford Eds., TMS-AIME, Warrendale, PA. (i986), pp. 117-128.
[2]. Gangloff, R. P., Fatigue Crack Growth Measurement and Data Analysis. S. Y.
Hudak and R. J. Buc-ci,Eds., American Society for Testing and Materials,
Philadelphia, (1981), pp. 117.
[3] Vosikovsky, O., and Rivard, A., "Growth of Surface Fatigue Cracks in a Steel
Plate," International Journal of Fati~,ue, Vol. 3, (1981), pp. 111-115.
[4] London, B., Shyne, J. C., and Nelson, D. V., "Small Fatigue Crack Behavior
Monitored Using Surface Acoustic Waves in Quenched and Tempered 4140 Steel,"
The Behavior of Short Fatigue Cracks. K. J. Miller and E. R. de los Rios, Eds.,
T
Mechanical Engineering Publications, London, (1986), pp. 537-552

[5] Baxter, W. J, "Gel Electrode Imaging of Fatigue Cracks in Aluminum alloys,"
International Journal of Fatigue, Vol. 5, (1983), pp. 37-42.
[61 Varvani-Farahani, A., " Non-propagating Fatigue Crack Behavior in different

Orientation of Rolled A1 2024-T351" MASc Thesis (1994), University of
Waterloo, Waterloo, Ontario, Canada
Sanjay Tiku l, Norman J. Marchand 2 and Bhikhu Unvala3
AN ADVANCED MULTIPLE FREQUENCY ACPD SYSTEM FOR CRACK

DETECTION AND CALIBRATION
REFERENCE: Tiku, S., Marchand, N. J., and Unvala, B., " A n Advanced Multiple
Frequency ACPD System for Crack Detection and Calibration," Nontraditional
Methods of Sensin[g Stress, Strain, and Damage in Materials and Structures, ASTM
STP 1318, George F. Lucas and David A. Stubbs, Eds., American Society for Testing
and Materials, 1997.
ABSTRACT: On-line crack detection and crack depth measurements have been carried
out using a novel sensor. The sensor is based on an advanced alternating current potential
difference (ACPD) technique. This sensor makes use of a new technology called current
focusing (CF) which allows the current distribution on the surface to be selectively
confined to a precise region (e.g., in a high stress gradient domain). The use of AC
currents in conjunction with the current focusing technique results in having a three -
dimensional control of the current density distribution in a solid. As a result, the technique
can be used on large structures as well as on standard specimens. The associated advanced
electronics, having variable frequency capabilities (500 Hz to 240 kHz) and variable
currents up to 5 Amperes has been designed and will be briefly described. To assess the
performance of the system and ACPD/CF technologies, a" fatigue testing program was
carried out on single edge notch specimens made of A1 7075 alloys. Naturally initiated
fatigue cracks in the range of 200 lain to 6mm have been detected. Potential difference
versus crack depth calibrations curves were obtained for various combinations of currents
and frequencies with a reliability better than 99%. Furthermore, off-line measurements
have been carried out to assess the applicability of the technique for NDI applications.
KEYWORDS: ACPD, current focusing, aluminum alloy, crack detection, crack

calibration, on-line detection, NDI.
tGraduate Student, Department of Metallurgy and Materials Engineering, Ecole

Polytechnique, Montreal, Canada, H3C 3A7 and Research Engineer with AMKA
Technologies) V.P. R&D, AMRA Technologies Inc., 4700 de la Savane, Montreal,
Canada, H4P 1T7.
3 Director, Matelect Ltd., 33 Bedfords Garden, London, W8-7EF, U K
56
Copyright9
(PDVSA by ASTM
TIKU ET AL. ON A MULTIPLE FREQUENCYACPD SYSTEM 57
An important aspect of structural integrity assessment is the detection and sizing of

cracks. The damage tolerant design (DTD) methodology allows a crack to exist, the size
of which is assumed to be the minimum detectable crack size, al. Using the fatigue crack
growth rate properties of the material, a safety inspection interval is established so that the
crack does not grow to the critical size. A critical parameter in the DTD is a~, so that a
decrease in ai can increase the safety inspection interval and thus reduce the down time
due to inspections. Therefore, there is an increasing demand for measurement of cracks
with higher reliability and the need for crack growth rates to be predicted with a higher
accuracy. To meet the afore-mentioned requirements, an on-line strain and damage
measurement sensor based on an alternating current potential difference technique has
been designed and developed. This paper provides a brief description of the technology
and the experimental procedure used to obtain the crack calibration curves.
ACPD TECHNIQUE
The potential difference technique makes use of an increase in electrical resistance

in conducting materials due to strain and cracking. The distribution of an AC cur;~nt
injected in the solid is as shown in the figure (la). The current distributes itself in the
shape of a balloon in the plane XY and is confined to a narrow thickness in the plane XZ
[1].. The potentials across AA' are same because they are measured between two
equipotential lines. In the XZ plane, the narrow region (or the depth) where the current is
confined is called the skin depth 5, and is governed by the following equation [2]:
5 = ~/p / ~Jrc la~go f (1)

where
p = resistivity (ohm-meter),
= relative permeability,
go = absolute permeability ( 4re. 10.7 H/meter),
f = frequency of AC current (Hz).
Equation (1) indicates that for a given material, the skin depth is inversely
proportional to the square root of the frequency of the applied current. Therefore, if we
increase the frequency of the applied current, the skin depth decreases. This in turn means
that the current density distribution near the surface increases leading to a higher
sensitivity to surface breaking cracks and applied stress or strain. For aluminium alloy
7075T651, the skin depth (8), is estimated to be around 3.6 mm for a 1 kHz AC current
and about 0.235 mm for a 240 kHz current frequency. That is why there is always a need
for high frequency electronics to achieve maximum advantages of ACPD.
Figure (lb) shows that the distribution of current density below the surface and
within the skin depth, 8 is not uniform but it decreases exponentially becoming almost zero
(0.0025 of the surface value) at a distance of six 5. In fact, the skin depth is termed as the
distance from the surface where the current density has reduced to around 36.86% of the
surface value. Due to this skin effect, the sensitivity to crack sizing is not affected by the
material thickness as long as the material or component is thicker than six 5.
gqnipotentials
A ( A I Streamlines. z
I / / II I1 I
/ , . , , ~/
(a) ~_Z / - ~ ,,./ ~

1.00
- 37% of current density at surface
ential Curve
(b) 0
8 28 38 48 58 65
Z -'-->
Fig. 1 - (a) AC current distribution in a solid. Co) Skin effect.
Since the current is confined to a narrow region, much smaller amounts of current
are required as compared to DCPD methods. Depending upon the size, materials with a
relatively high resistivity such as alloy steels and super alloys will require approximately 10
to 30 amperes of DC current and materials of lower resistivity such as aluminium alloys
can require 50 to 200 amperes [1,3]. Such high currents are difficult to handle and can
create serious problems of specimen heating. It may still be very difficult to measure strain
for low resistivity materials such as aluminium alloys with a DCPD technique. However in
ACPD, depending upon the requirements, very small currents ranging from 500 mill-
amperes to 5 amperes can be successfully used to measure short cracks, stress and strain
in steels, super alloys, low resistivity materials such as aluminium alloys etc. On the other
hand ACPD requires sophisticated electronics [4].
TIKU ET AL. ON A MULTIPLE FREQUENCY ACPD SYSTEM 59
CURRENT FOCUSING
Although, the sensitivity of ACPD is not affected by the thickness of the material,
the distribution of the AC current in the plane XY indicates that it will be affected by the
surface area of the material. The higher the surface area XY, the lower will be the current
density distribution, since the same amount of current is flowing in a large area. This
results in a lower sensitivity to the cracks and stress (or strain). A novel technology,
named current focusing, has been developed to overcome this problem [5]. With the help
of this technique, the current distribution can be confined to a narrow region rather than
being spread over the whole area. Current focusing is based on the principle that when a
conductor carrying an AC current is placed very near the return conductor, e.g. a plate,
then the distribution of the AC current in the return conductor is confined and tends to be
a mirror image of the current flowing in the upper conductor. The modified current
distribution in the return conductor is as shown in the figure (2a). In this case, the
electrical field is now concentrated between the conductor and the plate. The magnetic
field generated by the AC current flowing in the conductor forces the current in the plate
to flow below the conductor. A cross sectional view of the current focusing effect is
shown in figure (2b). The current can therefore, be constrained or focused in a region of
the plate near the conductor, and the effect can be maximized by bringing the conductor
(which from now on will be called the current focusing device) as near the plate as
possible. This means that the potential measurements across AA' are no longer the same
because the current distribution (instead of being spread out as balloon shape) will now be
focused below the current focusing device leading to an increased sensitivity to cracks,
straining etc.
An experiment was carried out to verify the current focusing effect. A current
focusing device was installed on an Al 7075 plate. Measurements of the potential at
various current frequencies (30, 60, 90, 120, 150, 180, 210 and 240 kHz) were carried out
near the current focusing device and away from the device in steps of one mm. The results
are shown in figure (3). Note that the potentials drop as we move away from the current
focusing device. Without the current focusing device, the potential would have been the
same at all points of measurement (see Fig. 1a). Thus, the control of the frequency of the
AC current combined With the current focusing technique results in having a three
dimensional control of the current density distribution in a solid. Taking into account the
complex geometries which are often encountered in real life structures (e.g. holes,
notches, corners, sharp bends etc.,), the ability to steer or focus the current in the area of
interest, i.e., where the cracking is expected, is of paramount importance. The other
significant advantage is that the ACPD along with current focusing can be used on
standard specimens as well as on large structures without significant change in the
calibration curve.
Conduc~/,~ X ~Y
Currtn!Cable
r
/
RegionofCurrenlFlow
l
(a)
M a g n e t ~
~ ElectricalField
(b)
Fig. 2 - (a) Modified current distribution due to current focusing. (b) Schematic of current focusing
effect.
70
9 30 kHz
\ 60 kHz
60 A,,~ 9 90 kHz
50 \ \ 9 210 kHz
\\ \o * 2*OkHz
E 40 \\%\
1:3
D- 30
O
<C
20
10
0
0 3 6 9 12 15
Distance from the current focusing device (rnrn)
Fig. 3 - A plot of ACPD versus probe distance from the current focusing de~ice.
ACPD ELECTRONICS
The electronics designed and developed is an advanced two channel ACPD system
named CGM 6-3. A simplified block diagram of the ACPD system is shown in the figure
(4). The voltage controlled oscillator, V.C.O., generates a sinusoidal wave form of
continuously variable frequency with a high stability of frequency and amplitude (< 0.1%)
and very low distortion (<0.2%). This waveform is fed to the voltage to current power
amplifier. The power amplifier generates a constant current (0.1% stability) of five
amperes at frequencies of up to 240 kHz. The signal from the specimen is initially
amplified by a specially designed transformer with a turns ratio of ten. The transformer has
the unique advantages of low noise, a very high common mode rejection ratio and
isolation. This is followed by a series of integrated circuits which can be selectively
switched in to give gain3 of 1000, 3000, 10 000 and 30 000. The pre-amplification and
final amplification is carried out in a separate shielded enclosure which is mounted near the
specimen with the main instrument several feet away. The switching waveform generator
involves the generation of two square waves, one in phase with the sine wave and one at
90~ with the sine wave. The signal is de-modulated by lock in switching or synchronous
detection to produce two DC signals, one corresponding to the resistive part (0~ and the
other corresponding to the inductive part (90~ After de-modulation the signal is
processed through a bank of filters. The filters can be selected from one of the preset
values of 1 Hz, 80 Hz, 1 kHz and 20 kHz or can be switched out completely. The filters
are followed by four off-set circuits and can be switched from zero to 5 volts using coarse
and/or fine step controls.
The resistive part or 0~ component is monitored all the time during testing. This
has several advantages over the continuous measurement of the magnitude of the ACPD
signal. Under ideal conditions, the ACPD signal OA, as shown in figure (5) is at 45~ with
respect to the phase of current passing through the specimen. At higher frequencies, the
inductive pick up becom'es higher [3,6,7]. The inductive component also depends on the
size of loop of the current and voltage cables and movement of cables [6,8]. Due to the
inductive pick-up, the actual ACPD signal now would be OB rather than OA. Thus, the
magnitude of ACPD signal varies with all these external factors which are not associated
with crack, stress, etc. However, the 0~ component, (OC) remains the same as indicated in
the figure (5). As a result, the signal (0~ to pick up voltage ratio is much higher and the
movement of the cables has little effect on the resistive component.
EXPERIMENTAL PROCEDURES / ON-LINE MEASUREMENTS
A single Edge Notch (SEN) specimen geometry as shown in figure (6) was selected for
use in carrying out crack detection and calibration measurements during fatigue tests. The
SEN geometry was selected because it reflected a complicated version of several locations
being monitored on a CF-18 aircraft. The notch geometry (Kt = 2.885) also represents the
bolt holes of a compressor disk. No special surface finish is required for the surface. The
materials used for the study were A1 7075T651 and A1 7075T7351. The cracks which
were monitored were naturally initiated fatigue cracks. The probe set up included a
current focusing device glued in the middle of the specimen and following the contour of
the notch along the specimen length. The potential or voltage probes were welded across
the notch and the current focusing device, thus giving an effective probe spacing of 7.5 - 8
11oI ..-.-L ................................................................................................

V o f .....................
I
Constantcurrent
V~A ( f ) - - - - . - i - - ~
0 - 5 Amperes
Switching 500 Hz - 240 KHz
Waveforrn
Vif " ' " - " 1 uu,~,m= i 9 ~ g. . . . . tor
90~/ O*
~out.o.-. .................................. / 1
L__
VoYgO'-- - ~
Synchronous amphfier I X
Detector
Fcut-off ...~............................... ,
cif ...... ~ .............. ! i
in/out._~.__, ';
i ! i i
VoXgO'.. ~ " 90" /
Vo XO'_._i..~$wJtched filtersy~
!
09
in/out - --.i- - : ' Channel X,
GainX-': ................................................................................
Fig. 4 - A simplified block diagram of ACPD system

900,
B
....... - / ~ ACPDafter inductivepick-up
/ I---* Inductivepick-up
45~ iV Phaseof Current

0 C 0o
Fig. 5 - A schematic view of ACPD component.
TtKU ET AL. ON A MULTIPLE FREQUENCYACPD SYSTEM 63
mm. All the probe attachments (for both current and voltage) were performed using an
ultrasonic welding technique, which is detailed elsewhere [9]. The fatigue testing w a s
carried out at a stress ratio (R) of zero and frequency of 1 Hz. The remote far field stress
applied varied between 131 MPa to 275.7 MPa for different tests. The output signals
from the CGM6-3 were digitized and recorded using digital voltmeter (DVM/Hewlett-
Packard model 34401A). The data acquisition was carried out at 100 PLC with auto zero
correction in the D\rM. This gave a sampling rate of one reading every 3.34 seconds. This
low sampling rate was selected because the inherent (internal) noise of the DVM is too
high when using faster sampling rates. This problem is now overcome using a faster DVM
with a high noise rejection ratio. The new system is presently being used for on-line
monitoring of strain and cracking on F-18 aircraft undergoing full scale testing and is
detailed elsewhere [10].
t
.......... 3/4"
..41
1/2-20 13'NF 4 places
- i,- ,,:(i--
4 places
Fig. 6 - The detailed geometry of the SEN specimen used in the testing program
RESULTS / CRACK CALIBRATION
A typical response of ACPD ( which in our case is always 0~ component) versus time
during fatigue cycling is shown in figure (7). As can be seen, the potential readings are
very stable in the beginning. However, there is an increase in the potential as soon as the
crack is detected by the ACPD probes. Several experiments were carded out to determine
the ACPD/crack response at various frequencies. Each calibration curve was obtained by
fatiguing several samples. The samples were fatigued until a pre-determined change of
potential was reached. Each specimen was then pried open (by applying a tensile load) and
the crack depth measured using scanning electron microscopy. The dominant crack depth
(a), and the corresponding change in potential (SV) were then correlated to obtain the
calibration curves at 30, 90 and 120 kHz frequencies while current and gain were fixed at
5 amperes and 3000 respectively. Second order polynomials were fit to data yielding the
following calibration equations for 30, 90 and 120 kHz frequencies respectively.
8V = 7.0 a 2 + 25.906 a - 5.092 (2)

r z = 0.996 and 0.2 _<a -< 6 ram.
5V = 43 39 a: - 0.614 a - 0.016 (3)
r2= 0.997 and 0.2 _<a <__0.43 ram.
5V = 25.42 a2 + 20.546 a - 0.825 (4)
r 2 = 0.999 and 0.2 _a < 5.32 ram, where r2 = reliability of regression.
Figure (8) shows plots of the crack depth versus change in potential for the above three
cases.
0.03
' • 0.02
>
C3
0.01
_.__;
0.00
2500 5000 7500 10000 12500
N u m b e r of Cycles
Fig. - 7 A typical response of ACPD versus number of cycles.
The calibration equations can be used to convert the ACPD versus time plots into
crack depth (a) versus number of cycles (N). Figure (9) shows the plot of crack depth
versus number of cycles derived from figure (7). As can be seen, crack growth ranges
from 401am to 4001am Thus, ACPD can be effectively used for studying short cracks and
for the characterization of fatigue properties of different materials. Finally it is noteworthy
that resistivities of the aluminium alloys were in the range of 50 nano ohms meter( nf~-
meter). Resistivities of other commonly used alloys such as titanium based alloys, nickel
based alloys and steels are much higher than that of aluminium alloys. As a result, the
sensitivity of ACPD to crack depth is much greater for these other alloys than it is for
aluminium alloys (as reported here).
A quadratic relationship was observed for correlating crack depth versus change
in potential for the three sets of data. This non linear behaviour can be attributed to the
fact, that the current is not uniform through the thickness of the solid but drops
exponentially ( Fig. lb) from the surface to the inside. Hwang and Ballinger [7] and Lu et
450 8.0
50 kHz, 5 Amp. 3000 Gain 90 kHz, 5 Amp, 3000 Oaln 9
AI 7075 T651 f At 7075 T651 /I t
r
375
t
6.0 i
300 I
i i
E /
v 225 ~ 4,r
p9
J
150
2.C
75
#
0.(
o.o o.~ o.z 0.3 o.* 0.5

0 1 2 3 4 5 6
Crock Depth,o (turn) Crock Depth, a (mm)
900
120 kHz, 5 Amp, 3000 Gain
AI 7075 T7351 ,a
800
/
700 /~
/
600
E 500
400
o9
oO
300
~
200 ~176
.o,~176
100 o.O~
.~.o~176176176
0
0 I 2 ,3 4. 5 6
Crock Depth, a ( r a m )
Fig. 8 - Calibration curves : crack depth versus potential difference.
al [11] have observed the same effect and reported a second order polynomial and power
law relationship respectively. However, this is in contrast with the observations made by
Di et al [12], Charlesworth and Dover [6], Jablonski [13] and Watt [4]. They reported a
linear relationship between crack depth and change in potential. Therefore, a two
dimensional finite element analysis was carried out to sort out the effect of crack growth
on the change in potential. Details of the analysis are given elsewhere [14]. Figure (10)
reports the result of the analyses. This plot shows a non linear relation for crack depth
(normalized with respect to skin depth) versus change in potential. The non linearity
comes in when the crack depth approaches the skin depth. However, the relation is almost
linear when the crack depth is greater than the skin depth.This is consistent with the
observation made by Verpoest et al [3].
0.40
0.35
E 0.30
v
E
t-.
0.25
0.20
0.15
0.10
t~
0.05
, J . ,
0.00
10000 11000 12000 13000
Number of Cycles
Fig. 9 - - A plot of crack depth versus number of cycles
0.040
0.038
~ e*
E
,~, 0.036
p"
.~149
>
*o ~
it""
0.034
0.032
0.0 0.5 1.0 1.5 2.0
Crack Depth (0)
Fig. I0 - FEA result : Crack depth versus potential difference.
OFF-LINE MEASUREMENT
Interrupted ACPD measurements were carried out during fatigue testing. This
was done to evaluate the capability of detecting and sizing cracks using discrete
measurements of potential during the fatigue life of components. Figure (1 I) shows a plot
of ACPD versus time, where interruptions in reading the AC potentials can be seen. At
point A the cycling was stopped, the system brought down to zero load and readings
T1KU ET AL. ON A MULTIPLE FREQUENCY ACPD SYSTEM 67
taken for some time. As can be seen, the readings have dropped. At point B, the CGM6-3
was disconnected and the readings (as expected ) dropped to zero (location C). The
electronics was then connected back and the cycling resumed. The readings came back to
their original value ( the difference being less than 0.1 mV). At points D, E, F and (3,
cycling was stopped again, the system brought down to zero load, readings taken for some
time and then the cycling resumed, It can be seen that the potential dropped every time the
loads were removed but returned to their original values as soon as the cycling was
resumed. This experiment was stopped when the potential difference (PD) during cyclic
loading reached 24.2 inV. Note that the change of 242 mV was also observed at zero
load, This experiment confirms that the present ACPD/CF technique is capable of
detecting and sizing cracks with and without the external loads. The interruptions did not
affect the ACPD responses to crack growth. Since there is no change in the ACPD value
due to interruptions, it can be concluded that ACPD/CF can be used to inspect parts
removed from service and help in making decisions about their future usage.
0,6 0,550
0,5 0.545
0.540
0.4
"6
>,~ 0.3
0.535
T
24.2
(3 0,530
rt
(3 0.2
<
c (*;
0.525
0.520
(b)
1
0.0 0.5%5
5000 loooo 15ooo 20000 25000 5000 10000 15000 20000 25000
Time (see) Time (sec)
Fig. 11 - (a) A plot of ACPD versus time. (b) Zoom of(a).
In some experiments, the frequency was scanned from 30 kHz to 240 kHz before
the fatigue test. The fatigue test was then carried out and the testing stopped at pre-
determined change of potential due to cracking. The system was then brought down to
zero load and the frequency was scanned again from 30 kHz to 240 kHz. The resulting
potential differences before and after the fatigue test (at various frequencies) are shown in
figure (12). As expected, the sensitivity to crack detection increases with the increase in
frequency, i.e. for the same crack depth the PD at 240 kHz is much higher than at 30 kHz.
As can be seen, the results are not affected by the fact that the measurements are taken at
zero load. This further adds to the proof that ACPD can also be effectively used as an
off-line or as a Nondestructive inspection (NDI) tool for crack detection and sizing.
15
14! o -- 0.8 mm
13
>
12
J
1t
10
>
1o
9
8
7
6
5 -- 9 ,,,i ....... j , ,,,,
3o 90 12o 21o 2,0

Frequency (kHz)
Fig. 12 - A plot of frequency versus change in the potential due to crack.
CONCLUSIONS
9 An advanced ACPD system combined with current focusing has been developed to
detect and size cracks for both on-line and off-line (NDI) applications.
9 The effectiveness of current focusing has been demonstrated. A three dimensional

control of current density distribution was achieved.
9 Crack calibration curves (quadratic functions) detailing crack depths versus potential
difference have been obtained at three different frequencies with the reliability of the
regression better than 99%. The experimental results are in agreement with the finite
element calculations of the surface potential as a function of crack depth.
9 The technique can be used to characterize naturally initiated short crack ( ___40 1am)
growth behaviour of materials.
ACKNOWLEDGEMENTS
This work was performed under contract from the Department of National Defence
(DND) of Canada (Contract - DSS 18svo. W8477 - 0 - AC59).
TIKU ET AL. ON A MULTIPLE FREQUENCYACPD SYSTEM 69
REFERENCES
[1] Dover, W.D., Chai'lesworth, F.D.W., Taylor, K.A., Collins, R., and Michael, D.H.,
"A.C. Field Measurement - Theory and Practice," The Measurement of Crack
Length and shape during Fracture and Fatigue, C. J. Beevers, Ed., Engineering
Materials Advisory Services LTD., West Midlands, UK, 1980, pp 222- 260.
[2] Kraus, J.D., Electromagnetics, McGraw Hill, Third Edition, 1984, pp 449-450.
[3_] Veropest, I., Aernoudt, E., Deruyttere, A., and Neyrinck, M.," An ACPD Method
for Detection and Measurement of Surface Microcracks during Fatigue Testing of
Wires," Advances in Crack Length Measurement C. J. Beevers, Ed,, Engineering
Materials Advisory Services LTD., West Midlands, UK, 1982, pp 295-322
[4] Watt, K.R., "A consideration of an A.C. Potential Drop Method for Crack Length
Measurement," The Measurement of Crack Length and shape during Fracture and
Fatigue, C. J. Beevers, Ed., Engineering Materials Advisory Services LTD,, West
Midlands, UK, 1980, pp 202-221.
[5] Unvala, B.A., U.S. Patent No. 5,202,641, 13 April, 1993.
[6] Charlesworth, F.D.W., and Dover, W.D. " Some Aspects of Crack Detection and
Sizing using A.C.Field Measurements" Advances in Crack Length Measurement C.
J. Beevers, Ed., Engineering Materials Advisory Services LTD., West Midlands,
UK, 1982, pp 253-276.
[7] Hwang, I.S., and Ballinger, R.G., " A multi-frequency AC potential drop technique
for the detection of small cracks," Measurement Science Technology/, Vol. 3, 1992,
pp 62-74.
[_8] Syahkal, D.M., Collins, R., " Induction Effects during Crack Measurement by the
AC Field Technique", Advances in Crack Length Measurement C. J. Beevers, Ed.,
Engineering Materials Advisory Services LTD., West Midlands, UK, 1982, pp 277-
294.
[2] Tiku, S., Schulze, H.W., and Marchand, N.J., " On-Line Strain and Damage
Measurement System", Second Annual Report FAPS 91P17, Department of
National Defence, Canada, 1993.
[10] Tiku, S., Marchand, N.J., and Finlayson, R.D.," An Advanced ACPD System for
Aircraft Structural Integrity Assessment", 1995 Proceedings of USAF Structural
Integrity Program Conference, (in Press).
[11] Lu, Z.J., Nicholas, P.J., and Evans, W.J., " Calibration of an ACPD Monitoring
System for Small Crack Growth in Corner Crack Specimens," Engineering
Fracture Mechanics, Vol. 50,No. 4, 1995, pp 443-456.
[1__2] Dai, Y., Marchand, N.J., and Hongoh, M., "Fatigue Crack Growth Measurements
in TMF Testing of Titanium Alloys using an ACPD Technique," .Sp.ecial
Applications and Advanced Techniques for Crack Size Determination STP 1251,
J.J. Ruschau and J.K. Donald, Eds., American Society for Testing and Materials,
Philadelphia, 1995, pp 17-32.
[1__3] Jablonski, D.A., "Measurement of Multiple- Site Cracking in Simulated Aircratt

Panels using AC Potential Drop." Special Applications and Advanced Techniques
for Crack Size Determination STP 1251, J.J. Ruschau and J.K. Donald, Eds.,
American Society for Testing and Materials, Philadelphia, 1995, pp 32-50.
[].~4] Tiku, S., and Marchand, N.J., "On-Line Strain and Damage Measurement
System", Third Annual Report FAPS 91P17, Department of National Defence,
Canada, 1994.
Seung H. Nahm I and Chang M. Suh 2
O b s e r v a t i o n on the G r o w t h B e h a v i o r of Small Surface Cracks U s i n g

R e m o t e M e a s u r e m e n t System
REFERENCE: Nahm, S. H. and Suh, C. M., " O b s e r v a t i o n on the Growth

B e h a v i o r of Small Surface Cracks U s i n g R e m o t e M e a s u r e m e n t System,"
Nontraditional Methods of SensinQ Stress, Strain, and Damaae in
Materials and Structures, ASTM STP 1318, Geoge F. Lucas and David A.
ABSTRACT: A new experimental system which uses a remote measurement

system(RIMS) constructed by means of an image processing technique was
applied to the observation of small surface fatigue crack behavior in
ICr-IMo-0.25V steel and type 304 stainless steel at room temperature as
well as at elevated temperature. This system consists of a long distance
focusing microscope, a CCD camera, a light source, a 3-axis controller,
a personal computer and a data translation card. The resolution of this
system is I.I ~m. This system allows one to observe the surface crack
length down to 30 ~m. Using this system, the growth rates of small
surface cracks on smooth specimens were studied in terms of stress
intensity factor range and compared with those of long cracks.
~YWORDS: remote measurement system, image processing technique, small

fatigue crack, iCr-iMo-0.25V steel, type 304 stainless steel, stress
intensity factor range
The size of pre-existing defects, such as inclusions, in materials is

getting smaller due to the improvement of steel making processes.
Experimental studies on the growth behavior of small cracks were
recently conducted because the growth life of small cracks initiated
from pre-existing defects occupies a large portion of total fatigue life
[!-~]. Many researchers [2,~,~] have recognized the difference between
the growth behavior of long cracks and that of small cracks. As a
result, advanced experimental methods to observe the growth behavior of
small cracks have been developed.
The replica method [i,~] is widely used to measure small cracks
because it has many advantages in application. However, the method is
labor and time intensive because the replication procedures must be
repeated with crack growth. Also, the replication method is not possible
for elevated temperature tests. Scanning electron microscopy (SEM) is a
precise method, but specially designed equipment is required for its
usage [!]- Therefore remote sensing techniques such as,
iSenior researcher, Korea Research Institute of Standards and

Science, P.O. Box 102, Yusong, Taejon, 305-600, Korea.
2Professor, Kyungpook National University, 1370 Sankyuk-dong,

Pook-ku, Taegu, 702-701, Korea.
71
(PDVSA LosbyTeques)
Copyright91997 ASTM pursuant to License Agreement.
International www.astm.orgNo further reproductions authorized.
photomicrosoopy[8], interferometry[~], and electrical potential[10] are

desirable to measure small cracks. Some drawbacks to these techniques
are that photomicroscopy and interferometry are not automatic measuring
techniques, and the electrical potential technique must be calibrated
before each test. Furthermore interferometry and electrical potential
techniques do not provide observation of crack behavior. Therefore, a
new measuring system is requested to observe the growth behavior of
surface fatigue cracks and measure the crack length simultaneously. In
this paper, a new remote measurement system (RMS) for the investigation
of surface fatigue cracks is introduced. The RMS accompanied with an
image processing technique successfully observed small surface fatigue
cracks initiated and grown on the surface of ICr-IMo-0.25V steel and
type 304 stainless steel. In addition, the growth rates of surface
cracks measured by RMS were analyzed in terms of the stress intensity
factor range.
EXPERIMENTAL
Setup of remote measurement system
The RMS consists of a long distance focusing microscope, a light

source, a three-axis controller, a monitor, a digital indicator, a CCD
camera, a personal computer, and a data translation card(image card).
Figure 1 depicts the schematic diagram of the system. The image on the
monitor is obtained by the CCD camera attached to the microscope whose
focusing distance is approximately 840 mm, and magnified up to 480• A
unit cell size of the CCD camera is 8.4 Bm x 9.8 p~n. Thus, each
individual pixel of the input image represents 0.7 p2n x 0.8 ~m in
actuality. The resolution of the increment of crack extension is i.i ~m.
As a result, cracks larger than 30 Bm can be distinguished by this
system. For high temperature tests, an electrical resistance furnace
FIG. 1--Schematic diagram of experimental setup of remote measurement

system.
NAHM AND SUH ON A REMOTE MEASUREMENT SYSTEM 73
equipped with a transparent window is utilized. Image data of cracks are

stored into the computer for further processing after the test. The
required m e m o r y for one image frame is about 0.3 Mbyte.
The RMS was calibrated to confirm measurement accuracy. During the
calibration, a light source was installed with proper intensity and
angle, and a 200 mm long ruler was measured. The accuracy of the RMS was
confirmed by comparing a scale on the ruler with the value read from the
digital indicator of the RMS. The mean value of the four measurements of
the 40 mm length was 39.68 mm. Therefore, the error between the actual
length and the m e a s u r e d value was - 0.8%.
The accuracy of various measuring methods was also compared with
respect to a small defect (diameter = 180 ~m) made by an electro-
discharge machine. Table 1 contains values m e a s u r e d by scanning electron
microscope, optical microscope, RMS, and replica methods. The scanning
electron microscope had the best accuracy, and the RMS provided accuracy
almost as good.
The RMS allows automatic observation and measurement of surface cracks
at elevated temperature or in an inert gaseous atmosphere. The
measurement and observation can be done after the test since the growth
behavior of the crack is recorded continuously on the computer during
the test.
TABLE 1--Comparison of a c c u r a c y among the

four m e a s u r e m e n t methods.
SEM Microscope RMS Replica
180~m 176~m 179~m 171~m
Test m a t e r i a l s and test p r o c e d u r e
The test materials were iCr-iMo-0.25V steel and type 304 stainless
steel which are widely used for turbine rotor and high temperature
components. The chemical compositions are contained in Table 2. The
average grain sizes of both materials, as m e a s u r e d by the line intercept
method, were 25 ~/n and 30 ~m respectively. Tensile tests were conducted
at room temperature and elevated temperature. Tensile specimens with a
length of 41.9 mm were machined to a diameter of 6.35 mm. Tensile tests
were performed at i0 mm/min according to ASTM E 8 [II] with a closed
loop servohydraulic testing machine. Hardness tests were conducted with
a micro-Vickers hardness tester. Mechanical properties of the test
materials are contained in Table 3.
The fatigue specimens, prepared as shown in Fig. 2, were designed to
examine the initiation and growth behavior of small surface cracks in
the specimen gage section. The specimen radius of curvature provided a
sufficient flat region and a well-focused image for the RMS system. In
the type 304 stainless steel, an artificial defect (diameter 150 ~m and
depth 70 ~m) was introduced in order to initiate a crack. The defect on
the surface of specimens was machined by Nd:YAG laser.
The fatigue tests of iCr-iMo-0.25V steel were p e r f o r m e d at room
temperature in ambient atmosphere and also at 538~ in an argon gaseous
environment. The type 304 stainless steel tests were p e r f o r m e d at room
temperature and at 538~ in ambient atmosphere only. The fatigue
conditions of all tests were I0 Hz frequency, a 0.05 load ratio and a
sinusoidal waveshape. When ICr-iMo-0.25V steel was tested at elevated
temperature, argon gas flowed into the electrical resistance furnace at
TABLE 2--Chemical composition of test materials (Wt,%).
Material C Si Mn P S Ni Cr Mo V As Sn Sb
Cr-Mo-V 0.31 0.23 0.76 0.006 0.001 0.36 i.ii 1.32 0.27 0.006 0.005 0.001
304S.S. 0.07 0.35 1.89 0.032 0.014 8.50 18.0 -
TABLE 3--Mechanical properties of test materials.
Material Temperature Yield Tensile Elongation reduction Hardness

stress strength of area
(~ (MPa) (MPa) (%) (%) (Hv)
Cr-Mo-V 24 665.2 823.1 18.8 59.4 266

538 533.5 580.6 22.7 55.5 -
304 S.S. 24 313.8 647.2 57.0 72.0 179
538 208.0 505.2 40.8 66.5 -
R28 unit : m m
f19 R18
9q-i A
i
1
=_15,_ 19.! 30 _I_ 19_1_15,i

......
i. 98 .L
FIG. 2--Geometry of surface fatigue specimen.
a rate of 5 I/min in order to prevent oxidation. The crack length was

m e a s u r e d at the 80 % of m a x i m u m fatigue load while the cyclic loading
was interrupted. The image of the crack was a c q u i r e d and stored into the
p e r s o n a l computer, d u r i n g the interruption, for further p r o c e s s i n g after
the test.
Image of surface crack at elevated temperature
Figure 3 contains images of a surface fatigue crack on a iCr-IMo-0.25V

steel smooth specimen at 538~ in argon gaseous environment. Figure 4
contains images of a surface fatigue crack in type 304 stainless steel
at 538~ in ambient atmosphere. Although little oxidation occurred on
the surface, good images could not be obtained because of the convection
of air in the furnace. As a result, the image quality obtained by RMS
was worse than that by SEM. Consequently the CCD camera had to be
replaced by a digital camera in order to improve the image.
Initiation of a small crack in 1Cr-lMo-0.25V steel
Cracks in iCr-IMo-0.25V steel were initiated from inclusions, which

could be classified into two groups depending on the initiation site.
The first group of cracks were initiated from the inclusions on the
specimen surface and the second were initiated from the inclusions
beneath the surface. Figure 5(a) represents a SEM fractograph of a ICr-
iMo-0.25V steel specimen fractured by a 447 MPa fatigue load at room
temperature as one example belonging to the first group. Many small
inclusions on fracture surface are apparent. This implies that the
cracks were initiated from these inclusions. The chemical composition of
matrix and inclusion are contained in Table 4. The composition of the
inclusion is very different from that of the matrix. About 71% of the
composition of the inclusion is AI, and the major element of the matrix
is Fe, being 93%. This means that the inclusion is mainly composed of
aluminum oxide.
Figure 5(b) represents a SEM fractograph of a iCr-iMo-0.25V steel
specimen fractured by a 373 MPa fatigue load at 538~ as one example of
the second group. Figure 5(b) indicates that most of the inclusions
gathered within a slender band which was located beneath the specimen
surface and skewed by 30 degrees with respect to the specimen surface.
Most of the crack initiations occurred from these bands. The fatigue
life of specimens with cracks initiated beneath the surface (Fig. 5(b))
is relatively longer than that of specimens with cracks initiated on the
surface (Fig. 5(a)). Therefore the fatigue life is affected by the
location of crack initiation than sizes and numbers of inclusions.
TABLE 4--Comparison of chemical compositions of an

inclusion and matrix
Component Inclusion (Wt,%) Matrix (Wt,%)
A1 70.8 -
Ca 17.3 -
Fe 6.9 93.5
P 4.0 -
Ti i. 1 -
Mn - 3.2
Cr - 2.3
Mo - 1.0
Si - 0.I
FIG. 3--Representative i m a g e s of a crack for I C r - I M o - 0 . 2 5 V steel smooth

s p e c i m e n at 538~
(a) 3 6 5 , 0 0 0 cycles, 2a = 0.264 mm,
(b) 3 7 2 , 0 0 0 cycles, 2a = 0.610 mm,
(c) 3 7 7 , 0 0 0 cycles, 2a = 1.010 mm,
(d) 3 8 5 , 0 0 0 cycles, 2a = 1.978 mm.
FIG. 4--Representative images of a c r a c k for type 304 stainless steel

specimen with an a r t i f i c i a l d e f e c t at 538~
(a) 1 , 6 4 0 , 0 0 0 cycles, 2a = 0.306 mm,
(b) 1 , 7 1 4 , 0 0 0 cycles, 2a = 0.626 mm,
(c) 1 , 7 2 6 , 0 0 0 cycles, 2a = 0 . 9 5 9 mm,
(d) 1 , 7 4 9 , 0 0 0 cycles, 2a = 1 . 7 1 9 mm.
FIG. 5 - - S E M f r a c t o g r a p h s of I C r - I M o - 0 . 2 5 V steel s m o o t h specimen: (a) in

case of crack i n i t i a t e d on the surface, (b) in case of crack
i n i t i a t e d b e n e a t h the surface.
Relationship between the fatigue crack length and the cycle ratio
Figure 6(a) contains surface crack length versus cycle ratio data for
ICr-IMo-0.25V steel at room temperature and at 538~ The data lie
within a relatively narrow band regardless of stress level and
temperature. These results have the same tendency as those of other
reports [~,12]. They indicate that more than 80% of the total life is
consumed for the crack to grow up to 800 ~m. Cracks larger than 800 Mm
grow relatively fast and eventually lead to dynamic fracture. Therefore
the critical crack size which distinguishes between small and large
cracks seems not to change with different stress levels and
temperatures. This implies that the understanding of the small crack
growth on the smooth specimen m a y be more important for life prediction
than the large crack growth.
Figure 6(b) contains surface crack length versus cycle ratio data for
type 304 stainless steel specimen at room temperature and at 538~
While the data points at room temperature fall on one line regardless of
stress level, those at 538~ don't fall on that line unlike ICr-IMo-
0.25V steel. This indicates that about 70% of the total life is consumed
for the crack to grow up to 800 Mm at room temperature and about 95% of
the total life is consumed at elevated temperature.
Variation of the aspect ratio
The crack length in the depth direction, as well as on the surface,

should be considered to understand the characteristics of a small crack
in ICr-IMo-0.25V steel. However, the aspect ratio (a/c) of the crack
could not be measured directly by using the RMS because the RMS can
observe the crack on the specimen surface only. In order to measure the
aspect ratio, several specimens fatigued to different surface crack
lengths were fractured under monotonic loading at low temperature. The
crack lengths on the surface and in the depth direction from the
fractured surface were m e a s u r e d by an optical microscope.
Figure 7 presents the aspect ratio(a/c) versus the ratio of crack
depth to the specimen thickness(a/t). The aspect ratio is about 1.3 at
low values of a/t, decreases until a/t becomes 0.6, and converges to
0.84. This result is in good agreement with the results of Sub et al.
[12,13]. The relationship of aspect ratio and a/t can be expressed by
equation (I).
a/c = 1.2445-1.8451(a/t)+3.0591(a/t) 2 -l.6773(a/t) 3 (i)
Growth behavior of surface fatigue crack
Figure 8 represents long crack growth rate data from CT specimens of

ICr-IMo-0.25V steel. The stress intensity factor range, AK, was
calculated using the equation of the ASTM E 647-91 [14]. The data tested
at room temperature in ambient atmosphere are similar to those at 538~
in ambient atmosphere and also in argon gaseous environment. This
implies that the crack growth rate of ICr-IMo-0.25V steel is not
dependent on test temperature and test environment. The crack growth
rate can be expressed by equation (2).
dc / dN = 1.4791 x 10 -9 ( AK )3.40 (2)
Takeuchi's data [15] from CT specimens of ICr-IMo-0.25V steel at

elevated temperature are also p r e s e n t e d in Fig. 9. The data tested at
538~ are similar to Takeuchi's data at 400~ in ambient atmosphere.
, ~ [ !
I ' I ' I '
"" " Cr-Mo-V steel

E 6 R=0.05, 10Hz
E
"-" - R.To 538~
O
Cq 5 O 637 MPa 9 480 MPa
~ 544MPa 9 426 MPa
" 491 MPa 9 373 MPa
t..- 4 A 441 MPa
~)
L_ -
o 2
0.0 0.2 0.4 0.6 0.8 1.0

Cycle ratio, N/Nf
(a)
7 ' I ' 1 '
E 6
E
o
c'q
1:3
5
4
304 stainless steel (Defect)
R=0,05, 10Hz
314MPa O
298MPa E]
283MPa z~ J
R.T. 538~
9 1
_e
o 3
t..
O
G)
2
o
0~
1:: 1
O0
0 ......i .... I , I
0.0 0.2 0.4 0.6 0.8 1.0
Cycle ratio, N/Nf

(b)
FIG. 6--The relationship between surface crack length, 2c and cycle

ratio, N/Nf: (a) for ICr-IMo-0.25V steel, (b) for type 304
stainless steel.
1.4 I I I I
D Cr-Mo-V steel
R=0.05, 10Hz
o 1.2
d
1.0
o
EL
<
0.8
0.6 I I I I
0.0 0.2 0.4 0.6 0.8 1.0
Crack depth/Thickness, a/t
FIG. 7--The variation of aspect ratio.
10 -2 i , i i , ' ' ' 1 , i i i , ,,=
Cr-Mo-V steel j
R=0.05, 10 Hz
10-3 "O R.T.,Air ~ "
[] 538~ Air
o>,, 538~ Air. ~ .
o 10 .4 "Takeuchi's data ,.~t" -
E
E /k 400~ R=0.1 ,-~Z~
v
V 550~ R=0.1
Z 105
"10
"10
10-6
10-7
10 00
AK (MPa m 1/2)
FIG. 8--Dependence of long crack growth rate of CT specimen upon stress
intensity factor range at room temperature and 538~
However Takeuchi's data at 550~ in ambient atmosphere appear higher

than those of this study. This result m a y be caused by stress ratio and
o x i d a t i o n differences.
The surface crack growth rate data from the smooth specimen of ICr-
IMo-0.25V steel were expressed in terms of stress intensity factor
range. The equation for surface cracks s u g g e s t e d by N e w m a n - R a j u [16] was
used to calculate the stress i n t e n s i t y factor range. Figure I0 presents
the r e l a t i o n s h i p between the surface crack growth rate (dc/dN) and the
stress intensity factor range (AK). In this figure, the solid line
represents the results using Eq(2) for the CT specimen. As shown in
Fig. 9, the small surface crack growth rate data at room temperature are
very s c a t t e r e d b e l o w the stress i n t e n s i t y factor range of 15 M P a m I/2, and
are located above the line of Eq(2). However the crack growth rates for
the smooth specimen at stress i n t e n s i t y factors larger than 15 M P a m In
approaches those of the CT specimen. A similar result was reported by
Kitagawa et al. [2].
The growth rates of surface cracks in ICr-IMo-0.25V steel at elevated
t e m p e r a t u r e in argon gaseous environment are also p r e s e n t e d in Fig. 9.
The growth rates of surface crack at e l e v a t e d temperature below the
stress i n t e n s i t y factor range of 25 M P a m In are relatively faster than
those at room temperature, but the growth rates of surface crack are
r e t a r d e d at stress intensity factors larger than 25 MPam I/2. A c c o r d i n g to
these data, one can conclude that the growth rates of small surface
cracks are faster than those of long cracks for an equivalent stress
i n t e n s i t y factor range.
Figure i0 presents the r e l a t i o n s h i p b e t w e e n the surface crack growth
rate and the stress intensity factor range for 304 stainless steel. The
scatter of crack growth rates below the stress intensity factor range of
15 M P a m I/2 is r e l a t i v e l y larger than those above 15 M P a m In. However, the
crack growth rate data above 15 M P a m In lie w i t h i n a n a r r o w band
regardless of stress levels. A n d the crack growth rate of type 304
stainless steel at 538~ is faster than that at room temperature for an
equivalent stress intensity factor range. James' data [17] for CT
specimens at 538~ is also p r e s e n t e d in Fig. 10. James' data was
p r o d u c e d by fatigue tests p e r f o r m e d at a frequency of 40 cpm, which is
lower than the frequency used in this study. The d i s c r e p a n c y of both
data m a y be a t t r i b u t e d to specimen g e o m e t r y differences and test
conditions.
CONCLUSION
The growth b e h a v i o r of a small crack i n i t i a t e d and grown on a smooth

specimen of ICr-IMo-0.25V steel and type 304 stainless steel was studied
using the remote m e a s u r e m e n t system with image p r o c e s s i n g technique. The
o b t a i n e d results were as follows:
I. The i n i t i a t i o n and growth of a small surface fatigue crack could be
o b s e r v e d directly by using the RMS. This m e t h o d could reduce the
labor and the testing time in c o m p a r i s o n to conventional testing
methods.
2. The crack size could be m e a s u r e d and p e r m a n e n t l y stored in a personal
computer. It was possible for this system to measure surface fatigue
crack lengths as small as 30 ~m at room temperature.
3. When the growth rate of small cracks was p r e s e n t e d in terms of stress
i n t e n s i t y factor range, it was faster and more scattered than that of
the long crack at an equivalent stress intensity factor range.
4. The growth rate of surface cracks at elevated temperature was faster
than that at room temperature for a r e l a t i v e l y small stress i n t e n s i t y
factor range. However, the growth rate of the surface crack was
r e t a r d e d at a r e l a t i v e l y large stress intensity factor range.
10"2 i~ , , , ,,,,,,,,i,,,,I ,,~ , . . . . ' I
Cr-Mo-V steel /
.=0.05,lOHz /
L Room temperature /
IO-~I~8 N ~ a .o/,,
(D
O
O 10-4_ " " : ata
" 538OC ~ - Z
E
E I -
v - A 373 MPa i'11~/"-"
z
10"51
0"6 - l/;,~z~a
%
,%
, ~-
10-7 . . . . ,I. . . . . . . . . . . . . . .
10 100
AK (MPa m la)
FIG. 9--Dependence of small crack growth rate upon stress intensity
factor range for ICr-IMo-0,25V steel smooth specimen at room
temperature and 538~
&~
10 -2 I I I I I I I I / i i ! ! i |1
304 stainless steel (befect)

R=0.05, 10Hz
R.T. 538~
10-3 "314MPa O I f !
cD 298MPa [3 #
O 283MPa ,%
10-4 James' data 9
E
E
10 .5
z
10-6
10-7 i I i I ~ 1 1 1 I i i i i i i
10 100
AK (MPa m 1/2)
FIG. !0--Dependence of small crack growth rate upon stress intensity
factor range for type 304 stainless steel specimen with an
artificial defect at room temperature and 538~
REFERENCES
[i] Howland, C., Hicks, M. A. and Jeal, R. H., "The Relevance of Short
Crack Behavior to the Integrity of Major Rotating Aero Engine
Component," Proceedinas of 2nd Conference on Small Fatiau@ Cracks,
U.S.A., 1986, pp. 607-622.
[2] Kitagawa, H., Takahashi, S., Suh, C. M. and Miyashita, S.,
"Quantitative Analysis of Fatigue Process-Microcracks and Slip
Lines under Cyclic Strains," ASTM STP 675, 1979, pp. 420-449.
[3] Sub, C. M., Hwang, N. S. and Park, M. K., "Fatigue Crack Growth,
Coalescence Behavior and Its Simulation on Multi-Surface Cracks,"
Trans. of KSME, Vol. 18, No. 2, 1994, pp. 716 - 728.
[4] Saxena, A., Wilson, W. K., Roth, L. D. and Liaw, P. K., "The
Behavior of Small Fatigue Cracks at Notches in Corrosive
Environments," International Journal of Fracture, Vol. 28, 1985,
pp. 69-82.
[5] Lankford, J. and Davidson, D. L., "The Role of Metallurgical
Factors in Controlling the Growth of Small Fatigue Cracks,"
Proceedinas of 2nd Confer@nce on Small Fatiqu@ CraGks. U.S.A.,
1986, pp. 51-71.
[6] Swain, M. H., "Monitoring Small-Crack Growth by the Replication
Method," ASTM STP 1149, 1992, pp. 34-56.
[7] Davidson, D. L., "The Experimental Mechanics of Microcracks," ASTM
STP 1149, 1992, pp.81-91.
[8] James M. Larsen, Jay R. Jira, and Kakkaveri S. Ravichandran,
"Measurement of small Cracks by Photomicroscopy : Experiments and
Analysis Method," ASTM STP 1149, 1992, pp. 57-80.
[9] William N. Sharpe, Jr., Jay R. Jira, and James M. Larsen, "Real-
Time Measurement of Small-Crack Opening Behavior Using an
Interferometric Strain/Displacement Gage," ASTM STp 1149, 1992, pp.
92-115.
[i0] Richard P. Gangloff, Donald C. Slavik, Robert S. Piascik, and
Robert H. Van Stone, "Direct Current Electrical Potential
Measurement of the Growth of Small Cracks," ASTM STP 1149, 1992,
pp. 116-168.
[ii] "Standard Test Methods of Tension Testing of Metallic Materials,"
Annual Book of ASTM Standards, Part I0, E8, 1992, pp.130-149.
[12] Suh, C. M. and Kang, Y. G., "A Basic Study on Growth
Characteristics of the Small Surface Crack in 2.25Cr-IMo Steel," J.
of Ocean Enqineerina and Technoloqy, Vol. i-I, 1987, pp. 104-110.
[13] Suh, C. M. and Kim, G. N., "A Study on the Initiation and Growth
Behaviors of Surface Crack in a Type 304 Stainless Steel at Room
Temperature," Trans. cf K$ME, Vol. 8, No.3, 1984, pp. 195-200.
[14] "Standard Test Method for Measurement of Fatigue Crack Growth
Rates," Annual Book of ASTM Standards, Part I0, E647, 1992, pp.
674-701.
[15] Takeuchi, E., Matsuoka, S. and Nishijima, S., "Near-Threshold
Fatigue Crack Propagation Behavior and Frequency Effect for Cr-Mo-
V Steel at an Elevated Temperature," Trans. of JSME(A), Vol. 53,
No. 487, 1987, pp. 459-467.
[16] Newman, J. C. and Raju, I. S., "Stress-Intensity Factor Equations
for Cracks in Three-Dimensional Finite Bodies," ASTM STP 791, 1983,
pp. 238-265.
[17] James, L. A., "Frequency Effects in the Elevated Temperature Crack
Growth Behavior of Austenitic Stainless Steels - A Design
Approach," Journal of Pressure Vessel Technoloav. Vol. i01, 1979,
pp. 171-176.
Fahmy M. Haggag 1, J. A. Wang 2, M. A. Sokolov2, and K. L. Murty3
USE OF PORTABLE~IN SITU STRESS-STRAIN MICROPROBE SYSTEM TO

MEASURE STRESS-STRAIN BEHAVIOR AND DAMAGE IN METALLIC
MATERIALS AND STRUCTURES
REFERENCE: Haggag, F. M., Wang, J. A., Sokolov, M. A., and Murty, K. L., "Use of
Portable/In S/tu Stress-Strain Microprobe System to Measure Stress-Strain Behavior
and Damage in Metallic Materials and Structures," Nontraditional Methods of Sensing
Stress, Strain, and Damage in Materials and Structures, ASTM STP 1318, George F. Lucas
and David A. Stubbs, Eds., American Society for Testing and Materials, 1997.
ABSTRACT: A novel portable~in situ Stress-Strain Microprobe (SSM) system was used
to measure true-stress/true-plastic-strain (or behavior of several metallic materials, welds,
and their heat-affected-zones (HAZs) in various metallurgical and damage conditions. The
SSM system utilized an automated ball indentation (ABI) technique to measure elastic
modulus, yield strength, or curve, strength coefficient, strain-hardening-exponent (uniform
ductility), and to estimate fracture toughness (from the ABI-measured flow properties) in
carbon steels, stainless steels, nickel alloys, aluminum alloys, titanium alloys, zirconium
alloys, etc. Numerous ABI tests were also conducted on several nuclear pressure vessel
steels (NPVSs) in the unirradiated, neutron irradiated, and post-irradiated thermally-annealed
conditions. For all these test materials and conditions, the ABI-derived results were in good
agreement with those from conventional standard test methods. Furthermore, the
nondestructive ABI test results rigorously indicated the various levels of neutron-
embrittlement damage and the percentage of ductility recovery following thermal annealing
of the NPVS specimens. In situ/nondestructive structural applications of the SSM system and
its ABI technique have been demonstrated by testing a circumferentially welded stainless
steel pipe and a full-thickness section of a nuclear pressure vessel (using 90 o V-blocks and
magnetic mounts for temporary attachment of the SSM testing head to the pipe and the steel
section, respectively). All SSM localized tests were computer-controlled and conducted in
less than 2 minutes per ABI test; depending on the desired strain rate. Example test results
on metallic structural components and samples are presented in this paper.
KEYWORDS: nondestructive, in situ, structural integrity, ball indentation, partial

unloading, welds, heat affected zone, yield strength, flow properties, metals
~President, Advanced Technology Corp., 115 Clemson Dr., Oak Ridge, TN 37830
2Research Scientist, Oak Ridge National Laboratory, Oak Ridge, TN 37831
3professor, North Carolina State University, Raleigh, NC 27695
85
Copyright9
(PDVSA by ASTM
INTRODUCTION
A portable~in situ stress-strain microprobe (SSM) system was developed recently by

Advanced Technology Corporation (ATC) to test minimal material and determine several
mechanical properties (e.g., yield strength, flow properties, strain-hardening exponent,
strength coefficient) of metallic structures including their welds and heat-affected zones.
The SSM system utilizes an automated ball indentation (ABI) technique which is
nondestructive and provides a localized direct measurement of the stress-strain curve at
quasi-static strain rates at room temperature. The SSM system and test methods [1.2] are
based on well demonstrated and accepted physical and mathematical relationships which
goveru metal behavior under multi axial indentation loading. A summary of the automated
ball indentation (ABI) technique is given in Ref. [2] while more details are given in Refs. [3-
15]. The testing head of the SSM system can be configured for ABI field and laboratory
testing as well as for testing of miniature tensile test specimens. The ABI technique is based
on strain-controlled multiple indentations (at a single penetration location) of a polished
surface by a nonlinear-geometry indenter (tungsten carbide spherical indenters of 0.25 to
1.57-mm diameter). The indentation depth is progressively increased to a maximum user-
specified limit (typically less than 15% of the indenter diameter) with innovative
intermediate partial unloadings. The applied indentation loads and associated penetration
depths are continuously acquired during the ABI test and used to calculate the incremental
stress-strain values based on elasticity and plasticity theories and few semi-empirical
equations. The ABI test is considered nondestructive since it generates a shallow spherical
depression with no sharp edges or stress concentration sites. Furthermore, the compressive
residual stress left in the ABI test area will retard crack initiation.
The microprobe system currently utilizes an electro-mechanically-driven indenter,

high resolution penetration transducer and load cell, a personal computer (PC), a 16-bit data
acquisition/control unit, and copyrighted ABI software. Automation of the test, where a PC
and a test controller were used in innovative ways to control the test (including a real-time
graphics and digital display of load-depth test data) as well as to analyze test data (including
tabulated summary and macro-generated plots), made it accurate and highly reproducible.
Results of ABI tests (at several strain rates) on various base metals, welds, and heat-affected-
zones (HAZs), at different metallurgical conditions are presented and discussed in this paper.
Excellent agreement was obtained between ABI-derived data and those from conventional
ASTM methods. Gradients in the yield strength and flow properties and correlations to the
material microstructure in the weld and HAZ areas are discussed in Ref. [_8]. A 347 stainless
steel (SS) flat specimen was also tested and the ABI results compared to its material
certification. In situ SSM configuration (Fig. 1) was used successfully to test a 100-mm outer
diameter 347 SS pipe containing a circumferential weld. New SSM applications on welds,
HAZs, neutron-irradiated, and thermally-annealed steels are presented in this paper.
PORTABLE~IN SITU STRESS-STRAIN MICROPROBE SYSTEM AND ABI

TESTING
Figure 1 shows the different components of microprobe system used in this work.
HAGGAG ET AL. ON USE OF A PORTABLE/IN SITU SSM 87
These include: 1) a compact testing head, shown here in its in situ field configuration 2) a
small electronics cabinet which contains the data acquisition/control unit, other boards for
signal conditioning and control, and the drive of the servo motor, and 3) a portable personal
computer. Other testing head mounts such as magnetic holders were successfully used.
FIG. 1--A portable Stress-Strain Microprobe (SSM) system is used here for in situ testing
of circumferentially welded type 347 stainless steel pipe (I00 mm outer diameter and 5 mm
thick). The pipe is held in a wooden stand to demonstrate this field application of the SSM
system. Four aluminum V-blocks were used to temporarily attach the testing head (two
column load frame) of the microprobe system to the pipe, allowing a 360 ~ inspection of the
stress-strain curve gradients in the weld and its HAZ. The inset photo is an enlargement of
the 100 mm pipe, chuck (20 mm diameter) holding a 1.57 mm spherical tip indenter, and a
displacement transducer mounted in a horizontal bracket. (The bench-top system, which is
not shown here, has a support platen for laboratory specimen testing.)
ABI DATA ANALYSIS
Yield Strength Determination from an ABI Test
In a tensile test the uniaxial deformation is usually contained in the constant volume
of the specimen's gage section. Hence, after the completion of elastic/linear loading of a
metallic specimen, plastic-yielding and work-hardening commence and continue until
necking occurs. In contrast, in an ABI test the elastic and plastic deformation are not
distinctively separated. With increasing indentation penetration depth, an increasing volume
of test material is forced to flow under multi axial compression caused by the indenter. Thus,
in an ABI test both yielding and work-hardening occur simultaneously during the entire ABI
test without a definite single-yield-point (because there is no constant deformation volume
in an ABI test). Consequently, an accurate determination of yield strength should be based
on the entire load-displacement curve of the ABI test as explained later. It should be
emphasized here that an ABI test consisting of, for example, 7 loading/unloading indentation
cycles (see Fig. 2-a), there will be 7 consecutive yielding processes (of new material each
time the indenter has penetrated deeper into the test material) as well as 7 consecutive
processes of work hardening (of both old and new material). Hence, the yield strength
analysis in ABI testing should account for this simultaneous occurrence of yielding and
strain-hardening of test material under ABI multi axial compression. Details of yield
strength determination from ABI tests are given elsewhere [1,2,3].
The reason for the approximately linear relation of indentation load versus depth (Fig.
2-a) is because of the dual nonlinear processes occurring in opposite directions (i.e. the
nonlinear increase of load versus depth due to the spherical geometry of the ball indenter is
being offset by the power-law work-hardening behavior of the metallic test material). Hence,
ABI tests do not exhibit the traditional segmented behavior (linear elastic followed by
nonlinear/work-hardening) of the tensile load-displacement data.
For each ABI loading cycle the total penetration depth (ht) is measured while the load
is being applied, then converted to a total indentation chordal diameter (dO using the
following equation:
dt = 2 (htD- ht2)~ (1)
where D is the indenter diameter. Data points from all loading cycles up to dt/D = 1.0 are
fit by linear regression analysis to the following relationship:
P/d~2 = A (d/D) m2 (2)
where P is the applied indentation load, m is Meyer's coefficient, and A is a test material (or
specimen) parameter obtained from the regression analysis of test data of dt/D versus p/~2.
The test material parameter (A) is then used to calculate the yield strength (Oy)of the material
using the following equation:
Oy= BmA (3)
where B m is a material-type constant (e.g., a single value o f l ] m = 0.2285 [2,3] is applicable

to all carbon steels whether cold rolled, hot rolled, or irradiated). The value o f B m for each
class or type of material is determined from regression analysis of various tensile yield-
strength values (measured from specimens with different heat treatments and flow properties
and machined from different orientations) and their corresponding "A" values as measured
from entire ABI curves (up to d/D = 1.0). In equation 3 above, the units of A and Oyshould
be the same.
Flow Pronerties
The homogeneous plastic flow portion of the true-stress (o~/true-plastic-strain (ep)

Indentation Load-Depth Plot

FIg.(C:~TCIM~kTAMrlELD-~+O01 - ATC_DATA)
2.2
z /
2.1
2
1+9
J
lt+ll
1+7
L6
l.5
in
1,4 in milp~ i I
i.3
ImlliR~:ll I
1,2
uunlP",dlllil
l.l IP~Bti
i n n
a
1
O.9
O.8
o.7
mEG
BFJ
r,4m /
O.If
O~+
0.4 j// i /
O,3
0+2
O,I /7'~ !
I
/
O
O O.04 0.08 0.12 0.16 0.2 0,24
tndem, tat, lon Depth (ram)
Ca)
True-Stress/True-Plastic-Strain
FiK.(C:~ATCI~DATA%FIELD-~,O01 -- A T C _ T S S _ )
800 ,
I ~ ~'~ ~ "~'-r'~ G~'~~'

600
500 f
i I I
,0o
~ 300
200
lO0
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.15 0.2
True Plslt| 9 StrLla
(b)
FIG. 2--In Situ ABI test results o f the heat-affected-zone (HAZ) in a circumferentialty
welded 347 stainless steel pipe (shown in Fig. 1) using a 1.57-mm-diameter tungsten carbide
indenter: (a) indentation load-depth plot, (b) true-stress/true-plastic-strain data and curve
fitting.
90 SENSINGSTRESS, STRAIN, AND DAMAGE tN MATERIALS
curve can be represented by the familiar power law equation:
ot = K I~,pn (4)
where
n = strain-hardening exponent and K = strength coefficient
It should be noted that this representation is not a necessary requirement for determining the
indentation-derived oc% data as will be shown later (equations 5 and 6) but it can be used
to determine the strain-hardening exponent over the % range of interest. Furthermore, a
single power curve may not fit the entire ot-ep curve as noted in ASTM Test Method for
Tensile Strain-Hardening Exponent (n-Values) of Metallic Sheet Materials (E 646-78).
A computer program is used to solve the following equations and to thereby

determine the flow curve from the ABI data.
% = 0.2 dr (5)
ot = 4P/~dp~6 (6)
where
dp = {0.5 CD [l~2+(d~2)2]/[h~+(dv/2)2-hvD]} 1/3 (7)
C = 5.47P(1/E l + 1/E2) (8)

1.12 qb _< 1
6= t 1.12+zing l<qb_<27 (9)
t 6m,x qb>27
qb = %E2/0.43ot (10)
5ma• = 2-87Ctm (11)
z = (Sm~,,,-1.12)/ln(27) (12)
In the above equations, o t is the true stress, ~ is the true-plastic-strain,pd is the

plastic indentation diameter, D is the diameter of the ball indenter, P is the applied
indentation load, lap is the plastic indentation depth, E 1 is the elastic modulus of the indenter,
E2 is the elastic modulus of the test material, 6 is a parameter whose value depends on the
stage of development of the plastic zone beneath the indenter, ~m is a parameter proportional
to the strain rate sensitivity of the test material or specimen (e.g., for low strain-rate-sensitive
materials ~m = 1.0), and "In" is the natural logarithm.
The engineering ultimate strength (UTS) can be calculated from the ABI test results
(if the material stress-strain curve follows a power law) as follows:
UTS = K . (n/e) n (13)
ltmlr,R.Balxlar
The Brinell hardness number (HB) can also be determined from the ABI test using
the maximum indentation load ( P ~ in Kgf) and the final impression diameter (d fin mm) and
the indenter diameter (D in mm) using the following equation (Standard Test Method for
Brinell Hardness of Metallic Materials, ASTM E 10-84):
HB = 2P~/[~D(D - (D2- dr:)~ (:4)
Fracture TouRhness
A simple technique is described [32.,2~] for estimating the fracture toughness by

coupling the ABI-measured flow properties with a modified but empirically calibrated
critical fracture strain model. This technique is currently limited to ductile fracture
applications. The critical fracture strain model for ductile fracture prediction can be
expressed in the following form [16,17]
K~c = Constant (el'. ~o'. E. Oy)~ (15)
where Kj~c is the initiation fracture toughness calculated from Jic, ~* is the critical fracture
strain, 6" is the characteristic distance ahead of the crack tip over which the strain must
exceed e[, E is the elastic modulus, and % is the yield strength. The modification [18] of
this model involved: (1) the use of measured uniform strain from tensile tests or the strain-
hardening exponent from ABI tests instead of the critical fracture strain value required in the
original model, and (2) the assumption of an empirically calibrated value for the
characteristic distance, ~o*,for each class of material. The determination of the critical
fracture strain, c f, requires testing of several circumferentially notched round tensile
specimens each having a different value of its notch root radius. Since this approach is
costly, values of the critical fracture strain, or', were not determined for the materials used
in this work and "n" values were used instead.
The modified critical strain model can now be written as:
KjI c = Constant (n. C*. E. oy)~ (16)
The characteristic distance, ~o*,for ductile fracture is usually a multiple of the inter-
particle spacing and currently it should be regarded as essentially an empirically obtained
quantity. Although this dimension is presumably of relevance to the microstructural aspects
of fracture initiation it is plausibly related to the yield strength, strain-hardening exponent
(a measure of work-hardening), and the strength coefficient of the test material. However,
more research is needed to better quantify and define this dimension in order to be able to
use this method of estimating fracture toughness for applications where this characteristic
distance is expected to change (e.g. due to radiation embrittlement).
A new empirical model, used successfully [2] for estimating the fracture toughness
in A508 class 4 forging, is described by the following equation:
Kjic= Constant (K. d. n. Oy)~ (17)
where d is the grain size of the test material, K is the strength coefficient, n is the strain
hardening exponent, and Oy is the yield strength. The remaining constant needs to be
determined and verified by correlations with more extensive data. This model eliminates the
need for the characteristic distance and uses three parameters measured from ABI tests
(namely, Oy, K, and n). Furthermore, the grain size could be determined, nondestructively,
in the field for structural components using portable metallography equipment. However,
more fracture toughness and ABI measurements are needed to quantify the applicability of
this model to many heats of materials.
In Situ ABI Testing of Structural Comt)onents
A flat 347 stainless steel (SS) specimen obtained from an aerospace alloy (Heat No.
F846) was tested prior to testing the 347 SS pipe to establish a comparison between ABI and
tensile test results. The ABI-measured yield strength of 315.8 MPa (from one 7-cycles test)
was in good agreement with the tensile yield strength of 317.2 MPa (indicated on this
material's test report). A total of five ABI tests were then performed on the 100 mm outer
diameter 347 SS pipe (5 mm thick) containing a circumferential weld (308 SS). The testing
head of the portable~in situ stress-strain microprobe system was clamped on the pipe using
four 90 ~ V-blocks as shown in Fig. 1. This mounting method allowed the testing head to be
rotated 360 ~ and clamped rigidly for ABI testing at any location of the weld, HAZ, or the
base metal. A value o f l ' 3 m = 0.191 (Ref. [2]) was used for all ABI tests on stainless steel
samples and pipe materials. An example of the in situ ABI test results from the welded 347
SS pipe is shown in Fig. 2 and the results are summarized in Table 1 below. These ABI test
results show that the flow properties measured by the microprobe at three circumferential
weld areas are in good agreement with each other and are consistently slightly lower than
those at the base metal and the HAZ test locations. The above in situ tests also successfully
demonstrate the potential applicability of the microprobe system to nondestructively test
welded pipes and pressure vessels in the petroleum, fossil and nuclear power plants, etc,
ABI Testing of Welds and Irradiated Materials
The bench-top configuration of the stress-strain microprobe was used in testing

laboratory specimens of a double-V weld from a high strength steel. The ABI test results at
three test locations - base metal (BM), heat-affected-zone (HAZ), and weld metal (Weld)-
are shown in Fig. 3. This figure shows that the flow properties (true-stress/true-plastic-strain
curve) of the HAZ is not necessarily bracketed by those from the BM and Weld on both sides
of the HAZ material. It might be higher (Fig. 3) or lower than both of them (as seen in other
FIG. 3--Comparison of ABI test results on base metal (BM), heat-affected-zone (HAZ),
and weld metal (Weld)in a double-V weld of a high carbon steel.
TABLE 1-Summary o f l n Situ ABI test results from welded 347 SS pipe.
ABI-True Stress/ ABI-Yield Ultimate

Test True-Plastic Strain Strength Strength Brinell
Area (Equation) (MPa) (MPa) Hardness
Weld Metal (308 SS):

Test No. 1 (It (MPa) = 990 %.198 283 589 169
Test No. 2 o t (MPa) = 920 ep.190 283 555 164
Test No. 3 ot (MPa) = 971%190 300 586 172
HAZ:
Test No. 4 ot (MPa) = 1060 %.191 331 638 186
Base Metal (347 SS):
Test No. 5 ot (MPa) = 1097 Cp"197 325 654 188
ABI tests not in this paper). The nondestructive aspect of the ABI technique allows testing
welded joints without the need to destructively machine miniature specimens which might
also release residual stresses (generated from the welding procedure). Furthermore, the
localized ABI testing allows testing very narrow and/or irregular geometry HAZ areas. The
SSM system can be used to map out gradients in the stress-strain behavior of welded
structural components nondestructively in situ. Hence, structural integrity and/or proper
welding procedures and post-weld heat treatments can be evaluated.
The ABI technique was used to assess the degree of neutron-embrittlement damage
and the percentage ductility recovery due to post-irradiation thermal annealing. To establish
a reference condition, ABI tests were conducted on the end tabs of broken, flat, miniature,
tensile specimens of A533B-1 pressure vessel steel. Both tensile and ABI tests were
conducted using the same SSM system. Excellent agreement between stress-strain curves
from both ABI and tensile tests is shown in Fig. 4. The ABI-measured stress-strain curves
(Fig 5) show that the 343 ~ thermal annealing resulted in a partial recovery while the
thermal annealing at 454 ~ resulted in full recovery of mechanical properties of the
irradiated A533B [Heavy Section Steel Technology (HSST) Plate 02]. This was consistent
with the results from destructive fracture toughness and Charpy impact tests [21]. The
results shown in Fig. 5 demonstrate the capability of the SSM system to quantify the degree
of embrittlement and recovery due to mitigation procedures of nuclear reactor pressure vessel
steels. Furthermore, the SSM system can also monitor, nondestructively in situ, the re-
embrittlement rate of nuclear pressure vessels following their thermal annealing during their
life-extension.
CONCLUSIONS
(1) The ABI technique of the SSM system was successful in accurately determining the
yield strength and measuring the flow properties of welds in several metallic materials and
FIG. 4--Comparison between tnie-stress/true-plastic-strain curves obtained from ABI and

tensile tests on A533 B- 1 pressure vessel steel.
1200
I I I I I I I
I HSST PLATE 02 I
I100 160
I000
~s
140
o
OL
900
U)
tJJ 120 t~
~
n," BOO re-
I-- I--
01 - / fill
700
rr"
/ s~" 100 ~
I--
/,/ I...-
600 UNIRRADIATED
-- IRR + ANNEALED a t 343~ BO

500 ZRR § ANNEALED at; 454~
400 I I I I I I I 60
0 2 4 6 8 I0 t2 14 16
TRUE PLASTIC STRAIN ~,}
FIG. 5--Stress-strain curves from ABI tests on A533-B1 (HSST Plate 02) in four
conditions, i.e. unirradiated, neutron-irradiated (embrittled), and two post-irradiated
thermally-annealed conditions (to mitigate its radiation damage).
96 SENSINGSTRESS, STRAIN,AND DAMAGE IN MATERIALS
structural components. The gradients in mechanical properties of weld metals and their
HAZs were successfully determined from ABI tests conducted on both laboratory specimens
as well as on structural components.
(2) The SSM system quantified the degree of neutron-embrittlement damage as well as
the percentage recovery in mechanical properties resulting from post-irradiation thermal
annealing of nuclear pressure vessel steels.
(3) The in situ nondestructive capability of the SSM system can be used to monitor the
aging of bridges and other critical components and to verify their structural integrity.
Furthermore, welding and repair procedures can be accurately assessed.
REFERENCES
[_1] Haggag, F. M., "Field Indentation Microprobe for Structural Integrity Evaluation,"
U.S. Patent No. 4,852,397, August 1, 1989.
[2_] Haggag, F. M., "In-Situ Measurements of Mechanical Properties Using Novel

Automated Ball Indentation System," Small Soecimen Test Techniaues Applied to
Nuclear Reactor Thermal Annealine 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. 27-44.
[3_] Haggag, F. M., Nanstad, R. K., Hutton, J. T., Thomas. D. L., and Swain, R. L., "Use
of Automated Ball Indentation Testing to Measure Flow Properties and Estimate
Fracture Toughness in Metallic Materials," Applications of Automation Technology
to Fatigue and Fracture Testing. ASTM STP 1092, A. A. Braun, N. E. Ashbaugh, and
F. M. Smith, Eds., American Society for Testing and Materials, Philadelphia, 1990,
pp. 188-208.
[4] Haggag, F. M., Nanstad, R. K., and Braski, D. N., "Structural Integrity Evaluation
Based on an Innovative Field Indentation Microprobe," Innovative Approaches to
Irradiation Damage and Failure Analysis, D. L. Marriott, T. R. Mager, and W. H.
Bamford, Eds., PVP Vol. 170, American Society of Mechanical Engineers, New
York, 1989, pp. 101-107.
[_5] Haggag, F. M. and Nanstad, R. K., "Estimating Fracture Toughness Using Tension
or Ball Indentation Tests and a Modified Critical Strain Model," Innovative
Approaches to Irradiation Damage and Failure Analysis, D. L. Marriott, T. R. Mager,
and W. H. Bamford, Eds., PVP Vol. 170, American Society of Mechanical
Engineers, New York, 1989, pp. 41-46.
[6] Haggag, F. M., Wong, H., Alexander, D. J., and Nanstad, R. K., "The Use of Field
Indentation Microprobe in Measuring Mechanical Properties of Welds," Recent
Trends in Weldin~ Science and Technolo~. TWR'89. Proceedings of the 2nd

International Conference on Trends in Welding Research, Gatlinburg, Tennessee,
U.S.A., 14-18 May, 1989, 1990, pp. 843-849.
[7_] Haggag, F. M., "Application of Flow Properties Microprobe to Evaluate Gradients

in Weldment Properties", ~tcrnational Trends in Welding Sciences and Technology,
S. A. David and J. M. Vitek, Eds., ASM, Metals Park, OH, 1993, pp. 629-635.
[8] Haggag, F. M. and Bell, G. E. C., "Measurement of Yield Strength and Flow
Properties in Spot Welds and Their HAZs at Various Strain Rates,"International
Trend~.jn Welding Science and Technology, S. A. David and J. M. Vitek, Eds.,
ASM, Metals Park, OH, 1993, pp. 637-642.
[9] Haggag, F. M., "The Role of Ltiders Strain in Predicting Flow Properties in Steels
from an Instrumented Hardness Test," M.S. Thesis, Department of Chemical and
Nuclear Engineering, University of California, Santa Barbara, California, 1980.
[1ol Haggag, F. M. and Lucas, G. E., "Determination of L~iders Strains and Flow
Properties in Steels from Hardness/Microhardness Tests," Metallurgical Transac~igns
A. Vol. 14A, 1983, pp.1607-1613.
JR] Au, P., Lucas, G. E., Sheckherd, J. W., and Odette, G. R., "Flow Property
Measurements from Instrumented Hardness Tests," Non-destructive Evaluation in
the Nuclear Industry. Metals Park, OH., 1980, pp. 597-610.
[12] Tabor, D., "The Hardness of Metals," Clarendon Press, Oxford, 1951.
[1--3] Francis, H. A., "Phenomenological Analysis of Plastic Spherical Indentations,"
Tra~.sactions of the ASME~July 1976, pp. 272-281.
[14] George, R. A., Dinda, S., and Kasper, A. S., "Estimating Yield Strength from
Hardness Data", Metal Progress. May 1976, pp. 30-35.
[1--.5] Server, W. L. and Oldfield, W., "Nuclear Pressure Vessel Steel Data Base," EPRI
Report NP-933, 1978.
[]6] Pandey, R. K., and Banerjee, S., "Strain Induced Fracture in Low Strength Steels,"
Engineering Fracture Mechanics, Vol. 10, 1978, pp. 817-29.
[17] Ritchie, R. O., Server, W. L., and Waullaert, R. A., "Critical Fracture Stress and
Fracture Strain Models for Prediction of Lower and Upper Shelf Toughness in
Nuclear Pressure Vessel Steels," Metallurgical Transactions A, Vol. 10A~(1979) pp.
1557-70
[ 18] Haggag, F. M., Reuter, W. G., and Server, W. L., "Recovery of Fracture Toughness
of Irradiated Type 347 Stainless Steel Due to Thermal Stress Relief: Metallographic
and Fractographic Studies," Proceedings of the 2rid International Symposium on
Environmental De~adation of Materials in Nuclear Power Systems-WaterReactors,
Monterey, Califomia, September 9-12, 1985, 1986, pp. 509-14.
[19] Pavinich, W. A., "The Effect of Neutron Fluence and Temperature on the Fracture
Toughness and Tensile Properties for a Linde 80 SubmergedArc Weld," Proceedings
of the 2nd Intemational Svmoosium on Environmental Degradation of Materials in
Nuclear Power Systems-Water Reactors. Monterey, California, September 9-12,
1985, 1986, pp. 485-95.
[2_0] Haggag, F. M., Server, W. L., Lucas, G. E., Odette, G. R., and Sheckherd, J. W.,
"The Use of Miniaturized Tests to Predict Flow Properties and Estimate Fracture
Toughness in Deformed Steel Plates," Journal of Testing and Evaluation. JTEVA,
Vol. 1, No. 1, Jan. 1990, pp. 62-69.
[21] Sokolov, Mikhail A., Donald E. McCabe, ShafikK. Iskander, and Randy K. Nanstad,
"Comparison of Fracture Toughness and Charpy Impact Properties Recovery by
Thermal Annealing of Irradiated Reactor Pressure Vessel Steels," _Seventh
International Symposium on Environmental Degradation of Materials in Nuclear
Power Systems - Water Reactors, Breckenridge, Colorado, August 7-10, 1995, 1995,
pp. 771-782.
Optical, Non-Contacting Strain
Measurement Devices
Peter C. McKeighan, Steven B. Seida, Ernest A. Franke, and David L. Davidson ~
BISMAP: A NON-CONTACT BIAXIAL DISPLACEMENT MEASUREMENT

SYSTEM
REFERENCE: McKeighan, P. C., Seida, S. B., Franke, E. A., and Davidson,

D. L., "BISMAP: A Non-Contact Biaxial Displacement Measurement System,"
Nontraditional Methods of Sensing Stress, Strain, and Damage in Materials and
Structures, ASTM STP 1318, George F. Lucas and David A. Stubbs, Eds.,
American Society for Testing and Materials, Philadelphia, 1997.
ABSTRACT: The development and validation of a prototype BISMAP ~iaxial

Dis_placement Measurement by M_~hine Vision _Processing) system is described. The
non-contact biaxial displacement measurement system adapts machine vision
technology to create a unique extensometer. The BISMAP system is built around a
vision processing system that tracks and recognizes surface texture features around
specific control points by using a normalized correlation technique. The system has
numerous advantageous features including non-contact measurement in two-
dimensions, variable gage length, direct surface measurement (no attached target) and
real-time measurement of multiple, discrete points. The performance of the system is
demonstrated with evaluations of the monotonic (and in some cases cyclic) stress-
strain response for four different materials: aluminum, LEXAN, rubber and baboon
tendon. The system has a measured accuracy of 370 lae with • lae variability when
compared to a strain gage. It is best-suited for measuring strains in excess of 5000 lae
and there is no practical upper limit on the measurable strain with retraining. The
BISMAP system shows promise for development from prototype into a laboratory
instrument capable of measuring strains in pliable or delicate materials.
KEYWORDS: extensometer, biaxial, non-contact, stress, strain, displacement, vision

system, normalized correlation, aluminum, LEXAN, tendon, rubber.
INTRODUCTION
Understanding the deformation behavior of new or exotic materials often

requires full-field measurement not suited to the capabilities of conventional
tSr. Research Engineer, Sr. Research Engineer, Institute Scientist and Institute Scientist,
respectively, Southwest Research Institute, 6220 Culebra Rd., San Antonio, TX 78238.
lOl
Copyright9
(PDVSA by ASTM
extensometers. Moreover, performing the mechanical testing required for these

materials can be expensive once material cost and increased testing complexity (often
including elevated temperature) is considered. Although some optical full-field
measurement techniques are available [1] (e.g., brittle- or birefringent-coating, moir6
and photoelastic methods), these approaches lack spatial resolution, can be time
consuming to implement and interpret and are not well-suited to most aggressive
environments.
Several non-contact laser extensometers have been commercially developed

[2,3] although these transducers require that targets be affixed to the specimen. With
other systems [4,5] it is difficult to relate microstructural differences within the
specimen to the measured deformation. This is a problem for non-homogeneous
materials exhibiting composite-type structures. One further disadvantage with many
available systems is that they measure displacement along only one axis. This
precludes the observation of heterogeneous deformation that might occur either near a
notch or due to material inhomogeneities. A system similar to the BISMAP system
detailed herein has been developed over a period of years beginning in the mid-1980's
by Sutton et al. as described by Lyons et al. in reference [7]. This system and the
BISMAP system differ with regard to image correlation algorithms used, texture
modification to the specimen and illumination source as well as overall camera
resolution. Nevertheless, the system described [7] clearly illustrates the suitability for
the overall optical technique for high temperature environments which is also
envisioned as a potential application for the BISMAP system.
The ideal new generation of extensometer would be easy to use, require no

unusual specimen preparation, record in three-dimensions, not need targets requiring
attachment, have high accuracy and a rapid sample rate, permit the choice of gage
length and be reasonably priced. The BISMAP system described in this paper satisfies
some of the above listed characteristics. Strictly speaking, the term BISMAP in this
paper refers to the complete system including software and hardware. Most notably
the system measures in two-dimensions, does not contact the specimen, uses a gage
length of choice, requires no targets and performs in real-time. BISMAP uses image
processing to measure displacements recorded from nine points on the surface of a
specimen. The system developed so far is considered a prototype designed to
demonstrate the feasibility of the technical approach.
In this paper, statistically relevant comparative measurements are provided to

demonstrate the accuracy, resolution and repeatability of the system. A strain-gaged
specimen made from LEXAN polycarbonate was used for this comparative assessment.
In addition, the system is demonstrated using three very different materials (metal,
rubber and tendon) that span a four order of magnitude stiffness range. The
advantages, as well as limitations, of the BISMAP system can be clearly observed
from this wide ranging study.
McKEIGHAN ET AL. ON BISMAP 103
METHODS AND MATERIALS
System Overview and Ot3eration
The displacement measurement system is based on the use of normalized image

correlation techniques to track the position of features on the surface of a specimen.
The basic components of the prototype BISMAP system shown in Figure 1 include a
telemicroscope, digital camera, machine vision system and microcomputer. The image
of the specimen passes through the telemicroscope and is captured by a monochrome
charge coupled device (CCD) camera with a 512x480 resolution. This image is sent
to a Cognex machine vision system for processing.
FIG. 1--Major components of the BISMAP system.
The Cognex image processing computer provides hardware and proprietary

software for training models to detect features around tracking points in addition to
general purpose image processing tools. During prototype development, two different
Cognex machine vision systems were used. The Cognex 2000 and 4400 computers
use Motorola 68000 and 68040 microprocessors, respectively. The Cognex 4400 also
has a newer version of the system software. The displacements of the tracking points
are transmitted serially (via RS-232) from the Cognex computer to the data
presentation system for calculation of stress and strain.
When initializing the system, an operator first establishes the area of the
specimen to be viewed. This is accomplished by selecting the appropriate
telemieroscope and magnification. The desired number of tracking points, or grid
points, are then chosen and located. A typical electronic grid is illustrated and defined
in Figure 2. The best resolution occurs when this grid is positioned to use most (75-
90%) of the CCD camera image. One of the grid points is then selected as an overall
reference and the machine vision system stores the characteristics of the features
surrounding each point. The process for developing a spatial pattern of features
around the control points is termed "training." As load is applied to the specimen, the
Cognex computer searches the area surrounding each control point and uses
normalized correlation algorithms to determine the new coordinates for each grid
point. These coordinates can only be determined if the machine vision system
recognizes the features identified during training. Given the new coordinates, the data
presentation computer can calculate biaxial (horizontal and vertical) displacements.
Consequently, the strain in each direction, or alternatively Poisson's ratio, can be
calculated, plotted or stored on disk for further analysis.
FIG. 2--IUustration of computer-generated grid points (a) overlayed on rubber

specimen and (b) with general terms defined.
User Interface
Although the BISMAP system presented is only in a prototype form, the user
interface is a well developed, menu-driven system allowing significant flexibility.
This user interface was written using the LabWindows software from National
/nstmments (Austin, TX). Graphical displays for viewing recorded data are available
as is the abilityto view program status from a fde logging major events. Displacement
and straindata are both displayed on-screen in real-time and written to disk.
An option is available to view a displacement map of the grid points as a test

progresses. The original location of the grid points are indicated with vectors
originating from each illustrating the measured displacements. This graphical
representation provides insight into specimen behavior that may be especially useful
when testing non-homogeneous materials.
Measurement Methodology
The purpose of the training process alluded to earlier is to develop a model that
can be subsequently used to track a given grid point. In this context a model is a
pattern of grey levels surrounding the grid point that spatially represents the features.
One parameter initially experimented with is the number of grey levels used when
capturing the features of an image area. Experience has shown that sixteen levels
provide the optimum balance between speed and effectiveness.
The machine vision system uses Cognex training routines supplemented by

some additional custom refinement algorithms to develop the model. After completing
the initial training phase, the Cognex algorithm tests the quality of the models for each
control point. Although the Cognex software trains well, two additional tests have
been developed to assess the character of the surface texture.
For the first custom check, standard deviation of the intensity of each pixel in
the model is calculated. If the standard deviation is less than 4.0, the model is
retrained by either moving the grid points to new locations or increasing the model
pixel area. Experience with using the Cognex software has indicated that this pixel
value standard deviation is a good quality test of the trained model. The second check
ensures that the models can replicate grid point position in a stationary (unloaded)
specimen. For each point, the machine vision processing system captures four images
and relocates the grid points. If the model is not found centered at the point, the
system will be retrained.
A normalized correlation searching algorithm is used to fred image areas

similar to the trained models. Normalized correlation is a measure of the geometric
similarity between an image and a model, independent of any linear differences in
image or model brightness. The advantages of this technique over some other
automated feature-identification schemes are immunity to image noise, tolerance to
contrast fluctuations independent from image complexity and repeatable subpixel
accuracy. A more in-depth discussion and mathematical foundation for the normalized
correlation technique can be found in references [7,8].
As noted earlier, the image processing system limits the region searched to an
area slightly larger than the model and centered around the model's previous position.
This allows faster, real-time tracking than possible if the entire image were marched.
When searching, the normalized image correlation generates a shape score quantifying
differences between the image and model. The Cognex system uses these shape scores
in the vicinity of the control point to interpolate with a proprietary set of equations
and microprocessors to calculate fractional pixels and estimate subpixel position.
These Cognex searching functions yield a subpixel interpolation accuracy of 0.1 to
0.25 pixels [9]. Since a given displacement measurement is the difference between
two pixel locations, the total displacement resolution is in the range ~=0.2-0.5 pixels.
This subpixel displacement resolution can be used to calculate the theoretical

resolution of the BISMAP system. The gage length L between a pair of tracking
points selected during initialization can be expressed as a percentage of the total CCD
camera field-of-view. Given 512 pixels in one direction, the gage length is defined by
the quantity 512(~/100) where 11 is the percentage (typically ~ is chosen to be 70% or
greater). This implies that the system strain resolution er is
Er-- - -
(1)
L 512 (11/100) 5.12 11
for a CCD camera with a total pixel area of 512x512. This resolution can be
maximized by (a) aligning the loading axis with the camera axis having the greatest
number of pixels and Co) selecting the tracking points to fill most of the field-of-view.
System resolution limits are given in Table 1 for a variety of subpixel

resolutions and percentage of image used, T1. The range of strain resolution is 500-
2000 Be. The actual value observed in practice will depend upon a number of factors
influencing the quality of the captured image (most notably the surface texture of the
specimen). For the tests detailed herein, ~q was generally 70% or greater which
implies the best theoretical resolution should be approximately 500 Be.
TABLE 1--Theoretical strain resolution, e,, for a 512 x 512 pixel CCD camera.
Subpixel Percentage of System Strain Resolution, e r

Resolution, ~ CCD Image Used, ~ xlO" (microstrain)
0.2 85% 4.596 ( 460 )
(best case) 70% 5.580 ( 558 )
50% 7.813 ( 781 )
0.5 85% 11.489 ( 1149 )
(worst case) 70% 13.951 ( 1395 )
50% 19.531 ( 1953 )
Materials Utilized and Testin2 Procedures
Assessing the performance of a unique measurement system is a difficult and

challenging task. In many cases the established metrics, developed to compare
traditional measurement systems, do not address the unique strengths or capabilities of
this new system. With this in mind, the performance of the BISMAP was investigated
using a variety of materials designed to illustrate different aspects of this non-
contacting system.
Four materials were used to demonstrate the performance of the system as

shown in Table 2. These materials span a wide range of behavior with the stiffness
(Young's modulus) varying by four orders of magnitude. The aluminum and
polycarbonate samples were machined into standard 150 mm long tensile bars with a
gage width 12 mm by a thickness of either 6 mm (polycarbonate) or 4 mm
(aluminum). The rubber and baboon tendon (fresh frozen) were tested in the as
received condition with the section dimensions as shown in Table 1.
TABLE 2--Details for the different materials investigated.

Handbook Approx. Size Surface
Material Classification Modulus, GPa (gage FOV) Texture
7075 aluminum 70.0 150 mm tensile spec. ion etched
(1.0 x 0.8 ram)
LEXAN transparent 2.4 150 mm tensile spec. paint spray
polyearbonate (5 x 4 mm)
bicycle fire rubber ~0.001 8ram x 20ram sect. ion sputtered
inner tube (17 x 13 r a m ) platinum coating
baboon biomaterial N/A 10ram x 30mm sect, aluminum powder
tendon (12 x 9 ram) (swabbed)
The performance of the BISMAP system requires sufficient surface texture to

provide contrasting features. Four different techniques (as shown in Table 1) were
used to provide additional texture to the specimens. These techniques have been
developed through extensive experience using the stereoimaging technique to measure
displacements [10,11]. Although having to usually add texture to the surfaces is a
disadvantage for the BISMAP system, the four methods used in this project and tested
in Table 2 demonstrate that a variety of techniques are available to achieve enhanced
surface contrast.
If a given specimen does not have sufficient optical texture, the techniques
detailed are used to add it. The resulting texture is of an optical character and does
not generally influence mechanical behavior of the sample. However, care must be
taken to insure that adding texture does not adversely impact properties. For instance,
it would be undesirable to roughen a surface that will be tested cyclically so as to
avoid influencing the inherent fatigue performance. In all cases sound experimental
judgement must be exercised to avoid influencing the mechanical behavior studied.
The stress-strain behavior of the materials was determined by subjecting each

specimen to a tensile test. Although most tests were highly nonstandard, testing
basically conformed to relevant ASTM procedures in terms of techniques and strain
rates. Testing was performed in a pseudo-static mode at a relatively low, incremental
strain rate (in the range of 0.001-0.005 mm/mm per minute) in all cases so as to
minimize the potential for transient effects. In a highly deformable material such as
rubber, even the best gripping arrangements will result in applying nonuniform stress.
Furthermore, nonuniformities in the sample itself may cause widely varying spatial
strains.
Although the rubber and baboon tendon specimens were difficultto grip using
conventional methods (see Figures 3 and 4), these deformable materials are well suited
to the field displacements measured with the B I S M A P system. A typical extensometcr
would provide a single, average measurement that will not be representative of the
character of the strain distribution. B I S M A P on the other hand, with nine grid points,
allows strainto be potentially measured for 12 different gage sections (six in each
direction) hence allowing greater flexibilityfor processing and interpretingthe
resulting data and strain distribution.
FIG. 3--Test and grip arrangement FIG. 4--Test and grip arrangement
for the rubber inner tube. for the baboon tendon.
The LEXAN polycarbonate material was used to more rigorously assess

BISMAP system performance by comparing results to a resistance strain gage. The
strain gage used had an 8.9 mm gage length and was mounted on the surface of the
specimen opposite that optically monitored as shown in Figure 5. The comparative
indicator for assessing BISMAP system performance is the foil resistance strain gage,
a transducer that by itself can manifest some degree of inaccuracy [1]. Although not
an absolute strain indicator, the strain gage should provide accurate strains on the
order of 1% using the established gage installation techniques and transducer
instrumentation employed in this work. This level of accuracy is believed sufficient to
assess overall BISMAP performance.
FIG. 5--LEXAN specimen with a barely visible (back side) strain gage.
In addition to performing a single test over the elastic range of LEXAN, a

measurement of global repeatability was performed. This was assessed by dividing the
elastic range of the tensile test into three distinct regions with full unloading (to zero
load) and BISMAP system re-initialization performed between each segment. During
this testing, the BISMAP strain measurement was repeated eight times in succession
for each measurement point. The strain data presented is an average from the three
columns of strain for these eight measurements (24 total).
Two separate evaluations were also performed on the rubber sample. First, a
cyclic stress-strain test was evaluated for two complete cycles in a low relative strain
(r regime. Measurements were subsequently made for the rubber when high
strains (r or 100% strain) were applied.
RESULTS
General Svstem Performance
The performance of the two Cognex computers (a model 2000 and newer 4400)
differed primarily with regard to processing speed. The time spent training is
generally of less importance when assessing overall system performance (training is an
infrequent event). Nevertheless, this time was generally in the range of 30 to 90
seconds. The time taken to capture and track an image has a greater impact on overall
system speed. Regardless of material type and surface texture, the model 4400 was
able to capture an image and locate the nine control points in about 0.4 second.
Conversely the model 2000 was less consistent and generally required in the range of
0.8 to 1.3 seconds. The performance of the older model 2000 system was highly
dependent on surface texture.
The processing speed is of utmost importance for real-time data generation.

One of the primary variables influencing the speed of the processing algorithm is the
size of region around the tracking point searched. Experience with the Cognex
systems has shown that features in the search area surrounding each tracking point do
not change significantly in shape as the specimen is deformed. This implies that the
image processing algorithm can constrain its search and consequently provide a faster
analysis. In practice a search region ten pixels around the tracking point is sufficient.
During typical operation the system requires retraining every 25-30% strain applied.
Given a specimen with good texture, the retraining success rate was typically not an
issue and proceeded without difficulty. Furthermore, retraining did not generally
require application of any certain load or strain level.
It is also worth reiterating that the BISMAP system developed herein is in a

prototype form. Given this essential nature, the techniques used still remain in a
prototype form. For instance, point-to-point measurement efficiency (in terms of time)
is less of an issue in this work than demonstrating and documenting system
performance. Many of the issues related to this prototype nature will be discussed in
more detail in the next major section of this paper.
LEXAN Material
The strain gage mounted on the LEXAN specimen allows a direct comparison
between measured strain levels which leads to an assessment of accuracy. The stress-
strain behavior is shown in Figure 6 with the expected modulus value shown from
reference [12]. In general the BISMAP and gage strains agree well though each
differs slightly from the tabulated modulus. This is not considered significant however
since the difference is slight, the pedigree of the LEXAN not fully known and the
stress-strain behavior generally non-linear, as expected with polymeric materials.
The data shown in Figure 6 is further assessed in Figure 7(a) by calculating the
strain difference between the gage and the BISMAP result at each incremental stress
value. This error gradually increases, reaches a peak and decreases. The observation
of an increase followed by a decrease is significant because it implies that cumulative
(a) LEXAN
0.0006
==
i
J 0.0004 average
40 ~
LEXAN S (9
00.0037
0~ theoretical
0.0002 strain
.c resolution
30
,./.,/ O3
\'y 0.0000 . . . . I . . . . I . . . . I . . . . . I
Oi 0 10 20 30 40
IX
Stress, MPa
u) 20
LEXAN
(b)
0.0002
10 9 strain variability c)
/ . ;> defined by 2 p~
-- strain g a g e standard m
average deviations
9 BISMAP -1-
0.0001 0.000046
0.
z
0 . . . . I . . . . I '' '1 . . . . O9 _. ;__._,_.~_._.~._.__" .__"._;_. ,_.,
0.000 0.005 0.010 0.015 0.020 .r--
Strain 0.0000 . . . . I . . . . . . .i . . I . . . . . i o
10 20 30 40 z
Stress, MPa f.D
FIG. 6--BISMAP measured strain response of LEXAN FIG. 7--System performance assessing (a) strain accuracy "(7
compared with strain gage output. Each BISMAP and (b) overall strain measurement variability for the
strain is the average of three columns for eight LEXAN material. ..J.
_.r
repeated measurements.
(progressive) error is not an issue. The low magnitude of error is somewhat surprising
in view of the theoretical resolution calculations presented in Table 1. In practice, any
measurement differences less than a multiple (for example 2 or 3) of the theoretical
resolution is not experimentally significant. Using this test demonstrates that all of the
data in Figure 7(a) is not significant. Nevertheless, the implication is that the
BISMAP system is out-performing the specified pixel accuracy and achieving subpixel
accuracies at least 2x less than quoted.
The variability of the 24 measurements comprising each strain value is shown

in Figure 7Co) for each incremental stress level. This variability, for this case defined
by two standard deviations, is quite low. This level represents the strain variability
range defining the 95% confidence interval (assuming a normal distribution of strain
measurements). The data shows that each strain measurement is quite consistent with
overall measurement variability approximately • lie. Although the averaging
technique does limit real-time processing speed, the importance of multiple
measurements is clearly indicated by the two outlier variabilities apparent at 15 and 25
MPa. An examination of the data for these two stress levels does not yield any
obvious rationale for why the variability was 3-4x greater than observed for all other
points.
The results from the measurement repeatability exercise are indicated in the
stress-strain response in Figure 8. A direct comparison between Figure 8 and Figure 6
suggests that the difference between the BISMAP measurement and the strain gage is
comparable or even slightly less for Figure 8. Nevertheless, the apparent discontinuity
between the first and second loading is somewhat of a concern. The magnitude of this
strain difference is approximately 600 lie which is on the order of both the theoretical
resolution and the maximum error observed in Figure 6. Since a similar discontinuity
was not observed in the strain gage measurements, the offset is BISMAP system
generated as opposed to some type of mechanical or loading anomaly.
Rubber. Aluminum and Tendon Material
The cyclic and monotonic stress strain results for rubber, shown in Figure 9(a)
and 9(b) respectively, illustrate BISMAP system performance at higher strain levels.
The hysteresis and non-linearity of the stress-strain response under cyclic loading is
typical for this type of material. In general, the responses shown exhibit good
repeatability for the cyclic loading.
System performance under very large (almost 100%) strain measurements is

apparent from Figure 9(b). As the strain increased the sample stretched and distorted
until the grid points moved to the edge of the image at strains of 0.35 and 0.70. At
this point the load was held constant, new control points chosen and the models
retrained. The retraining did not result in any apparent discontinuities in the material
response for rubber as was observed for LEXAN. The data from this high strain case
(a) _. , . . . . . ~
0.20
40 Q.
LEXAN
strain 0.10
gage
~"~/ ---e- Cycle 2
9 first loading
[] second loading
30 0.00
9 third loading
I = i = i i . . . . I
0.00 0.05 0.10 0.15

first Strain
loading "=="1 ~j,,.. third loading
20 (with retrain)
1.00
Rubber
second loading I 0.75
( w i t h retrain~ I g~
10 t~
0.50 m
-1-
0.25 z
~retrain I~n m
0 ~
, , , = . . . . . . . . . . . .
0.000 0.005 0.010 0.015 0.020 0.00 r

0.00 0.25 0.50 0.75 1.00 0
Strain z
Strain
-o
FIG. 8--Incremental stress-strain response with unloading FIG. 9--BISMAP-measurexi strain response of robber under
and retraining prior to reloading. (a) cyclic and (b) monotonic loading conditions. Each
strain is an average of 12 repeat measurements for .--L
GO
column 1.
was used to calculate an incremental (point-by-point) tensile modulus as shown in

Figure 10. The initial high modulus value tends to decrease and reach a plateau value
at stresses greater than 0.5 MPa. Little scatter is observed once the initial non-linear
region of the stress-strain curve is exceeded. Over the full range of the response the
modulus decreases by a 2.5x factor.
Rubber
~2
"0
0
E
ffl
e-
9 /.
= 1
0
>-
0 .... ~ ....
I I .... ', . . . .
0.00 0.25 0.50 0.75 1.00
Stress, MPa
FIG. 10--Incremental Young's modulus for rubber under monotonic loading.
Whereas the displacements measured with rubber are large, the strains required
to assess the response of aluminum in Figure 11 are much smaller. In fact, the strain
range of Figure 11 is 150x smaller than the range for rubber in Figure 9. Regardless
of this strain range difference the general response of the data in Figure 11 is linear as
expected for the elastic regime evaluated. However, the bars indicating strain range
are wide relative to the strain magnitude measured. In view of this observation,
accurate measurement of the elastic strain response of aluminum and similar stiffness
materials is considered beyond the capability of the current prototype BISMAP system.
However, it should be mentioned that BISMAP would be suitable for assessing overall
o-e response.
The stress-strain response for the final material considered, baboon tendon, is
shown in Figure 12 for two columns of grid points. The behavior exhibitod, with both
tension and compression across the gage section of the specimen, is indicative of
bending. In fact, this bending is apparent in the photograph of the gripped and loaded
tendon shown in Figure 4. What is particularly interesting is that even though the
Aluminum
-W
400
/
3OO 7"-
./
o..
7"
\
r
u~
200
/ bars indicate
range of re-
corded data
/:._
/
100
-./-
, , t I , = , , I , t , =
I
0.000 0.002 0.004 0.006

Strain
FIG. 11--BISMAP measured strain response of 7075 aluminum. Each strain point is
the average of 12 repeat measurements (the range of these measurements are
shown with the error bars) for column 1 strain.
0.5
Tendon
"~ col 1 col 3

0.4
reference ~,~ ~,~ _

points\
- - + T+ T+
T T
~. 0.3 ",*** _L
i t t
=o
0.2
+ Column I (loading)
0.1 0 Column 1 (unloading)
---II-- Column 3 (loading)
0 Column 3 (unloading)
0.0 9' ~ I I
-0.010 -0.005 0.000 0.005 0.010 0.015

Strain
FIG. 12--BISMAP measured strain response for baboon tendon. Each strain value
is the average of 12 repeat measurements for each column.
grips apply a tensile axial load, one side of the gage section is completely in
compression that increases as the tensile load is applied. This strain response is
presumably a consequence of non-uniformities in the sample itself. Prior to loading a
loosely attached membrane sheath with a featureless texture that encased the tendon
was removed. Moreover in compression the tendon behaved fairly elastic, whereas, in
tension it exhibited a response consistent with post-yield behavior. Although the
elastic limit of a tendon in tension is unknown, the behavior observed is consistent
with tearing of fibrous material.
The stress-strain response of the baboon tendon is included to illustrate the type
of behavior that can be measured with the BISMAP system. Although in the strictest
sense the compressive strains noted are surely accompanied by compressive bending
stresses, the strain measurements made on a micro-scale illustrate one of the key
inherent features of the system, namely the ability to provide multiple gage length
measurements. Whereas the measured load is a global, single value, the BISMAP
system can be used to assess a limited degree of fuller field strain measurement over a
narrow field of view.
This application provides an excellent demonstration of the fuller field, multi-

axis strain measurement capability of the BISMAP system and the corresponding
importance of this capability for non-homogeneous materiais. The observed
deformation behavior of this biomaterial is clearly complex and not well-suited to
results from an averaging extensometer.
Biaxial Measurements
The focus of the data presented so far has been measuring strain parallel with
the loading axis. The BISMAP system also provides strain measurement capability in
the orthogonal axis. Poisson's ratio, which is the ratio of these strains and is denoted
by v, can be calculated. The variation of Poisson's ratio with stress is shown for
rubber and tendon in Figure 13 for both the loading and unloading segments of the
stress-strain behavior. Differences in v for the loading and unloading portion of the
stress-strain behavior would be expected for the rubber and tendon exhibiting
hysteresis in the stress-strain response.
For most isotropic engineering materials Poisson's ratio typically varies between
0 and 0.5. This is the case for the rubber with measured values varying between 0.3 and
0.4. The lower values of v may be a consequence of the short gage length of the
specimen and the close proximity of the grips (the same also applies for the tendon
sample). The normal length-to-width aspect ratio for specimens, generally 3:1 or more,
was not possible with the limited material available. Consequently, grip effects appear
to be limiting the lateral strain in the sample, resulting in a lower Poisson's ratio for the
rubber.
McKEIGHAN ET A L ON BISMAP 1 17
Higher levels of v were observed for the tendon with values between 0.4 and 0.6
with the highest values observed during the unloading. As seen in Figure 12, the axial
strain nearly remained constant during unloading. The implication of a constant v for this
region is that the lateral strain is also remaining constant during the unload. Clearly, the
biaxial deformation behavior of this material is complex and assumptions for typical
engineering material behavior are not applicable.
0.7
0.6 [3
0.5
6
0.4
r i
o 0.3
0
n
9 Rubber(loading)
0.2
O Rubber(unloading)
9 Tendon (/Oi~ng)
o.1 Q Tendon(unloading)
0.o i 9 ' '' I .... I ...... : : ....

0.0 0.1 0.2 0.3 0.4 0.5
Stress, MPa
FIG. 13--Poisson's ratio for rubber and baboon tendon as derived from BISMAP-
measured axial and lateral strains. Only the tensile portion of the stress-strain
response of the tendon is shown.
DISCUSSION
The focus so far has been to demonstrate the feasibility of the BISMAP system
and to determine the strengths and limitations of the system. In this part of the paper,
the application and future development of the BISMAP system are examined and
developmental issues are addressed.
Unresolved Developmental Issues
The results presented were generated using a technique to average the data
from several repeat measurements to limit the influence of rogue displacements that
can occasionally be measured. The influence of spurious data is clearly observed in
the strain variability outliers shown in Figure 7Co). One way to improve data quality
would be to intelligently sort the over-sampled data and discard the outliers. Although
this would be straightforward to implement, the current BISMAP system does not
incorporate this feature.
The optimum strain level to retrain the system has also not been explored. In
order to achieve maximum resolution, the initial tracking points are selected so that
they span most of the field-of-view. As the specimen is loaded the control points
move out to the edge of the image. At this point two effects can contribute to greater
error in displacement measurement. First, the distortion of the magnified image is
greatest closest to the edge due to lens abberations. Second, the accuracy of the
normalized correlation algorithm is reduced since the features are the most distorted.
The disadvantage with more frequent retraining is that this could lead to the
type of displacement bias observed in the second loading interval shown in Figure 8.
Although this bias is small relative to the system resolution, the reason for its
appearance as well as the frequency of occurrence is unknown. These questions could
be easily answered with a series of well designed and controlled diagnostic
experiments.
Machine vision algorithms are inherently time consuming. The BISMAP

displacement output is displayed in real-time but the update rate between
measurements is relatively slow. The newest Cognex 4400 was capable of calculating
2.5 points/second. Assuming over-sampling occurs, the realistic output rate is a strain
value every 3-5 seconds. Although the applications presented in this paper could
accommodate this rate, some other tests will be limited by this sample rate.
Applying the BISMAP System
The results presented have indicated that the system is feasible, and system
performance for a variety of different materials was examined. However, using the
biaxial BISMAP extensometer for other materials or applications requires addressing
two issues. First, the strain range required must be sufficiently large when compared
to BISMAP system resolution. As a guideline, if the behavior observed results in
strains less than 5000 ~te, resolution will prove to be a problem. This was apparent
during the tensile testing of the aluminum sample.
Given an appropriate strain level, the second issue of importance is the surface
texture. Sufficient texture must exist to repeatability acquire and track the features
surrounding the control points. Experience has demonstrated that the standard
deviation of pixel intensity can be used effectively to assess the degree of texture. If a
surface in its natural form has a standard deviation less than 4.0, some external
method will need to be used to increase texture and pixel intensity. All of the four
methods identified in Table 1 have proven effective for supplementing surface
features.
McKEIGHAN ET AL. ON BISMAP 1 19
The ability to retrain image models during a test implies that there is no upper
limit on strain measurement. However, it should be noted that retraining at non-zero
loads may result in error accumulation since the strain response will be incrementally
developed. Nevertheless, retraining makes the BISMAP system particularly useful for
high deformation rubber, plastic, fabric and biological materials. In addition, since
BISMAP does not require either contact with or targets on the specimen, it can be
used to assess deformation in flexible and fragile materials that are difficult or
impossible to test with conventional extensometry.
Future Technological Developments
Modern computer technology is expanding at such a rapid pace that new

equipment quickly becomes obsolete. The system described herein has been
developed using components obtained within the past five years. Faster computers,
frame grabbers and image processing equipment will contribute to better system
performance and a decreased update time. Newer generations of CCD cameras will
also contribute to enhanced performance. The resolution calculation detailed in
equation (1) clearly illustrates the importance of camera pixel density.
Resolution can also be increased by making a slight modification to the optical

approach used. This is schematically illustrated in Figure 14 where either two cameras
can be used or an image splitter implemented optically. Either technique would result
in an increase in the effective gage length which will then have a proportional effect
on resolution. This technique can be combined with a higher resolution CCD camera
to yield a dramatic increase in resolution.
FIG. 14--Dual lens configuration designed to increase system resolution.
It is not impractical to use this type of a split image or dual camera approach
to gain a 5x increase in gage length. The newest generation of CCD cameras have
1200x1200 pixels. This represents a 2.3x increase in capability over the CCD camera
used in this work. Combining these two technological improvements results in a
resolution increase of nearly 12x. The concomitant change in the maximum resolution
(Table 1) is from 460 lae to 40 ~tr This enhancement would greatly increase the low
strain performance of the system using technology currently available, but it will
decrease the flexibility of the system to make measurements from geometric
discontinuities such as holes and notches. Although this technique will improve strain
resolution in the principal loading direction, the improvements incurred in the
orthogonal direction will be less.
Nr Generation BISMAP
The system presented represents a prototype developed to demonstrate the

overall concept. By using off-the-shelf Cogaex image processing technology, little
experience using machine vision is necessary to implement and use the system.
Although the refinements discussed earlier would increase the cost, the basic hardware
cost for the complete BISMAP system is approximately $30K. This hardware cost is
not considered high for this type of specialized, unique instrument.
Enhancements to the BISMAP system could be incorporated into a laboratory

instrument developed from the prototype system described. For instance, analog
outputs for closed-loop control of the test system could be provided. The system
could then be used for complicated testing under either constant or variable amplitude
loading that necessitate closed-loop strain control. With a laboratory instrument, it
will be possible to make measurements from pliable and delicate materials more easily
than currently possible. The BISMAP system is viewed as a supplement, not
replacement, for techniques currently used to measure high resolution displacements.
A new laboratory instrument could also be used for measurements where

conventional techniques are unsuitable. Some possible uses envisioned include:
(1) measurements of displacements of biomaterials and implanted devices in

situ. These measurements could be done in conjunction with finite
element modeling to obtain the relationship between loading and material
response;
(2) measurements of material response at very high temperatures in furnaces.

This application makes use of the non-contact and non-target
characteristics of BISMAP. The biaxial capability will also be useful for
measurements near geometric discontinuities. The two-axis understanding
of deformation behavior will provide data suitable for more fully
understanding the link between material structure (or microstructure on a
smaller scale) and mechanical response.
CONCLUSIONS
. A prototype biaxial, non-contact displacement measurement system has been

developed and demonstrated with strain measurements on aluminum,
polycarbonate, rubber and tendon materials. The overall system hardware costs
are approximately $30K.
. The operation of the system was verified by comparing measured strains to a
foil resistance strain gage. BISMAP strains differed by an average of 370 lar
(range of 200 to 600 lar with an average variability of 50 lar as measured on a
LEXAN tensile specimen using a strain gage as an absolute indicator. The
prototype system exhibited good repeatability approximately equal to the
overall strain accuracy and resolution.
. The BISMAP system achieved subpixel displacement accuracies typically better
than 0.1 pixels. The output data rate consisted of a strain value every 3-5
seconds (this includes time averaged sampling time to digitally filter the strain
readings). The system was capable of measuring large strains and, with
retraining, has no practical upper limit on the measurable strain.
. The BISMAP system shows promise for development into a laboratory
instrument for measuring strains or Poisson's ratio in pliable or delicate
materials. Its primary advantages include non-contact measurement in two-
dimensions, variable gage length, direct surface measurement (no external
targe0 and real-time measurement of multiple, discrete points.
ACKNOWLEDGEMENTS
The authors are grateful for support of this research by the Southwest Research
Institute Internal Research Program and the Advisory Committee for Research.
Messrs. Byron Chapa, Joe Fey and John Campbell were also instrumental in the
development of this prototype system.
REFERENCES
[1] Dally, J. W. and Riley, W. F., Exoerimental Stress Analyses, McGraw-Hill

Book Company, New York, NY, 1978.
[2] United Calibration, Inc. (Huntington Beach, CA), Laser Extensometer Model
Ext-62-LOE.
[3] Zygo, Inc. (Middlefield, CT), Laser Extensometer Model 1102HS.
[4] Jia, Z. and Shah, S. P., "Two-Dimensional Electronic Speclder Pattern
Inferometry and Concrete Fracture Processes," Experimental Mechanics,
pp. 262-270, September 1994.
[5] Chao, Y.-J. and Sutton, M. A., "Accurate Measurement of Two and Three
Dimensional Surface Deformations for Fracture Specimens by Computer
Vision," in Exoerimental Techniques in Fracture, J. S. Epstein, ed., VCH

Publishers, New York, NY, pp. 59-93, 1993.
[6] Lyons, J. S., Liu, J. and Sutton, M. A., "High-Temperature Deformation
Measurements Using Digital-Image Correlation," ExperimentalMechanics,
Vol. 36, No. 3, March 1996.
[7] Franke, E. A., Wenzel, D. J. and Davidson, D. L., "Measurements of
Microdisplacements by Machine Vision Photogrammetry (DISMAP)," Review
of Scientific Instrumentation, Vol. 62, No. 5, May 1991.
[8] Seida, S. B., Franke, E. A. and Davidson, D. L., "Investigation of Non-contact
Biaxial Strain Measurement and Control Using Machine Vision," Southwest
Research Institute Project 10-9806, Final Report, 1995.
[9] Silver, W. M. and Schatz, D., "AI Beefs Up Gray-Scale Correlation," ,The
Electronic System Design Magazine, pp. 89-92, March 1988.
[10] Davidson, D. L., "The Characterization and Measurement of Displacements and
Strain by Stereoimaging," Scanning Electron Microscopy/1979/II, Society for
Experimental Mechanics, AMF O'Hare, IL, pp. 79-86, 1979.
[11] Williams, D. R., Davidson, D. L. and Lankford, J., "Fatigue Crack Tip Plastic
Strains by the Stereoimaging Technique," Experimental Mechanics, Vol. 20,
No. 4, pp. 134-139, 1980.
[12] Plastics: Edition 7, The International Plastics Selector, Inc., San Diego, CA,
p. 275, 1985.
James L. Rand I and Debora A. Grand
OPTICAL MEASUREMENT OF BIAXIAL STRAIN IN THIN FILM POLYMERS
REFERENCE: Rand, J. L., and Grant, D. A., "Optical Measurement of Biaxial

Strain in Thin Film Polymers", Nontraditional Methods of Sensing Stress, Strain, and
Damage in Materials and Structures, ASTM STP 1318, George F. Lucas and David A.
Stubbs, Eds., American Society for Testing and Materials, Philadelphia, 1997.
ABSTRACT: A non-contacting optical method for the measurement of strain

simultaneously in two directions has been developed in the process of characterizing
the behavior of thin film polymers. The thin film specimen geometry is a cylinder
created by heat sealing two layers of the material together. The technique utilizes a
random speckle pattern that is painted on the surface of the specimen. Images are
captured by a CCD camera of a small portion of the painted region as the material
creeps under load. A digital image correlation technique is used to analyze the
movement of the speckle pattern over time. The resulting displacement field obtained
from these images is then used to determine the strain on the surface of the specimen.
This strain measurement technique has also been extended to the measurement of
biaxial strains on the surface of thick sheets as well as thin films.
KEYWORDS: non-contact strain measurement, biaxial testing, biaxial creep

response, image correlation, optical strain measurement
The constitutive behavior of thin film polymers has been an area of study by
the authors for many years. The initial work entailed the uniaxial creep response of
linear low density polyethylene (LLDPE) under constant load [1]. As the research
i President, Winzen Engineering, Inc., 12001 Network Boulevard - Suite 200, San
Antonio, TX 78249.
2 Staff Engineer, Winzen Engineering, Inc., 12001 Network Boulevard - Suite

200, San Antonio, TX 78249.
123
Copyright9
(PDVSA by ASTM
developed it became clear that, due to the anisotropy of the material created by the
extrusion process and the multi-axial loading of the structure, the need existed for the
simultaneous biaxiai measurement of strain in the film. A uniaxial test of this
material will produce a very pronounced tension field in the vicinity of the grip which
prevents a meaningful measurement of transverse strain. Therefore, it was necessary
to create a biaxial state of stress and develop a measurement technique to determine
the resulting strains in both directions simultaneously. Once the biaxial response of
the material was determined, a nonlinear, viscoelastic, anisotropic constitutive material
model was developed for the polymer film [2].
BIAXIALTESTING
There are many test techniques used to create a biaxial stress state in a material
and each test method has its own strengths and weaknesses. A standard method of
achieving a biaxial state of strain is the standard uniaxial test conducted in accordance
with ASTM Standard Test Methods for Tensile Properties of Thin Plastic Sheeting (D
882). However, the development of wrinkles in the axial direction of the sample
during this test prevents the measurement of transverse strain under conditions of
uniaxial stress. In addition to the presence of wrinkles, provisions must be made for
the time dependent variation in strain as described in ASTM Standard Test Methods
for Tensile, Compressive, and Flexural Creep and Creep-Rupture of Plastics (D 2990).
In an effort to eliminate the problems associated with wrinkling of the specimen, a
method was developed for creating a biaxial state of stress. The ASTM Standard Test
Method for Multi-Axial Tension Test for Geosynthethics (D 5617) involves the
pressurization of a circular diaphragm while observing the displacement of the center
of the specimen. In order to compute either a stress or a strain from the measured
quantity, the material must behave in a symmetric manner which would preclude its
use for anisotropic materials or unbalanced states of stress. An alternate method is to
simply clamp all four edges of a square sample of the material and load the two
principle directions separately. Unfortunately, this specimen geometry is notorious for
edge effects which produce a non-uniform stress and strain field. A final method is to
inflate a cylindrical specimen of the material. This test has the advantage that the state
of stress is biaxial and known from the measurement of pressure and geometry. All of
the mentioned test procedures have limitations which must be overcome in order for
the testing to be efficient and the data to be meaningful.
The cylindrical specimen is simply a thin-walled pressure vessel as shown in
Figure 1. As such, the magnitude of the radial stress is on the order of the pressure
which is negligible compared to the other normal stresses. Of greater interest are the
axial stress, CrA.~, and the hoop or circumferential stress, t~c,~, due to pressure which
can be shown to be:
APr APr
Oa'A~ - 2t and Oc, ap -
t (1)
where r is the radius of the cylinder, t is the thickness of the film, and AP is the
RAND AND GRANT ON OPTICAL MEASUREMENT 125
pressure differential across the film. By varying the test pressure, a variety of stresses
can be studied if the resulting strain can be measured. The limitation of this simple
loading is that the circumferential stress is always twice the axial stress; however, an
external axial force, F,, can be applied to the specimen to increase the axial stress
without increasing the circumferential stress. The stress, in a region away from the
grips, due the axial load can be determined by assuming that the force is uniformly
distributed around the circumference of the cylinder, such that:
Fa (2)
~ 2~rt
With the use of the cylindrical test specimen, many combinations of circumferential
and axial stresses can be created. The stresses on an element on the surface of the
cylinder are shown in Figure 1. The axial stress is the combination of the stresses due
to pressure in addition to the applied external force, while the circumferential stress is
clue only to a differential pressure.
O"A
T
ac <
__1
~A
Figure 1 -- Cylindrical Pressure Vessel
EXPERIMENTAL SETUP
The cylindrical specimens are mounted inside an insulated chamber where the
temperature may be controlled from 173 K to 303 K, Extreme care must be taken
when mounting the sample since the alignmem and balance of the specimen relative to
the loading mechanism inside the chamber is critical. Useful data will only be
obtained with meticulous setup procedures.
Specimen Preparation
From the manufacturing process, there is a film orientation. The film is rolled
in the machine direction (MD) and the transverse direction (TD) is orthogonal to the
machine direction. The coordinate system formed by these directions will be referred
to as "material" coordinates. When a structure is formed by this material, such as a
cylinder, the geometric description of the object is referred to "structural" (cylindrical)
coordinates. The specimen can be constructed with any arbitrary material direction
corresponding to the axis of symmetry of the cylinder. However, in order to collect
the data required for the development of a constitutive relationship, only three
orientations as shown in Figure 2 are considered. In the first orientation, the axial
direction of the cylinder is the machine direction of the film. The second orientation
has the transverse direction of the film in the axial direction of the cylinder while the
third cylinder orientation has the machine direction of the film rotated 45 ~ from the
axial direction of the cylinder. In order to maintain the machine direction across the
heat seal, the two layers of film are oriented at 90 ~ to each other.
I MD
MD
\
MD
MACHINE TRANSVERSE 45~ DIRECTION

DIRECTION DIRECTION
Figure 2 -- Orientation of Cylinders
For the specimens to have the same diameter at every point along the axis of
symmetry, the cylinders are constructed in the "layflat" configuration from two layers
of nominal 20 gm LLDPE film placed between glass plates which have a ground
straight edge. A torch is used along the edge of the glass to fuse the two layers of
film together, the excess material is cut away from the edge of the plate, and the torch
is used again to heal any discontinuities in the seal. The film cylinder has a diameter
of 17 cm which corresponds to the outer diameter of standard PVC pipe caps which
are used to react the pressure differential and introduce the external axial load into the
film. Once the pipe caps are attached to the cylinder a distance of 61 cm apart the
cylinder is pressurized with air to a pressure differential of 37.5 Pa to remove any
wrinkles in the surface. A random speckle pattern is applied to the film with an air
brush using ink that has been specially formulated for printing on polyethylene. The
specimen is then mounted inside the environmental chamber.
Environmental Chamber Configuration
A schematic of the environmental chamber is shown in Figure 3. The bottom

pipe cap is fixed to an interior support. Through this end cap, the cylinder is
pressurized with compressed air and the differential pressure measured. A pulley
system with a counter weight is used to support the top end cap. It is with this system
that the external axial force is applied. Sheaves rather than pulleys are used which
result in a nearly 100% transfer of the force which may be monitored with a load cell.
When a test temperature below room temperature is required, the chamber is cooled
with liquid nitrogen and a temperature controller maintains the chamber to +1 K of the
specified temperature.
A = Temperature Controller
m B = Cryogenic Solenoid Valve
C = Temperature Monitor
D = Pressure Regulator
E = Pressure Gauge
F = Shear Force
9 D3 Liquid
Ni~'ogen G = PVC Pipe Caps
I
H = Load Cell
I = Compressed Air
j z Video Target
1, 2, 3, = Temperature Probes
| = Sheave
hs~at~viroam~
~ber
Figure 3 -- Environmental Chamber Configuration
Specimen Alignment
Special consideration is given to the alignment of the cylinder inside the

chamber. Since the top end cap is suspended from the pulley system, the weight
distribution of the clamping attachments is balanced out so that the end cap is level.
The bottom end cap is aligned directly below the top end cap and is shimmed to be
level. The camera is placed a set distance away from the chamber and views the
cylinder approximately 16 cm from the bottom end cap through the chamber window.
The camera is aligned with the cylinder by hanging two plumb lines from the front
and the back of the top end cap. When properly aligned, the two lines will appear as
one. These plumb lines are removed prior to testing.
Image Collection
A small area on the surface of the cylinder is observed with a CCD camera.
The rectangular array of 512 by 480 pixels is used to capture the light intensity of this
area. The pixel spacing is not the same in the two directions but does have a fixed
relationship of 5 to 4. For the purpose of subsequent image correlation, a reference
image is required in an unloaded state of stress. A small pressure differential of 37.5
Pa is used to remove any wrinkles in the cylinder for the reference image. The
intensity histogram of the reference image is checked to verify the distribution and
intensity levels. After loading the cylinder to a given stress state, images are taken at
discrete times as the specimen deforms under constant load. The deformed images are
compared to the reference image with the image correlation technique to determine the
strain.
STRAIN MEASUREMENT
Normal strain is the elongation per unit length of a material. With many
materials, the elongation can be measured by the crosshead displacement or with
extensometers such as in tensile tests. However, both the modulus and stiffness of
polyethylene films are small which precludes the introduction of any mechanical
device such as a resistance strain gage. In addition, since film thicknesses may vary
by ten percent over a very small distance, the measurement of crosshead motion would
produce only an integrated value of strain over the entire length of the specimen. In
order to obtain acceptable resolution of the data, precise measurement of relative
displacement is necessary over a very small gage length. Optical techniques are
becoming popular for the measurement of strain. One of the simplest methods
involves placing a grid or reference points on the surface of sample and photographing
the deformation due to stress. The displacement of the reference points must be
manually measured from the photographs to determine the strain. This technique,
although simple, is quite labor intensive and was used in the uniaxial testing of thin
films[l]. The elongation was determined by recording the movement of two gauge
marks located 61 cm apart with video cameras. It should be noted that the large gage
length was required to obtain the desired resolution. In the biaxial testing, however,
it is impractical to have large diameter cylinders to obtain the desired strain resolution;
however, this technique could be used in conjunction with a microscope as was
reported by Brinson [3].
Another optical method, the digital image correlation technique, was
determined to have sufficient resolution and has been proven successful in the
measurement of strain [4],[5]. This technique uses the cross-correlation integral to
determine the location of maximum correlation between two digital images. If
identical images were compared, the correlation would be one. In the biaxial testing,
the digital intensity image taken of a specimen under load (named the deformed
image) is compared to the image taken before the specimen is loaded (named the
reference image) to determine the horizontal and vertical displacements of the image.
A random speckle pattern of ink, applied to the specimen surface prior to testing,
creates a unique intensity distribution which is tracked through the image correlation.
It is obvious that the deformed image will not be identical to the reference image,
since the random pattern will become larger or smaller or rotate due to the applied
stresses, and the maximum correlation will be less than one.
The CCD camera used captures images which are 512 X 480 pixels; however a
central region of 380 X 380 pixels is used in the correlation. There are three main
steps involved in the image correlation which are used to determine the horizontal and
vertical displacements of a 160 X 160 pixel region. These steps are:
1.) Take a subset (50 X 50) from the center of the reference image and fred the
location of maximum correlation in the deformed image. The location of
maximum correlation defines the "coarse" horizontal and vertical displacements.
2.) The "coarse" displacements are then used to locate a 160 X 160 region of
the reference image which exists in the deformed image. A smaller subset size
(20 X 20) is then used to determine the location of maximum correlation. This
step is done to increase the resolution of the displacement. If the displacements
are too large, the region in the reference image will be out of the field of view
of the deformed image and the image correlation routine will end. This step is
repeated until the 20 X 20 subset has mapped the entire 160 X 160 region.
3.) Once the pixel location of the maximum correlation is determined, a

subpixet interpolation is conducted to again increase the displacement resolution.
Since the camera is looking at a fixed location in space, as the cylindrical

specimen strains the target area from the reference image will be displaced from its
original location. With the current parameters, the target material captured in the
reference image can be displaced 1.143 cm horizontally and 0.889 cm vertically when
captured in the deformed image and still be in the camera field of view. This
corresponds to a circumferential strain of 7.2% and an axial strain of 5,5%. These
strain values are dependent upon the distance between the target and the fixed end of
the cylinder as well as the configuration of the CCD array. If greater displacements or
strains are encountered during processing, the image correlation routine will
automatically end. The image correlation routine itself can detect strains up to 12.5%
based on maximum pixel movement.
In order to determine strain from displacement, the definition of strain must be
recalled in terms of the two-dimensional structural coordinates observed by the
c~m'lera:
au ; ey- av . au + av (3)
where u is the displacement in the horizontal, x, direction and v is the displacement in

the vertical, y, direction. Since the strain is the gradient of the displacement, one
approximation would involve the assumption that the strain is constant between the
point of observation and the fixed base of the cylinder. This may not be a good
assumption in the region of the base, and no similar method exists in the
circumferential direction. However, the displacement is computed for every 20 X 20
array in the 160 X 160 field. If the strain were a constant in this field, the
displacements would appear as a plane tipped at some angle. This angle may then be
used to determine the derivative of each displacement in both directions. These
derivatives may then be combined to compute both the normal strains and the shear
strain. Differentiation of digital data from point to point would require a large number
of points in order to "smooth" the data. Therefore, a plane is fitted to the
displacements by a least squares method in order to determine the displacement of the
center, the angle of the plane with each axis and subsequently the strains. This
procedure is conducted on both the horizontal and vertical displacement fields. A
typical horizontal displacement field is shown in Figure 4.
Figure 4 - Horizontal Displacement Field at 88 minutes for a TD Cylinder

Loaded with 0.5 MPa Circumferential and 3.5 MPa Axial at 296 K
DATA R E D U C ~ O N
Once the horizomal, u, and vertical, v, displacement fields are determined from
the image correlation routine, a plane is fitted to each field by the method of least
squares. The slopes or angles with respect to the horizontal and vertical of each of the
planes are used to determine the strain in the specimen. The horizontal displacement
field produces the measured circumferential strain, Sc, ~u,~d =/gu/ax, and one part of
the shear strain, au/ay. The vertical displacement field produces the axial strain,
s~,m~,,~ = av/ay, and the second part of the shear strain, Ov/ax. The shear strain
components must be corrected for the nonuniformity of the CCD array since the pixel
spacing in the horizontal direction is different from the pixel spacing in the vertical
direction. The measured shear strain, Yxy,~ = ~ , is determined from:
Yxy = 1.25 au + 0.8 av (4)

9" ~ " ~ ay ax
The axial and circumferential strains do not need to be corrected for this
nonuniformity, since the displacement of the pixels is in the same direction as the
pixel spacing.
As the target material undergoes deformation, the optical image of that material
will be altered for a variety of reasons, not the least of which is lens distortion or a
change in the target location. Therefore, a number of tests were conducted where the
target was a rigid body and allowed to displace or rotate a known amount without
strain. The target area may undergo rigid body displacements in three directions:
horizontal, vertical and out-of-plane. Images were collected and correlated as
described previously in the section entitled Strain Measurement and the "apparent"
strain was determined for each type of rigid body motion. The "apparent" strain
occurs in the absence of any actual strain and the computed strain values must be
corrected for these effects.
The "apparent" strain due to rotation was determined by rotating a rigid disk,
with a speckle pattern painted on the surface, + 7 ~ Images were taken every degree
and were compared to a reference image taken with no rotation. The image
correlation routine measured a strain in the material even though there was only rigid
body rotation. The amount of rotation, m, was determined from the following :
1.au &
o, = g - g) (s)
The resulting "apparent" strain due to rotation was determined to be a function of the
rotation as shown in Figure 5. The "apparent" circumferential strain due to rotation
was found to be equal to the "apparent" axial strain due to rotation and can be
modeled as follows:
1.2\o~
This "apparent" strain due to rotation, in units of percent, must be subtracted from
both the measured circumferential and axial strains to correct for perceived strain due
to rigid body rotation. As seen in Figure 5, there is a small "apparent" shear strain
due to rotation; however it was assumed to be sufficiently small to not require a

correction.
Motion perpendicular to the surface of the cylinder, or out-of-plane, will occur
if there is rigid body motion as well as when there is circumferential strain. The
circumferential strain resulting from a pressure differential across a cylinder will
produce an increase in the cylinder radius. This increase in radius displaces the
cylinder material out of the original plane. In a cylindrical coordinate system, the
circumferential strain is the change in the circumferential dimension with respect to its
original position, or:
AI
EC - (7)
ro
Therefore, whenever a circumferential strain occurs it will be measured as 0u/0x by

the image correlation method. However, this strain cannot occur without moving the
target area toward or away from the camera by a very small amount. Therefore, a test
was conducted by moving an unstrained target a small, precisely measured, distance
toward the camera. It was found that an "apparent" strain developed in both directions
which was proportional to the small displacements from the original focal plane.
When this displacement was normalized with respect to the cylinder radius the
apparent strains were found to be linearly related to the normalized displacements.
The constants, k~ and k 2, represent the linear relationship between the apparent strains,
s c and SA, and the out-of-plane displacement created by the circumferential strain.
The measured circumferential strain must be corrected by this empirically
determined strain to achieve:
EC = EC~neasured-ea~arent,rotaaon (8)
1 +k 1
where the constant k~ was equal to 0.1281. A similar correction is applied in the axial
direction due to the motion out-of-plane. This correction is also proportional to the
circumferential strain and is therefore given by
E axia l = (EA,mea~urcd - Eapparent,rotation) - ~ + k l ~. C,mea.s'urc.d apparent~rotation j

(9)
where k 2 = 0.14437. It was found that horizontal and vertical rigid body translations
within the reference plane do not produce any "apparent" strains.
0.4
0 EC i x : u m ( o ~ , a l
~_ A ~ Axmi
0.2 + y~y
U~
e- "Apparent" Strain Model
+ + .~ + + + +
e-- 0.0
c--
UJ
§ ..%
~ -o.2
J
i i I I i I I I I I ~ I
< -0.4
=
-10 -8 -6 -4 -2 0 2 4 6 8 10
Rotation, co
Figure 5 -- "Apparent" Strain due to Rotation
SAMPLE DATA
Data has been collected at temperatures between 296 K and 223 K. For each
test, the specimen was soaked for 10 minutes at the test temperature in a stress free
state to allow all thermal strains to develop. The sample was then slightly pressurized
and the reference image captured. Shortly after the reference image was acquired, the
cylinder was loaded with the desired pressure and external force and the test was
initiated by manually starting the image capture program for a duration of 30 minutes.
Images are acquired at 30 second intervals at first and 3 minute intervals as the test
progresses.
The captured images are then processed by the image correlation program
resulting in a set of displacement files in units of pixels. At each time, planes were
fitted to the data and the strains determined from the slopes of the planes. The
measured data was then corrected for rigid body rotation and rigid body out-of-plane
translation, as discussed in the previous section. An example of the computed strains
as a function of time is shown in Figure 6. In the presentation of the data, the circles
are consistently used to denote the circumferential strain, the triangles are used to
denote the axial strain, and the squares denote the shear strain. All strains reported are
due to the applied loads at a given temperature and do not reflect any deformation due
to thermal contraction.
The purpose of collecting creep strain data for a variety of loading conditions is
to determine the appropriate constitutive relationship of the material. This relationship
is developed in material coordinates from data collected in structural or laboratory
coordinates. The relationship between these two coordinate systems is geometric and
given by the following transformation:
Ey "= sinZ0E1 + COSLOE2 + s i n 0 c o s O E 6

(10)
~= = cos20~ + sin20~2 - sinecose%
If the axis of the cylinder is either in the machine or transverse material directions,
then the two strains measured in laboratory coordinates are equal to corresponding
strains in material coordinates since 0 is either 0~ or 90~ However, the shear strain
in material coordinates, c6, may be found from the difference in normal strains if the
cylinder is constructed such that the axis of symmetry, is 450 from the material axis.
Substitution of this angle into equations (10) yields
r = ~r - % (11)
The shear strain obtained from a cylinder constructed in this manner is shown in
Figure 7.
4
O MD Strain
A TD Strain
D Shear Strain
& A ~ &A & AAAAAA A , ~ , A & _ -
2
&
._= 1
r 0 ID D O D D ~ D DDDE:E~ 1:3I D D ~
r
.B
=_
(D
0 0 0 0o0 o0oo0o0 o 0 0 o ~
t-
ILl -2
-3
3 4 5 6 100 2 3 4 5 6 101 2 3 4 S 6 102
Time (Minutes)
Figm'e 6 -- Strain as a Function of Time for a TD Cylinder Loaded with 0.5 MPa
Circumferential and 3.5 MPa Axial at 296 K
6 I~ ~ ~ I I I q t I II b I . . . . . I I i I . i I I1~ ~ [
~3 0 0 0 O0 0 0 00000 0 0 0 0 ~
0
~2
9 1
O9
0
2 3 4 s 6 10 0 2 3 4 s e 101 2 3 4 s 6 10 2
T i m e (Minutes)
Figure 7 -- Shear Strain as a Function of Time of a 45 ~ Cylinder Loaded with 4 MPa

Circumferential and 2 MPa Axial at 296 K
DISCUSSION OF RESULTS
The effort described in this paper represents the new application of a recently
developed technique intended to observe the strain behavior of a linear low density
polyethylene film which is nonlinear, viscoelastic and anisotropic. The image
correlation methods utilized in this research are well documented in the literature.
However, the results of such analysis must be understood to be the best correlation of
two dissimilar displays of pixel intensity. This is not a measurement of strain in the
normal sense, but instead generates an amount of data which the engineer may process
in any manner which will provide meaningful information. Therefore, the research
task was to fred a method to not only relate pixel location to displacement, but to
evaluate the derivatives of displacement in any direction. The relationship between the
observed camera image and the displacement of the target was made somewhat more
challenging by the fact that the target surface had a radius of curvature in one
direction while the camera sensor was a plane. As a result, the camera image would
sense an "apparent" strain for certain rigid body movements of the target. It was
necessary to "calibrate" the camera by developing a numerical technique which would
eliminate the effects of any rigid body motion even though it was caused by the same
strain that was to be measured.
A significant achievement in the application of this technique is the observation
that shear strain, in material coordinates, is directly obtainable from the difference in
the normal strains produced in a 45 ~ cylinder. Data obtained from this test give a
direct measurement of nonlinear viscoelastic shear response of an anisotropic material
without the application of a torque and resulting rotations in laboratory coordinates.
The research described in this paper represents a new application of an optical
technique for the transient measurement of biaxial normal and shear strains in thin
films without the normal problems associated with end effects, wrinkling or
disturbances from intrusive gages. The results are of sufficient accuracy that the
biaxial strain measurements have been used in the development of a nonlinear,
anisotropic, biaxial, viscoelastic constitutive model for linear low density polyethylene.
The development of this constitutive model has been made the subject of a separate
paper [2].
The two dimensional image correlation technique can be applied to more than
just cylinders constructed of thin films. For example, a 1 mm thick sheet of
commercially available geomembrane material was loaded uniaxially with 4 MPa of
stress in the machine direction. The resulting strains are shown in Figure 8. In
addition, the recovery of the material can be monitored by removing the weight and
continuing the image correlation process. It is believed that this is the first reported
use of the image correlation procedure to quantify the strain in a geomembrane.
CONCLUSIONS
Based on the research effort presented, it may be concluded that an image

correlation technique may be used to measure the two dimensional strains caused by a
biaxial state of stress. The alignment of the entire system is critical in obtaining
accurate data. In addition, care must be taken to correct all measured images for rigid
body displacements as well as rotation. It has been shown that shear strain, in material
coordinates, is easily obtained by orienting the specimen in a direction 45 ~ to the
material coordinates.
The image correlation technique developed for the biaxial testing of thin films
has also been extended to thicker materials and different stress states. Due to the
nature of the film tested, it was necessary to construct a cylinder of the material to
prevent wrinkling. However, when thicker materials are tested, fiat sheets may be
loaded uniaxially to produce biaxial strains.
3 A A AA A AAAAAA A AAAA~SA
2 A A
.c__
.t* 1
w,w
0O
0 [] D O DOE] [] D D n o D O [] nODDCs
f..
.r-
m -1 o 0
.E 0 000
o Oooooo o Oooooo
-2 0 TD Strain
UJ
A MD Strain
-3
O ShearStrain
9 i
-4
3 4 56 100 2 3 4 s 8 .vtnl 2 3 4 5 8 102
Time (Minutes)
Figure 8 -- Strain as a Function of Time for a Sheet of Geomembrane Material

Loaded with 4 MPa in the Machine Direction at 296 K
REFERENCES
[1_] Rand, J. L., Henderson, J. K. and Grant, D. A., "Nonlinear Behavior of Linear
Low-Density Polyethylene," Polymer Engineering & Science, Vol. 36, No. 8,
April 1996.
I2] Rand, J. L., Grant, D. A. and Strganac, T., "The Nonlinear Biaxial
Characterization of Balloon Film," 34th Aerospace Sciences Meeting & Exhibit,
January 15-18, 1996, Reno, N-V, AIAA Paper No. 96-0574.
[3_] Brinson, H. F. "Micro-Measurements of Mechanical Properties for Adhesives and

Composites Using Digital Imaging Technology," Proceedings of the 10th
International Conference of Experimental Mechanics, Lisbon, Portugal, July 18-22,
1994.
Peters, W.H., Ranson, W.F., Sutton, M.A., Chu, T.C., and Anderson, J.,
"Application of Digital Correlation Methods to Rigid Body Mechanics", Optical
Eneineerinz. Vol 22, No. 6, 1983.
Chu, T.C., Ranson, W.F., Sutton, M.A., and Peters, W.H., "Applications of Digital
Image Correlation Techniques to Experimental Mechanics", Experimental
Mechanics, Sept. 1985.
Jaime F. Chrdenas-Garcia, ~Mason M. Wu, ~ and Javad Hashemi j
A REVIEW OF STRAIN MEASUREMENT

TECHNIQUES USING THE GRID METHOD
REFERENCE: C~denas-Garcia, J. F., Wu, M. M., and Hashemi, J., "A Review of
Strain Measurement Techniques Using The Grid Method", Nontraditional Methods of
Sensing Stress~ Strain, and Damage in Materials and Structures, ASTM STP 1318, George
F. Lueas and David A. Stubbs, Eds., American Society for Testing and Materials,
Philadelphia, 1997.
ABSTRACT: The grid method is the oldest and simplest method for strain measurement.
It is an experimental technique in which a grid is applied to the surface of a specimen. A
grid is defined as one or more lines, dots or other shapes assembled in regular or irregular
patterns. A comparison of the deformed grid on the specimen, undergoing a general state of
loading, to that of a reference state yields the state of deformation. The emphasis in this
paper is on reviewing experimental techniques emphasizing optical, non-contact, non-
interfering and automated means of making deformation measurements using the grid
method. Thus, the objectives of this paper are: to briefly review the basic principles and
applications of the grid method in transparent and opaque solids, to describe the state-of.
the-art, and, to point in the direction of its future evolution.
KEYWORDS: grid method, diffraction pattern, elastic deformation, plastic deformation,

full-field strain measurement, optical diffraction gratings, optical measuring instruments,
strain tensor, video imaging, image processing
A number of experimental techniques have been developed for measurements of

strain and/or deformation over two-dimensional fields, including moir~ [1, 2], holography
[3], and speckle [4]. All of these techniques are termed full-field because a single picture or
image of the specimen surface is all that is required to obtain a qualitative assessment of the
Associate and Assistant Professor, respectively, Department of Mechanical Engineering,

Texas Tech University, Lubbock, TX 79409-1021
2 Senior Engineer, Bechtel Corporation, 3000 Post Oak Boulevard, Houston, TX 77252-
2166
138
Copyright9
(PDVSA by ASTM
to License Agreement. No www.astm.org
further reproductions authorized.
CARDENAS-GARC1A ET AL, ON USING THE GRID METHOD 139
deformation over the entire area of interest. To obtain quantitative data, it is necessary to
perform more detailed analysis and/or interpretation of the fringes generated by these
techniques. If done manually this analysis is laborious, time consuming and requires highly
skilled practitioners. The end result is usually discrete data along loci of points on the
specimen. In summary, all of these so called full-field techniques depict qualitative data
which is continuous, but further analysis yields quantitative data which is discrete.
Px
P2
Pz
Pt
Figure I - Schematic diagram showing various types of grids on the surface
of a specimen undergoing generalized2D loading
The grid method is an established experimental technique in which an identifiable

grid (whether naturallyoccurring,or artificiallyapplied)existson the surfaceof the
specimen of interest.The grid may consistof one or more lines,dots or other shapes,
assembled in regularor irregularmotifs or patterns. A comparison of the deformed grid on
the specimen, undergoing a general stateof loading,to thatof a reference stateyieldsthe
stateof deformation. A schematic diagram of variationsin grid configuratiom on a
specimen undergoing generalized2D loading is shown in Figure I. Note in particularthe
five dots shown in the upper fight-handcomer of the specimen, which are shown before
deformation (filled-indots) and afterdeformation (open dots). Itis clearthatusing this
basic 2D grid,consistingof three dots,a 2D strainand/or deformation measurement is
possible.
There are several characteristics of the grid method which complement the
experimental techniques mentioned above and make it particularly attractive for use in
deformation measurements. Grid techniques in their simplest form can be used to give a
very gross or very fine determination of strain, depending on the size and the number of
patterns used. Thus, depending on the relative scale of the regular or irregular patterns,
strain can be fotmd on a macro- and/or micro-scale. Further, depending on the rapidity of
analysis, the grid method can be adapted for use as a small area or as a full-field strain
measurement technique.
Consider the 2D specimen grid shown in Figure 2 consisting of equidistant parallel

and orthogonal lines (a cross-grating), e.g., with a pitch of 0.1 mm (10 lines per mm). This
grid allows the interrogation of the specimen on a macro- and/or micro-scale, depending on
the gage length used for the required measurements. It would be similar to having
thousands o f orthogonaUy-oriented strain gages of many sizes over the whole specimen
surface. When the analysis is done on a macro-scale a ruler is sufficient to measure strain,
e.g., if the 5 mm gage marks in Figure 2 are used to make a deformation measurement. If
done on a micro-scale, a microscope or an optical comparator may be used, as may be done
using the 0.1 mm marks in Figure 2. If the data gathering process could be automated and
done very rapidly, then a full-field measurement would be the end result. Thus, if enough
P2
f
Figure 2 - Schematic diagram showing the use of a grid for micro- and macro-scopic
measurements on the surface of a specimen
CARDENAS-GARCIA ET AL. ON USING THE GRID METHOD 141
measurements are made with the required spatial density, this simple technique permits the
gathering of discrete quantitative strain data which can be viewed on a continuous
qualitative basis. This is the opposite of the previously mentioned full-field techniques
where the qualitative data is initially obtained on a full-field basis, but the quantitative data
can only be accessed on a discrete basis.
To summarize, the grid method is the simplest and most direct approach for the
measurement of deformation. Its use has been evolving over time, in parallel with
complementary technological developments. Some examples of evolving technology
which have helped the evolution of the grid method are: the ability to make reproducible
rulings or gratings, and the development of the laser. These developments have made
possible the non-contact interrogation of the grid. More recently, the development and
common use of fast micro-computers and video technology have led to new approaches to
the use of the grid method which emphasize the automation of the data gathering process.
These developments motivate this paper. Consequently, the emphasis in this paper is on
experimental approaches emphasizing optical, non-contact, non-interfering and automated
means of making deformation measurements using the grid method. Thus, the objectives of
this paper are: to briefly review the basic principles and applications of the grid method in
transparent and opaque solids, to describe the state-of-the-art, and, to point in the direction
of its future evolution.
BASIC CONSIDERATIONS
As mentioned before, the grid method provides the most direct approach to measure
deformation. To apply it requires the use of markers on the specimen which are tracked as
the specimen deforms when undergoing loading [5, 6]. The 1D measurement of strain
requires at least two markers, denoted as points P and Q, as shown in Figure 3.
u
Before D fformation After Deformation
(_
Figure 3 - 1D measurement of deformation and/or strain
If the length between the two markers is measured before and after deformation, the
deformation is calculated from the following equation to be,
u=l-lo (1)
where u is the specimen deformation, and, 1oand 1are the initial and final gage lengths,
respectively. The normal strain is determined as,
U
e - (2)
lo
The simplest way to make this measurement is by manually keeping track of the
relative position of points P and Q before and after deformation, i.e., by keeping track of the
gage lengths as the material deforms. If the initial gage length is of the order of 25 mm, the
use of a ruler might be sufficient to make a deformation and/or strain determination for a
highly elastic material. If more precise measurements are required on an inelastic material,
the use of an extensometer, an optical microscope, or an optical comparator will allow such
grid method measurements.
Figure 4 iUustrates the 2D measurement of strain, which requires at least three

markers (P, Q, R) on the plane of the body. The lengths between these markers, which are
initially normal to each other, are denoted by (Io) x and (lo)y. The corresponding final
lengths, (lo) ~' and (lo)y', for markers P', Q' and R' are separated by angle 8~y.
~ ~ oR
) Before Deformation After Deformation
Figure 4 - 2D measurement of deformation and/or strain
The practical use of the grid method requires, as in the 1D case, the determination of
the relative position of the designated markers, before and after deformation. Once this
relative position determination, with the required accuracy, has been made then the normal
strains are given by,
l ! -- (o),
l ! - (Zo),
= (lo), e = (/o), (3)
and the shear strain is defined by,
y,y = - 0,, (4 )
where 0xr is the angle between line segments P'Q' and P'R'.
The strain values can be readily determined using these equations once a manual
assessment of the relative positions of the markers is made. This manual procedure can be
time-consuming and more or less cumbersome depending on the number of data points
used and/or the required accuracies demanded by the experimental application. Note also
that the markers can be made up of multifarious shapes and sizes, on a miero- and/or macro-
scale. This points to a weakness of using the grid method; except in the simplest
experimental situations, it is difficult, if not time-consuming, to make a multitude of
experimental measurements over the full-field of a specimen surface. This is what is
required in many practical applications. Thus the need exists to make the grid method more
amenable to automated interrogation to ease its applicability in experimental mechanics.
APPLICATIONS
The only comprehensive review of the grid method is that done by Parks [7]. Parks
defines the grid method as "... a method of applying a grid to the surface, or sometimes on
an interior plane of a specimen, measuring the distance between discrete points on the grid
both before and after loading, and an analysis of these measurements. The analysis of the
data consists in first obtaining the difference in position of points, before and after loading,
to determine displacements, and then dividing the difference between these displacements
by the distance between points to determine strain." Parks defmes the grid "... as an array of
lines or dots which indicates discrete points on the specimen. The array usually has a
repeated motif, with lines or dots occurring at some regular frequency. Often the array is
rectangular... Polar grids of radial and circumferential lines are also common...
Occasionally, irregular scratches are also used ..." In his review, Parks emphasizes the
various aspects of the grid method, i.e., choosing the grid pitch, the pitch precision, grid line
visibility, the application of the grid, recording the grid images, measuring the grid,
analyzing the grid and solutions of problems by grid techniques.
The approach taken in this paper, to select grid method applications for review, is
that of searching for other early and more recent approaches taken by researchers which add
relevancy to a review of the grid method. Approaches have been identified which get away
from a labor intensive approach to grid method analysis, and which attempt to overcome the
low level of accuracy of grid methods as compared to other strain measurement approaches.
These are two limitations which Parks notes in his review and which have been made the
primary focus here. In particular, methods which rely on optical, non-contact means of
making measurements are emphasized due to the potential that such measurements have for
automation.
To put these various applications in perspective let us list the desired characteristics
of an ideal grid method strain measurement approach, or for any idealized displacement
and/or strain measurement approach: (a) An optical, non-contact, non-interfering means of
making strain measurements; (b) a fully automated approach to grid method strain analysis
by using video technology coupled to micro-computers; (e) strain measurement on a
macro- or micro-scale by optimizing the size and number of grid markings used; (d) full-
field or small-area strain measurements as determined by the required rapidity and needs of
the analysis; (e) quantitative, discrete collection of strain data yielding a graphical display
of qualitative full-field information, versus other methods where qualitative, full-field data
yields discrete quantitative information; and, (f) high level of accuracy using grid methods
to make them comparable to other strain measurement approaches.
The applications to be reviewed fall into two broad categories: (a) experimental
techniques which track the motion of markers in the surface simulating manual data
acquisition (Digital Image Correlation falls under this category); and, (b) experimental
approaches which rely on the interrogation of periodic markings on the specimen using the
diffraction properties of a beam of monochromatic light and subsequent interpretation of the
Fraunhoffer diffraction peaks (Optieat Fourier Transform method which avoids the direct
measurement of the position of specimen surface markers).
Digital Image Correlation
Peters et al [8, 9, 10] have proposed and implemented a Digital Image Correlation
(DIC) technique applicable to the computation of surface strains and displacements. It
relies on the analysis of intensity patterns picked up from the undeformed and deformed
digital images of the object under consideration, with a random pattern of white speckles on
its surface. To obtain the needed images for analysis, a video camera and digitizer, attached
to a digital computer, are employed. The digitization procedure permits the digital coding
of physical scenes imaged by the video camera into rectangular arrays made up of picture
elements or pixels. The digitized images are then stored on a computer hard disk. The
underlying assumption is that there is a unique one-to-one correspondence between these
intensity pattems and the respective object surfaces. Thus, the principle of application of
DIC to the measurement of displacement and strain relies on being able to recognize the
differences between a set of two video images, i.e., to recognize the existing disparities
between the two images by correlating the two images. The disparity distributions are then
CARDENAS-GARCIA El" AL, ON USING THE GRID METHOD 145
directly used to get displacement distributions, from which the strain distributions may be
evaluated.
The determination of local similarities existing between pairs of images stored as

structured data sets, or the problem of image registration, is a basic task in computer vision.
In image registration, cross-correlation is frequently used to detect local similarities
between two images. Consider two video images of the same object, one before and the
other after deformation. Define a subset ~ in the undeformed image, centered at point A(i,
j), as an (n x n) pixels reference area. Also, define a bigger subset ~' in the deformed
surface image, centered at the same position, as an (m x m) pixel search area (m > n). The,
normalized cross-correlation formula in discrete form is written as follows,
n n
~I(i,j), I'(i + x,j + y)
C(x,y) = '=' j=' (5)
l (,,j) EEz
2] " 9
i=l j=l
O+x,j+ *2
where C is the correlation coefficient, I(i, j) is the gray level pattern of the subset f2 centered
at the point (i, j), I'(i+x, j+y) is the gray level pattern of the subset f~" centered at the point
(i+x, j+y) and, x and y range from 0 to (m-n+l). The correlation coefficient C is calculated
over the whole [(m-n+l) x (m-n+l)] area. The local deformation (u, v) of point (i, j) is then
found by maximizing Equation (5) above.
Based on Equation (5) it is possible to obtain sub-pixel accuracy when determining

displacements. Assume, as was done above, that the reference area is defined by an (n x n)
pixel subset, and the search area is defined by an (m x m) pixel subset (m > n). The
computation of correlation coefficients needs to be carried out over the area [(m-n+ 1) x (m-
n+l)]. The result is an array of [(m-n+l) x (m-n+l)] correlation coefficients, which
represents a 2D correlation distribution in discrete form. Among the [(m-n+l) x (m-n+l)]
coefficients, there is always a maximum value. If the required image registration accuracy
is +1 pixel, and the subset in the search area at which the maximum correlation coefficient
value is found is located at the point (i+u, j+v), then (u, v) is the disparity of the two images
at the point A(i, j). In terms of displacement measurement this means that the displacement
(u, v) of the point A(i, j) in the original image has been determined.
Mathematically, the maximum correlation coefficient value is usually located

between two pixels, rather than at a discrete pixel position. To achieve greater accuracy, a
more precise estimate of the location of the maximum correlation coefficient value is
needed. This would yield better estimates of displacement and strain. To achieve this, a
continuous correlation distribution is constructed by fitting the [(m-n+ 1) x (m-n+ 1)] discrete
correlation coefficients to a 2D curved surface. From this theoretical continuous
distribution, it is not difficult to find the maximum value with an optimum search
technique, thus achieving sub-pixel accuracy in the determination of the maximum

correlation coefficient value.
DIC is an experimental technique that has the potential advantage of automatic

measurement and control. Also, the scale of measurement is easily adjusted for large or
small areas, allowing the detailed determination of strain over the whole area under
consideration. This makes it possible to either obtain an average strain over an area, similar
to an electrical wire strain gage, or to obtain point data at many locations over the whole
area. Additionally, the micro-imaging of a small area permits using the surface
microstructure or surface inhomogeneities as gage marks. This approach has the added
advantage that marks do not need to be placed on the specimen surface.
Optical Fourier Transform Approach
The first to use an Optical Fourier Transform (OFT) approach to strain measurement
was James F. Bell. He was able to manufacture rods having 10,000 threads per inch (tpi)
[ 11] and used them as diffraction gratings for the measurement of dynamic strain associated
with elastic wave propagation in rods [12], large displacement measurements with extreme
accuracy [13], and, using rulings of 32,000 tpi to make high temperature measurements of
dynamic strain [14]. The basic principle for using the diffraction-grating technique
developed by Bell relates to being able to detect "the angles of the various diffraction
images, produced when monochromatic light falls on a series of equaUy spaced grooves
ruled on a specimen surface, depend upon the spacing of the grooves. When the specimen
is subjected to either tensile or compressive strain, of any magnitude, either elastic or
plastic, at any temperature up to within a few degrees of the melting point, the spacing of
the grooves is changed, producing a calculable change in the angle of diffraction for each
order. Whether the strain is quasi-static or dynamic, the measurement of the angular
changes of any two diffraction images will provide direct, and optically accurate,
measurement of both strain and angle" [15]. The detection of the motion of the diffxaction
images are done using the photoelectric detectors. It requires the focusing upon V-shaped
slits of the images whose angular change is to be measured. Behind each V-shaped slit is a
photomultiplier tube with the requisite frequency response for the monochromatic light
used.
A modification of the use of gratings to an application on a flat surface was

implemented by Douglas [16]. He sought to investigate the strains in the elastic-plastic
region in high-strength metal-sheet material. In doing so he developed the means to
manufacture reflective high density gratings on flat surfaces and to photographically record
the diffraction phenomena from multiple gages. The level of strain sensitivity achieved
with the use of this technique was 800 mierostrain.
Another adaptation of the diffraetographie strain gage was achieved by Pryor and
North [17]. They used the diffraction of light from a single aperture to measure strain.
Some features of this approach included: Sensitivities as small as 20 microstrain were
obsea'vable by eye; essentially linear over all strain ranges when small angles are used;
insensitive to lateral strains; inexpensive as well as easy to apply, use and interpret;
mountable by means which do not reinforce the specimen; no contact between specimen
and readout; readable by the naked eye if desired; easily made so as to be temperature
compensating; usable in hostile environments; stable with time, temperature and humidity;
and, readable by photoelectric means permitting dynamic strain measurements. A
subsequent application to the measurement of the deflection of a simply supported beam
loaded at the center was also reported in the literature [18].
Boone et al. [19, 20] used a slightly different approach to the determination of
strains using what they call "electro-optical processing of pictures of the deformed test
gratings," because of their interest in being able to rapidly obtain a large amount of strain
data in a specimen. The principle is based on the measurement of the local spatial
frequency and the orientation of the deformed grating by determining the exact location of
the diffraction points. The obtained measurements were performed along only one strain
axis.
Sevenhuijsen reported the use of a similar diffractographic technique for the

measurement of elasto-plastic deformations in 2D [21, 22, 23]. The research used glass
plate negatives to record the deformation of a grid on a specimen. A laser beam permitted
the interrogation of the glass negative generating diffraction patterns which could then be
analyzed and strain calculated. Absolute errors less than 100 microstrain were reported.
More recently the Video Optical Diffractometer for Strain Measurement (VODSM)
was developed which allows for the automated collection of discrete quantitative data and
their qualitative display for ease of interpretation, with full access to the quantitative data.
Previous applications include the experimental determination of the J integral for cracked
specimens. This has been demonstrated in a center-cracked 5052-H32 aluminuna panel
loaded in tension [24] and also in a double-edge-notched tensile panel of high strength, low
alloy steel [25]. The process of data collection, analysis and in-plane, strain tensor
evaluation has been completely automated and the required sensitivity is of the order of 200
microstrain. The main advantage of this experimental technique for strain measurement is
its ability to perform as an automated technique for the determination of the in-plane
components of the strain tensor: longitudinal strain, transverse strain, shear strain and rigid
body rotation. It permits a full-field qualitative representation of the strains with complete
access to quantitative data [26].
It is worthwhile to review some of the basic elements of the OFT approach for strain
measurement. The schematic of a typical experimental setup is shown in Figure 5. It
shows the laser and the outgoing small diameter (~1 ram) laser beam as it impinges
normally on the photographic negative containing the undeformed or deformed grating
image. The laser beam diffracts through the photographic negative, passes through a lens
which serves to fix the size of the diffraction pattern, which is then projected on a semi-
transparent frosted-glass screen. The diffraction peaks projected on the screen are imaged
with a video camera.
Photo-grid Screen
Laser
I- Collimator Lens
Figure 5 - Schematic of a typical Optical Fourier Transform experimental setup
Typically, as shown in Figure 6, the OFT or diffraction pattern projected on the

semi-opaque screen correspond to the various frequencies that define the 1D gating, and uo,
the distance between the diffraction peaks, is given by
,ID
u~ Po (6)
where, D is the distance between the screen and the grating, ~. is the laser light wavelength,
and Po is the grating pitch. This equation shows that the spacing of the diffraction orders is
inversely proportional to the pitch of the grid. This means that if a 1D grating is glued to an
object, and a photograph is taken of the grid on the object, if the resulting negative is
interrogated with a normally impinging laser beam, then the spacing of the grid lines can be
determined by measuring on the projection screen the separation of individual orders in a
diffraction pattern of the grid.
,.tD ,.t:
fourth uo p~ p~ Optical Axis fourth
iiIiii1.o,.
~k order order order order order order order
o..
Po
Figure 6 - 1D grating element and corresponding Optical Fourier Transform
If this is done also after deformation of the object, the distance u] between the
diffraction peaks is given by
2, D
u,- e, (7)
then the resulting normal strain is calculated by the equation,
- -
e -
- l
/Do
-
?do -
- -
7-/1
( 8 )
Po ul
It should be noted that the equation shown in Figure 6 for uo is a modified version of
Equation (6) where D, the distance from the photographic negative to the screen, has been
replaced by the focal length, f, of the imaging lens of the video camera.
This same principle as used with a 2D grating, or a cross-grating, is shown in

Figure 7 which is also the result of the normal impingement of a small diameter (--6mm)
laser beam on a transparent cross-grating, with its eventual projection on a semi-transparent
screen.
y'
G 9Fouri
Deformed
Undeformed
ar~ ,~
i . l v~ u, d
.~ o...........iw
Trans[orm c~/~//
Opti
a Axis
E x
Figure 7 - 2D grating element and corresponding Optical Fourier Transform
As before, the corresponding distances between optical diffraction peaks before and after
deformation are given by,
2f ,~f
u~ Po ul - 111 (9)
and,
2f 2f
v~ Qo Vl - Q1 (10)
from which the normal strains are found to be,
/2 0 -- U1 Vo -- "1~1
e~,~ - 8yy - - - (11)
U1 V1
and, if we let a = d~+ 0 and 13= d~- 0, since we know that
du dv
- 4+0 = a - 4-0 = fl (12)
dx
the mathematical shear swain and rigid body rotation can be shown to be,
~- 2 o~- 2 (13)
All of these quantities can be obtained in an actual experimental setup from an evaluation of
the centroid locations associated with the diffraction peaks shown in Figure 7.
Fast Fourier Transform Approach
The OFT approach led to the implementation by Morimoto et al. [26, 27, 28] of the
Fast Fourier Transform (FFT) approach to displacement and strain measurement. The basic
principle behind this approach is the same as that of OFT, except that the actual video
images of the grating glued to the specimen are analyzed by means of the FFT. To
illustrate the basic concepts let us use Figure 8 which shows a 1D sinusoidal grating before
and after a 10% deformation. Figure 9 illustrates the corresponding FFT's of these
sinusoids. If the distances between the zeroth and first order peaks are found, the resulting
strain may be determined by the application of Equation 8 above. Morimoto has examined
1D and 2D specimens by means of the FFT approach. The main advantage of this type of
approach is that the same analysis as was performed using OFT is done, except that now the
experimental data is in the form of video images of the actual gratings on the specimens.
This eliminates the step which generates a photographic negative which is to be interrogated
using the OFT approach. Current implementations of this approach are useful only for
plastic deformation of the order of 1% or more strain. The biggest advantage is that from
consideration of images of gratings on a specimen before and after deformation it is
possible to obtain full-field displacement and strain data automatically. The major
disadvantage currently is that the resolution of the technique needs to be improved if this
experimental technique is to have a major impact in mechanical measurements.
Figure 8 - 1D grating before and after 10% deformation
Figure 9 - Fast Fourier Transform representations of 1D grating

before and after 10% deformation
FUTURE DIRECTIONS
There is mother problem of the FFT approach to deformation measurement and it

occurs when the deformation is not readily determined by the calculation of the first and
higher harmonics because the local deformation behavior of a given object might not
coincide with such an approach. It should be remembered that a FFT approach has an
averaging effect on the data, and it tends to ignore local effects. For this reason wavelet
analysis becomes pertinent, as it involves a fundamentally different approach. As implied
above, a pixel row may be considered as the signal to be analyzed. In wavelet analysis,
instead of seeking to break down a signal into harmonic functions, the signal may be
broken down into a series of orthogonal basis functions of finite length called wavelets
[30]. Each wavelet is located at a different position on the position axis. At the finest
scale, wavelets may be made very short; at a coarse scale they may be made very long.
The important point is that any particular local feature of a signal can be identified from
the scale and the position of the wavelets into which it is decomposed. The wavelet
analysis approach is capable of improving the application of the FFT as Morimoto et al.
have shown [31]. There is another wavelet based approach that can also be applied which
requires being able to precisely assess the location of highly discontinuous markers on a
specimen [32].
Looking at the future, the major impact of applying wavelets for the shape and
deformation measurement of 3D bodies is that higher resolution levels of strain
measurement may be achievable. Deformation measurements could conceivably be made
without the benefit of applying a grating on a material, i,e., utilizing only the naturally
occurring texture on the material surface. Because the applications of this approach can be
extended to the microscopic and nanoscopic scale, this is sure to have a lasting impact in the
experimental solid mechanics field, and is predicted to be a growth area in the future.
SUMMARY AND CONCLUSIONS
The primary objective of this paper has been to review the contributions of selected
researchers toward making the grid method an experimental technique which can be applied
in a practical way and be fully automated. It has emphasized mostly optical, non-contact,
non-invasive measurement approaches which are sure to continue developing as video and
micro-computing technology continue to become more commonplace in the laboratory and
society in general. This paper has also emphasized the fact that a grid has to be viewed in
some generic way as one or more lines, dots or other shapes, assembled in regular or
irregular patterns. The objective when searching for new measurement approaches relying
on the grid method is to eventually be able to assess the deformation on a specimen which
requires no previous surface preparation, to do it rapidly and over the full area of interest, to
do it on a macro- and/or micro-scale, and to have access to qualitative and quantitative data
as needed. It appears that wavelet based analysis procedures may provide answers in the
near future which may help guide us in this direction.
REFERENCES
[1 ] Durelli, A. J. and Parks, V. L, Moir6 Analysis of Strain, Prentice-Hall, Englewood

Cliffs, NJ (1970).
[2] Post, D., Han, B., and Ifju, P., High Sensitivity Moir6, Springer-Verlag, New York,
NY (1994).
[3] Vest, C. M., Holographic Interferometry, John Wiley & Sons, New York, NY (1979).
[4] Ransom, W. F., Sutton, M. A., and Peters, W. H., "Holographic and Laser Speckle
Interferometry," Chap. 8, Handbook on Experimental Mechanic s, Kobayashi, A. S.,
editor, Prentice-Hall, Englewood Cliffs, NJ (1987).
[5] Durelli, A. J., Phillips, E. A., and Tsao, C. H., Introduction to the Theoretical and
Experimental Analysis of Stress and Strain, chap. 2, McGraw-Hill, New York, NY
(1958).
[6] Dally, J. W. and Riley, W. F., Experimental Stress Analysis, 3rd Edition, chap. 2,
McGraw-Hill, New York, NY (1991).
[7] Parks, V. J., "Strain Measurement Using Grids," Optical Engineering, Vol. 21, No. 4,
pp. 633-639 (1982).
[8] Peters, W. H., and Ranson, W. F., "Digital Imaging Techniques in Experimental
Stress Analysis," Optical Engineering, Vol. 21, No. 3, pp. 427-431, 1982.
[9] Chu, T. C., Ranson, W. F., Sutton, M. A., and Peters, W. H., "Applications of Digital
Image Correlation Techniques to Mechanics," Experimental Mechanics, Vol. 25, No.
3, pp. 232-244, 1985.
[10] Sutton, M. A., Turner, J. L., Bruck, H. A., and Chao, T. A., "Full-field Representation
of Discretely Sampled Surface Deformation for Displacement and Strain Analysis,"
Experimental Mechanics, Vol. 31, No. 2, pp. 168-177, 1991.
[11 ] Bell, J. F., "10,000 Threads to the Inch! ," American Machinist, Vol. 100, No. 16, pp.
112-113 (1956).
[12] Bell, J. F., "Determination of Dynamic Plastic SWainthrough the Use of Diffraction
Gratings," Journal of Applied Physics, Vol. 27, pp. 1109-1113 (1956).
[13] Bell, J. F., "Normal Incidence in the Determination of Large Strain through the Use of
Diffraction Gratings," Proceedings of the 3rd US National Congress of Applied
Mechanics (ASME, Providence, Rhode Island, 1958) pp. 489-493.
[14] Bell, J. F., "Experimental Study of Dynamic Plasticity at Elevated Temperatures,"

Experimental Mechanics, Vol. 2, pp. 181-186 (1962).
[15] Bell, J. F., "Diffraction Grating Strain Gage", Proceedings of the Society for
Experimental Stress Analysis, Vol. 17, pp. 51-64 (1960).
[16] Douglas, R. A. "Strain-field Investigations with Plane Diffraction Gratings,"

Experimental Mechanics, Vol. 5, pp. 233-238 (1965).
[17] Pryor, T. R., and North, W. P., "The Diffractographic Strain Gage," Experimental
Mechanics, Vol. 11, pp. 565-568 (1971).
[181 Pryor, T. R., Hageniers, O. L., and North, W. P., "DisplacementMeasurement Along
a Line by the DiffractographicMethod," Experimental Mechanics, Vol. 12, pp. 384-
386 (1972).
[19] Boone, P. M., "A Method for Directly Determining Surface Strain Fields Using
Diffraction Gratings," Experimental Mechanics, Vol. 11, pp. 481-489 (1971).
[20] de Caluwe, M., Steurs, R., van Gilsen, W., and Boone, P., "Strain Evaluation by
Electro-Optical Processing of Specimen-GratingPhotographs," Electro-Optics/Laser
Proceedings International (United Kingdom, 1976) pp. 224-230.
[21] Sevenhuijsen, P. J., "The Development of a Laser Grating Method for the
Measurement of Strain Distributions in Plane, Opaque Surfaces,"presented at the 6th
International Conference on Experimental Stress Analysis, Munich, FRG, 1978
(unpublished).
[221 Sevenhuljsen, P. J., "Two Simple Methods For Deformation Demonstration And
Measurement," Strain, Vol. 17,No. 1, pp. 20-24 (1981).
[23] Sevenhuljsen, P. J., "Photonics For Deformations,"Proceedings of the V International

Congress for Experimental Mechanics (SESA, Montreal, Canada, 1984) pp. 787-794.
[24] Cardenas-Garcia, J. F., Read, D. T., and Moulder, J. C., "Experimental Study of Path
Independence of the J-integral in an Aluminum Tensile Panel," Experimental
Mechanics, Vol. 27, pp. 328-332 (1987).
[25] Read, D. T., Moulder, J. C., and Cardenas-Garcia,J. F., "Experimental Study of Path
Independence of the J-integral in an HSLA Steel Tensile Panel," Conference of the
European Group on Fracture (Delft, The Netherlands, 1986).
[26] Moulder, J. C., and Cardenas-Garcia, J. F., "Two-Dimensional Strain Analysis

Using a Video Optical Diffractometer," Experimental Techniques, Vol. 17, No. 5,
11-16 (1993).
[27] Morimoto, Y., Seguchi, Y., and Higashi, T., "Moird Analysis of Strain by Fourier
Transform," JSME International Journal, Series I, Vol. 32, No. 4, pp. 540-546
(1989).
[28] Morimoto, Y., Seguchi, Y., and Higashi, T., "Two-dimensional Moird Method and
Grid Method Using Fourier Transform," Experimental Mechanics, Vol. 29, No. 4,
pp. 399-404 (1989).
[29] Morimoto, Y., Seguchi, Y., and Daifuku, K., "Fourier-transform Moir6 Method
with Consideration of Misalignment," Experimental Mechanics, Vol. 34, No. 4, pp.
349-356 (1994).
[30] Newland, D. E., Random Vibrations, Spectral And Wavelet Analysis, 3rd edition,
Longrnan, Harlow and John Wiley, New York (1993).
[31] Morimoto, Y., and lmamoto, Y., "Application Of Wavelet Transform To

Displacement And Strain Measurement By Grid Method," Proceedings of the Society
for Experimental Mechanics Spring Meeting, Grand Rapids, Michigan, June 12-14,
1995, pp. 894-897.
[32] Ranson, W. F., and Vachon, R. I., "Wavelet Based Strain Measurement Applications
In Optical Stress Analysis," Proceedings of the Society for Experimental Mechanics
Spring Meeting, Grand Rapids, Michigan, June 12-14, 1995, pp. 894-897.
Fu-Pen Chiang 1, Qing Wang2, Fred Lehman3
NEW DEVELOPMENTS IN FULL FIELD STRAIN MEASUREMENTS

USING SPECKLES
REFERENCE: Chiang, F.-P., Wang, Q., and Lehman, F., "New Developments in
Full Field Strain Measurements Using Speckles," Nontraditional Methods of
Sensing Stress, Strain, and Damage in Materials and Structures, ASTM STP 1318,
George F. Lucas and David A. Stubbs, Eds., American Society for Testing and
Materials, 1997.
ABSTRACT: Speckle photography is a versatile full field technique for measuring

displacement and strain. However, the wet photographic process and tedium of fringe
analysis had made it impractical for application to industrial problems. This paper
describes a recent development of combining digital recording and image analysis that
has rendered the technique a practical tool. In addition, through the development of
SIEM (Speckle Interferometry with Electron Microscopy) the technique has been
extended into micromechanics domain whereby strain fields within a region only a few
micron square can be mapped with confidence. Several examples of this application are
presented.
KEYWORDS: speckle, interferometry, digital image processing, electron microscope,

micro-mechanics, experimental mechanics, stress/strain analysis,
The science of measurement may be defined as "the quantitative comparison

between a given parameter and an established standard." In the case of full field
deformation measurement, the standard is usually a certain regular geometric pattern. For
example, regular lines or grids are used as a reference in grating technique and various
1 Leading Professor and Chair, Department of Mechanical Engineering, State University

of New York at Stony Brook, Stony Brook, NY 11794-2300
2 Research Assistant, Department of Mechanical Engineering, State University of New

York at Stony Brook, Stony Brook, NY 11794-2300
3 Principle Engineer, Instron Corporation, 100 Roy~ll Street, Canton, MA 02021
156
Copyright9
(PDVSA by ASTM
CHIANG ET AL. ON MEASUREMENTSUSING SPECKLES 157
moir6 methods. Any change from this standard is a direct measure of the deformation.
However, the measurement standard does not have to be regular in principle. It can also
be an irregular geometric pattern, for instance, random speckles. Speckle photography is
just such a technique which utilizes a random speckle pattern as the gaging device to map
the full displacement/strain field. Mapping is done by a correlation calculation
(performed either optically or numerically) of a cluster of speckles to delineate its
collective movement rather than following the movement of individual speckles.
Compared with moir6 methods, it is much easier to create a speckle pattern than to print
gratings on an object surface. The surface can be flat as well as curved. In general the
speckle pattern will not produce any strengthening effect like what a moir6 grating layer
sometimes does, which could be very crucial in many applications. Taking advantage of
the unique feature of speckle photography, a new experimental micro-mechanics
technique, Speckle Interferometry with Electron Microscopy (SIEM), is developed. Its
background, principle and some applications are briefly introduced below.
OPTICAL SPECKLE METHOD
The speckle phenomenon has been known since the time of Newton. But the
maturity of understanding and application of this phenomenon were realized after the
invention of laser. When an optically rough surface is illuminated by coherent radiation,
the reflected wavelets scatter in all directions. Being coherent among themselves they
combine to form a complicated random interference pattern called laser speckle. A
typical laser speckle pattern is shown in Fig. 1. Since speckle pattern is caused by the
Fig. I A laser speckle pattern Fig.2 A white light speckle pattern
random interference of light scattered from various depths on the object surface, it acts as
a random "grid" naturally printed on the object surface. By correlating the speckle
patterns before and after deformation, surface displacement/strain is determined. The

correlation is done by either the point-wise probing of a laser beam or whole field Fourier
filtering. The basic principle and numerous applications can be found in the book by
Erf ~] and the review papers by Chiang [2-3].
While it is highly sensitive, non-contact, nondestructive and remote sensing, laser
speckle metrology suffers from pattern de-correlation due to factors such as out-of-plane
tilt, excessive strain, etc. Later it was realized that any random intensity distribution can
be considered as a sp.eckle pattern. This means that any random pattern can be utilized
for metrological measurements so long as the speckles are distinct and of sufficient
contrast. When either naturally present or artificially created random patterns are
observed or recorded using white (or incoherent) light, they are referred to as white light
speckle. Fig.2 shows a typical white light speckle pattern. Once recorded, the white light
speckle photographs can be processed with a laser to observe either the Young's fringes
or the whole field deformation fringes. A complete theoretical analysis and various
applications of white light speckle technique for strain analysis is given in the paper by
Asundi and Chiang [4]. Since laser speckle and white light speckle are either generated
or analyzed by an optical approach, they are called optical speckle methods.
SPECKLE METHOD WITH ELECTRON MICROSCOPE
The sensitivity of optical speckle methods is determined by the size of the

speckles. And the smallest speckle that can be created using optics has a dimension of
one half of the wavelength of the incident radiation. The laws of physics dictate that the
smallest observable object cannot be smaller than the wavelength of the radiation that is
used to form its image. Thus the fundamental limiting factor of any optical speckle
method is the wavelength of the radiation that is used to record the speckles. When a He-
Ne laser is used, the theoretical limit of spatial resolution is approximately 0.31.tm (1/2 of
the average wavelength of the laser radiation). But the practical limit is about 10 times
larger, i.e. 3Bm, which is essentially the lower bound of the optical speckle method.
One of the choices to break this fundamental barrier is to use an electron
microscope. Chiang [5] first introduced the concept of SIEM and demonstrated its
feasibility in 1982. Particles with sizes only a small fraction of a micron can be created
by either a vacuum or physical/chemical vapor deposition processes, or other surface
treatment techniques. While they are not visible through an optical microscope they are
easily observable under a scanning electron microscope (SEM) or transmission electron
microscope (TEM). Fig.3(a) shows such a pattern observed by TEM, which was recorded
on photographic film. Fig.3(b) shows a typical Young's fringe pattern from laser beam
probing. Obviously the sensitivity is increased by several orders of magnitude.
Comparing SEM and TEM, the former is more suitable for performing
mechanical tests. It does not require special sample preparation. Any samples, as long as
they can be put into the vacuum chamber of a SEM, can be easily observed at high
magnification. SEM has a very high resolution as well as large field depth (300 times
better than an optical microscope). However, since a SEM image is formed by scanning
CHIANG ET AL. ON MEASUREMENTS USING SPECKLES 159
the electron beam over a specimen surface in a given number of lines, the resolution of its
photograph is not comparable to an optical image. Speckles are usually sparse in order to
be clearly recorded. Therefore, optical filtering of SEM photographs will yield fringes of
low contrast and high random noise. To extract the deformation field from such speckle
patterns, a digital image processing technique was developed [6-7].
Fig.3 (a) Speckles formed by vacuum depositing gold particles on graphite as recorded
by transmission electron microscope; (b) Young's fringes representing
displacement of the order of 0.001 nm.
COMPUTER AIDED SPECKLE INTERFEROMETRY (CASI)
The comparison between two digital images can be performed in many different
ways. One of the straight forward methods is to construct a correlation function and
define a correlation coefficient, for example [8-9],
H [h, (x, y) - h2(x,y)]2 dxdy

least square correlation
CL _ a ~ h2(x'y)dxdy
II h, (x,y)h~ (~,y)d.xdy
and cross correlation Cc _ .4
~ h~(x'y)dxdy~ h2z
where
h~ and h2 = speckle patterns before and after deformation, respectively,
A = area of the speckle pattern of interest.
Since I~ (x, y) = h I [x - u(x, y), y - v(x,y)]

and
r Ou r Ou
u(x,y) = u0 + JVxxaX+ j N a y
ray rav
v(x,y) = Vo+ / ~ d y + | ~ a x ,
CL and Cc are functions ofu(x,y) and v(x,y). When the minimum value of CL or the
maximum value of Cc is reached by changing u(x,y) and v(x,y), both displacement and
strain are determined. The deficiency of this method is the huge amount of computation
involved in the numerical operation.
Another method is CASI, which evolved through several stages from the
understanding of the optical processing of speckle photography. By applying a Fourier
transform to both hI and h2, one has
H l ( f x , f y ) = 3{ht(x,y)}
H 2 ( f x , f y ) = ~{h2(x,y)}.
Then an interference between two speckle patterns is performed on the spectral domain:
H'(L'f~)H~(L'L) = expO'[(~,(L,f~)-O2(L,f~)]}
F(s = H~(L,f~)H2(L,f~)
where * , ( L . L ) and *2(fx,J~) are the phases of H , ( L , L ) and H z ( f ~ , f y ) ,

respectively. It is shown that
O,(L.f~)-~2(L.L ) = 2rc(uL + . f y).
By an additional Fourier transform, one obtains the halo function
G(r1,~) = 3{F(f~,fy)} = G(rl - u,~ - v),
which is an expanded impulse function located at (u, v). By detecting the crest of this
impulse, displacement between two speckle pattern is uniquely determined. Numerically
the Fourier transform is done by a FFT (fast-Fourier transform) algorithm. The
computational efficiency is greatly improved when compared with conventional
correlation methods.
CHIANG ET AL. ON MEASUREMENTSUSING SPECKLES 161
APPLICATON OF CASI I[!Q]
An example of the application is shown in Fig.4. Fig.4(a) shows a rubber shock

mount (Lord multiplane shock mount part no.156p-6), which is prepared by spray
painting it lightly with white paint to produce a speckle pattern. The mount is subjected
to loading along its central axis in the upward direction. White light is used as
illumination to capture the images before and after deformation. Fig.4 (b) and (c) show
the results of the analysis, where the displacement vectors are superimposed on the image
of the shock mount to give a quantitative presentation of the nature of the deformation
field. The top view of the mount displays a symmetric radial movement, i.e., the
displacement field radiates out from its center. From the side view it is clearly revealed
that the upper region of the mount is "bulging" outward and the lower region is shrinking
inward. It also shows the geometric effect of projecting the displacement field from a
cylindrical surface onto a plane.
Fig.4 The test on a rubber shock mount.
APPLICATIONS OF SIEM
Combining speckle photography, electron microscopy and digital image

processing makes SIEM a practical tool for micro-measurement. The following
applications demonstrate some aspects of the capability ofSIEM. A typical experimental
procedure consists of speckle deposition, which is currently achieved through vacuum
evaporation, in situ mechanical test inside SEM, image digitization, digital image
processing by CASI and data analysis.
Determination of interphase mechanical properties of composite [11]
One of the major difficulties in modeling the behavior of composite materials is

the lack of experimental data on the mechanical properties of a matrix/fiber interphase
which is often the result of the manufacturing processe (such as fiber coating and
diffusion bonding). The interphase material cannot be created in bulk form for characteri-
zation or, even if it could, the mechanical properties in bulk are not necessarily the same
as they are in the form of thin films. Thus, it is absolutely essential that the interphase
mechanical properties be characterized directly. However, an interphase layer is often a
few microns in thickness, which renders the problem inaccessible to conventional strain
analysis technique. On the other hand, it is an ideal problem for SIEM.
A thin slice (about 2 mm thick) of SiC/Ti metal matrix composite is loaded in
transverse tension inside a SEM. A unit cell containing a sandwich of fiber-interphase-
matrix with/without coated speckles is shown in Fig.5. Displacement fields determined
by SIEM at different stages during loading are shown in Fig.6. The more detailed results
are achieved at higher magnification, as shown in Fig.7 and Fig.8. A finite element
program is used to model the deformation field and to back out the mechanical properties
of the interphase. As a first approximation we assume the inter-phase is isotropic and
homogenous. After several iterations, the local deformation field obtained experimentally
matches that of the numerical simulation. As a result it is determined that the Young's
modulus of the interphase must lie within the range depicted in Fig.9.
Fig.5 A unit cell viewed at 1,000 magnification without and with speckle pattern.
Fig.6 Horizontal displacement fields at different stages during the loading process.
Fig.7 Interphase zone viewed at 4,000 magnification without and with speckle pattern.
Fig.8 Horizontal displacement fields at different stages during loading process.
164 SENSINGSTRESS, STRAIN, AND DAMAGEIN MATERIALS
~"
"so .............................................................................
i i
~...................4............ ~
i ...... ,-r...........
i i i ,, i
i ...........................................................
~....................
~.............4...............~.....................
i..........
-t i i
................... ; ............. ~ .................................. ~- ................. t ........... ~ .................. ~ ...............
z L ! I
', i ', r E '.
o i i i i i i i
0.08 0.1 0.15 0.2 0.25 0.8 0.218 0.4 0.48 0.5
Poisson's Ratio v
Fig.9 Range of effective Young's Modulus ofinterphase material.
Mechanical behavior of a single fiber 1[.~]
Fibers are the load bearing component in a continuous fiber-reinforced polymeric

composite. Its strength is a dominant factor in determining the composite's ultimate
strength. Usually, the diameter of a single fiber is around 10 microns. To measure the
mechanical response along all material axes at micron scale is not a easy task. To the best
of our knowledge the full-field of deformation in transverse direction of a fiber has not
been determined via a direct measurement. Fig. 10 shows a speckle pattern observed by
SEM on a graphite fiber with 51am in diameter. Our tests on Aramids fiber show some
interesting phenomena.
Nearly perfect uniaxial orientation of the polymer chain gives arimads fiber
highly anisotropic mechanical behavior. It has a high tensile strength but weak lateral
Fig. 10 A graphite fiber coated with speckles of diameter of about 30 nm.
bond and low compressive strength. In order to evaluate the Poisson's effect, a single
KEVLAR fiber of diameter 12 ~tm (Du Point, KM2) is loaded in longitudinal tension
inside a SEM. With speckles coated on its surface, displacement fields in both
longitudinal and transverse directions are determined, as shown in Fig. 11. With load
increasing, the magnitude of both longitudinal and transverse strains increases. Unlike
most solids whose Poisson's ratio is smaller than or equal to 0.5, here the ratio of
transverse shrinkage to longitudinal elongation is much larger than 0.5 and not a
constant. Surprisingly, transverse displacement field also shows a wavy pattern, which
has not been noticed before. This is attributed to the effect of fiber shape change as a
result of lateral separation of fibrils within the fiber. In most theoretical analysis, the fiber
is treated as an isotropic material. The result of this test clearly shows that the effect of
anisotropy can not be ignored.
20 240
0 [ 22o
-20 200
~ -40
~ ~o ~ 160
=.,~,=~,===,,~'~=="==== o. 140
~ -1110 ~5
| 120
~ -120
-140
~ 8o
~-~ -160
I.-- 60
4 6 8 110 112 1'4 16 ;! 4 6 8 10 12 14 16
Longitudinalaxis (urn) Transverse axis (urn)
Fig. 11 Longitudinal and transverse displacements of a KEVLAR KM2 fiber.
Near field phenomenon at bimaterial interfacial crack tip [13-14]
One of the long standing topics in mechanics research is to gain a fundamental

understanding of what makes material tougher. While macroscopic properties of
structural material such as fracture toughness and crack growth resistance are used for
engineering analysis and design, the actual fracture process takes place at the atomic
lattice level. In order to predict the fracture behavior in a more precise manner, the
influence of fundamental properties at the atomic scale on the macroscopic properties has
to be well understood.
Driven by this motivation, a study on the stress singularity at the crack tip is
carried out by measuring the displacement fields around an interfacial crack tip. The
strain distribution along the cohesive zone is also evaluated, by which the constitutive
relations could be examined. The specimens are made of two different epoxy resins. One
has Young's modulus E=2.38 GPa, Poisson's ratio v=0.39 and another E=0.66 GPa,
v=0.44. All the cracks are "true" interfacial cracks, which are produced by propagating
interface cracks. Results are shown in Fig.12(a). It can be seen that a weak (less than
square root) singularity exists within the K-dominant zone. For the mode I dominant
1-5 0.5
di*i/
t.4 0
Experimental data 0.4 ~'penmerttal data I
1.3
0.3
lJ
02
I.!
0.I
L.O
0.0
O.tl /"/ 1.
o~ 0.8
-0.!
-02
0.7
O.S
.,~.4
0.5
-O.5
0.4
--.0.I o / / " (b)
0.3
.,,0,? ~" i i t
Q.2 I , I , I , I
1.0 1.5 2.0 2.S 3.0 3.S 0.0 o.s LO L5 2+0 ?,+S
t,o+[ r ( u m ) ] Loi [ r (/~m) ]
Fig. 12 Plot of log normal crack opening displacement against log of distance from the
interfacial crack tip
case, the slope near the crack tip is obviously larger than 0.5, which means that the stress
singularity index is less than 0.5. This observation is further confirmed by examining a
smaller region around the crack tip, as shown in Fig. 12(b). Comparing this result with the
Williams asymptotic solution for an interfacial crack between dissimilar isotropic
bimaterials, it is concluded that the K-dominant annulus with the predicted singularity
exists near the crack tip. However, the predicted oscillation of stress or displacement
does not occur. It is because there is another zone (with weak singularity) within the K-
zone and the Williams solution no longer exists in that region very near the tip.
Effect of adhesive thickness on adhesion 1[1[~]
Many efforts have been devoted to the study of joints due to the extensive use of
adhesive-bonded joints in modem structures. An unsolved debate in the science of
adhesive joints is whether or not the adhesive behaves in bonded states the same way as it
does in bulk form. In addition, the effect of thickness of the adhesive layer on its
adhesion is a main issue of the debate.
In this work a series of butt joints with different thickness ranging from 617 Ixm
to 38 ~tm are tested. The material of the adherend is PSM-5 and the adhesive is PC-6C.
Both adhesive and adherend are epoxy but PC-6C is more ductile than PSM-5. The
deformation field within and around the butt joint is determined by SIEM. The results
from two typical specimens are shown in Fig. 13. It is found that there exist three different
regions at the joint, where the gradient of the displacement is different. The regions are in
the adherend, in the center of the adhesive, and in the interfaces between the adhesive
and adherend. The displacement gradient in the center region decreases with a decrease
in the thickness of the adhesive layer. The apparent stiffness of the adhesive in the center
CHIANG ET AL, ON MEASUREMENTS USING SPECKLES 167
region is severely affected by the layer thickness. The strains obtained by experiments are
larger than the theoretical prediction. When the adhesive layer is thinner, the effective
strain (the displacement across the adhesive layer divided by the layer thickness) is
smaller, i.e., the joint appears tougher.
0.,8 8 - -, .... + .... n "'Vl .... I ....
O.St .d.- i .-~

~1 . ]~ i. ~176 ' ~- H
'
I~ ;9
4,
0.! ~, i e!
]
i i o !
0.0 (a) i o'e~ LO.4XPa
! 04* ~ 1 7 6 1 7 6 1 7 6 1 7 6 1 7e6I - 1 . 4 4 X o (b) ,!! ooo~"~176,i'
-O.l
e.=O.53X i ~ i ce-t0.$MPa
~
-O.t 9 i -i 9i ~ i It" 1-47Z
gooo~ o
; Ad, h e s t v e ~ i Adhe=lv~ i eg-O.98Z
-0.4 .... i, , , , . . . . _, ,.. ";"" ........... i .........

-1.5 -1.0 -0.5 0.0 0.6 1.0 1.5 -1.w -1.0 -0.6 O.O 0.6 1,0 1J5
3,/t y/t
Fig.13 Displacement in loading direction of different joints across adhesive layers

(a) layer thickness t=38/am; (b) layer thickness t=-617/am
Deformation characteristic at kink band tip in fibrous composite 1[~.]
The compressive strength of fibrous composites is lower than their tensile

strength. Kinking is a primary compressive failure mode and has been intensively studied
during the past three decades. Major efforts were devoted to the prediction of the stress at
which kinking occurs -- incipient kinking stress. However, the whole kinking process,
especially the kinking kinematics, has not been well understood. Since the specimen fails
instantaneously and catastrophically under compressive loading, experimental
investigation on the kink process is very difficult to perform.
It is realized that a kink occurs at the scale of several fiber diameters and a
microscope has to be used. To control the location of kink initiation, a notch is usually
introduced. But the notch brings about a complex stress field which makes analytical
modeling more difficult. We developed an experiment scheme which allows us to control
kink initiation and propagation in a specimen without a notch. Various phenomena
during the kink growth were clearly observed and recorded. Fig.14 shows a kinking band
during steady-state propagation. The displacement field near kink band tip is shown in
Fig.15. Shear deformation dominates the field within the kink band and at the band tip.
But it is still a mix-mode field rather than a pure mode II.
CONCLUSIONS
The principle of a recently established micromechanics technique, SIEM, is

described. Its capability of mapping the full field deformation within a small region at
micron scale is demonstrated by several applications. The image processing is performed
by CASI, which is proved to be fast and accurate.
Fig.14 A kink band during propagation Fig.15 Contour map of horizontal displace-
ment field around kink band tip
ACKNOWLEDGEMENTS
The supports provided by AFOSR under the Grant No. F496209310219 and ONR
under the Grant No. N000149151380 are greatly acknowledged.
REFERENCE
[1] Eft, R. K., Speckle Metrologv, Academic Press, 1978.
[2] Chiang, F. P., "A New Family of 2D and 3D Experimental Stress Analysis
Techniques Using Laser Speckles," Solid Mechanics Archives, 3(1), ppl-32, 1978
[3] Chiang, F. P., "Speckle Metrology," Metals Handbook, 9th edition, Vol. 17, pp 432-
437.
[4] Asundi, A. and Chiang, F. P., "Theory and Applications of the White Light
Speckle Method for Strain Analysis," Optical Engineering, 24(4), pp 570-580,
1982.
[5] Chiang, F.P., "Speckle Method with Electron Microscope," Society for Experi-
mental Stress Analysis, Spring Meeting, HI, May 23-28, 1982.
[6] Chen, D. J. and Chiang, F. P., "Computer-Aided Speckle Interferometry Using

Spectral Amplitude Fringes," Applied Optics, 32(2), pp 225-236, 1993
[71 Chen, D.J., Chiang, F.P., Tan, Y. S. and Don, H.S., "Digital Speckle-
Displacement Measurement Using a Complex Spectrum Method," Applied Ootics,
32(11), pp 1839-1849, 1993.
[8] Rosenfeld, A. and Kak, A. C., Digital Picture Processing, Second edition,
Academic press, 1982.
[9] Peter, W. H., Ranson, W. F., "Digital Imaging Techniques in Experimental Stress
Analysis," Optical Engineering, 21(3), pp 427-431, 1982.
[10] Lehman, F., "Local Deformation Field Measurement Using Digital Speckle
Photography," Proceedings of th.e Conference of Local Strain and Temperature
Measurements in Non-Uniform.~ie!ds at Elevated Temperature, Berlin, Germany,
March 14-15, 1996.
[11] Wang, Q. and Chiang, F.P., "Experimental Characterization of Interphase

Mechanical Properties of Composite," Composite Engineering, in press, 1996.
[12] Wang, Q. and Chiang, F. P., "Experimental Study on the Mechanical Behavior of a
Single Fiber," Material Research Society, Fall Meeting, Boston, MA, November
27-December 2, 1994.
[13] Yan, X. T., Wang, Q. and Chiang, F. P., "On Stress Singularity at the Interfacial
Crack Tip - An Experimental Study," Proceedings of the American Society for
Composite, 9th Technical Conference, Newark, DE, September 20-22, 1994,
pp1036-1041.
[14] Wang, Q. and Chiang, F. P., "Micro-mechanical Behavior of Polymer Interface

Reinforced with Copolymers," Proceedings of Material Research Society, Fall
Meeting, SymposiumQ, Boston, MA, November 27-December 1, 1995, in press.
[15] Wang, Y. Y., Wang, Q. and Chiang, F. P., "Study of Effect of Thickness on
Mechanical Properties of Adhesive Material by SIEM," Proceedings of the 18th
Annual Meeting of Adhesive Society, Hilton Head, SC, February 19-22, 1995, pp
325-327.
[161 Wang, Q. and Chiang, F.P., "Micro-buckling Behavior of Graphite/PEEK

Composite," Society of Engineering Science, 32"d Annual Technical Meeting,
New Orleans, LA, October 29-November 2, 1995.
Ultrasonic and Infrared Techniques
Dennis J. Buchanan 1, Reji John2, David A. Stubbs 3, Dianne M. Benson4, and Prasanna
Karpur2
ULTRASONIC LONGITUDINAL AND SURFACE WAVE METHODS FOR I N

S I T U MONITORING OF DAMAGE IN METAL MATRIX AND CERAMIC
MATRIX COMPOSITES
REFERENCE: Buchanan, D.J., John, R., Stubbs, D.A., Benson, D.M., and Karpur, P.,
"Ultrasonic Longitudinal and Surface Wave Methods for In Situ Monitoring of
Damage in Metal Matrix and Ceramic Matrix Composites," Symposium on Nontra-
ditional Methods of Sensin~ Stress, Strain, and Damage in Materials and Structures,
ASTM STP 1318, George F. Lucas and David A. Stubbs, Eds., American Society for
Testing and Materials, 1997.
ABSTRACT: This paper discusses the development of a unique set of integrated

Nondestructive Evaluation (NDE)/Mechanical test techniques for characterization of
ceramic matrix composites (CMC) and metal matrix composites (MMC). These
techniques are sensitive to the initiation and accumulation of damage in CMC and MMC.
These techniques use longitudinal (bulk) wave and surface wave ultrasonic energy to
interrogate the composite specimens, in situ, during mechanical testing. Changes in
surface and longitudinal wave characteristics are compared with stiffness changes for
extensometer measurements during tension, creep and fatigue testing of SiC/BMAS (bar-
ium-magnesium alumino-silicate) CMC. Surface waves were used to identify the onset of
surface and near-surface cracks in CMC. Changes in the surface wave characteristics
provide a better indicator of impending failure during fatigue testing of Sigmafri-6242
MMC than extensometer data. Longitudinal wave propagation was successfully used to
monitor the changes in the stiffness of the composite during fatigue loading conditions of
CMC and MMC. The changes in the peak-to-peak amplitude of the transmitted wave
signal appears to be an equal or better indicator of impending failure of the composite
than the extensometer data.
KEYWORDS: ceramic matrix composite, metal matrix composite, damage progression,

in situ, longitudinal wave, surface wave, mechanical testing, nondestructive evaluation,
ultrasonic, CMC, MMC
1Assistant Research Engineer, 2Research Engineer, 3Associate Research Engineer, and

4Graduate Research Assistant, Structural Integrity Division, University of Dayton
Research Institute, 300 College Park, Dayton, OH 45469-0128, USA.
173
Copyright9
(PDVSA by ASTM
INTRODUCTION
The development of new CMC and MMC material systems for use in advanced
aircraft engines and aerospace vehicles require light-weight, high-temperature, and high-
stiffness properties [1-3]. Traditional methods of measuring material properties with
extensometers, strain gages and optical techniques are not very sensitive to the initiation
of damage as seen through hairline surface cracks, internal matrix cracks, fiber debonding
and matrix delamination. Characterization of these advanced material systems requires
new nondestructive techniques that are sensitive to the initiation and progression of these
damage mechanisms.
Many investigators have used a variety of NDE techniques to detect and monitor
damage in monolithic and composite materials under tensile and fatigue loading
conditions [4-13]. Anson, Chivers and Put-tick [12] analytically investigated the detection
of pre-crack fatigue damage from velocity measurements in MMC using both bulk wave
and surface wave techniques. Kautz and Bhatt [5] showed good agreement between
calculated ultrasonic modulus and mechanical modulus using an acousto-ultrasonic
technique for CMC. Achenbach et al. [6] concluded that velocity and attenuation
measurements show variations with the number of fatigue cycles and suggest that this
provides data on near-surface fatigue damage in a aluminum alloy matrix with silicon-
carbide whiskers. MacLellan, Stubbs and Karpur [8] showed that the surface wave
technique was a better indicator of matrix cracking and specimen remaining life than the
extensometry modulus data during fatigue tests on MMC. Wooh and Daniel [13]
concluded that attenuation and wavespeed for longitudinal waves are sensitive to matrix
cracking and insensitive to fiber-matrix debonding in CMC during monotonic tensile
tests. Most of the previous investigations [4-13] have used either surface wave or
longitudinal wave techniques. Several authors [5-8,12-13] have combined traditional
mechanical and NDE techniques for tensile and fatigue loading conditions where the
ultrasonic energy is transmitted through the surface or through the thickness of the
specimen. Recently the University of Dayton Research Institute and Wright Laboratory
Materials Directorate (WL/MLLN) have integrated an existing mechanical test system
[14], with commercial longitudinal and surface wave transducers to acquire NDE data in
situ [15,16,17]. These studies have used surface waves and longitudinal waves that are
transmitted down the loading axis of the specimen with extensometer measurements to
provide a unique damage characterization system.
This paper discusses the development of an integrated NDE/mechanical testing

system with in situ damage characterization using longitudinal and surface wave NDE
integrated with conventional extensometry.
EXPERIMENTAL PROCEDURE
The MMC evaluated during this study was a Ti-6AI-2Sn-4Zr-2Mo (at%) matrix
reinforced with Sigma SM-1240 silicon carbide fibers, designated as Sigma, in a [0]6
BUCHANAN ET AL. ON ULTRASONIC LONGITUDINAL METHOD 175
layup. The microstructural analysis revealed a composite volume fraction of 0.25 [17].
The CMC consisted of barium-magnesium aluminosilicate (BMAS) matrix reinforced
with silicon carbide fibers with a [0/9013s layup. The microstmctural details of the
SiC/BMAS composite can be found in Ref. [18]. The volume fraction of fibers in the
CMC was 0.35.
The schematic of the mechanical test set up is shown in Fig. 1 [16]. The
monolithic friction grips were precisely machined to provide accurate alignment needed
for testing materials with small strains to failure [19]. For this integrated system, the
friction grips were redesigned to provide space for the longitudinal wave transducers.
The longitudinal wave transducers were passed through a hole in the grip, placed in
contact with the specimen end and held in place using a spring-loaded spacer block. The
surface wave transducers were attached to the top surface of the specimen with specially
designed spring-loaded clamps. Surface wave mode conversion wedges were used to
transform longitudinal waves into surface waves on the specimen. The depth of
penetration for the surface waves with a 10 MHz transducer is approximately 0.45 mm
for the CMC and 0.40 mm for the MMC. Fatigue tests using steel specimens
demonstrated that these transducer support mechanisms provided stable ultrasonic
coupling for more than a million cycles.
The NDE data acquisition system consisted of two separate systems, one for the
longitudinal wave transducers and the other for the surface wave transducers. Both
systems were computer controlled with a 100 MHz 8-bit data acquisition board for the
longitudinal wave system and a 500 MHz 8-bit data acquisition board for the surface
wave system. The transducers were driven with a spike-type pulser-receiver. The
ultrasonic energy was transmitted between the transducers and the specimen with a water-
based couplant. The frequency for the surface wave transducers was 10 MHz and the
frequency of the three sets of longitudinal wave transducers were 1, 5 and 10 MHz.
Figure 1. Schematic showing test set up with ultrasonic longitudinal and surface wave
transducers with one surface wave grip removed for clarity [16].
The mechanical displacement of the specimen was measured using an elevated

temperature extensometer to correlate the changes in the ultrasonic NDE signals to the
mechanical response of the material. A photograph of the experimental setup with a
CMC specimen is shown in Fig. 2. The specimen length between the friction grips is
approximately 110 mm. The quartz rods for the extensometer were modified to measure
the displacement across the 50 mm gage length of the specimen. The grips for the surface
wave transducers were removed to provide an unobstructed view of the surface wave
transducer wedges.
Figure 2. Photograph of friction grips with extensometer, ultrasonic longitudinal and

surface wave transducers.
ANALYSIS OF ULTRASONIC WAVEFORMS
A schematic of a typical time-domain signal for a longitudinal wave is shown in

Fig. 3. The time of propagation of the wave as measured by the receiving transducer is
denoted as ts, and the maximum difference between a successive peak and valley is
denoted as the peak to peak (p-p) amplitude.
Damage accumulation in the specimen can be monitored by changes in the wave

propagation time, p-p amplitude and frequency content of the ultrasonic signals.
MacLellan, Stubbs and Karpur [8] showed that the surface wave p-p amplitude decreased
during fatigue testing of MMC. Wooh and Daniel [13] reported changes in longitudinal
wavespeed and attenuation of the ultrasonic wave through the specimen thickness during
in situ monotonic tensile testing of CMC. This study reports changes in propagation
time, p-p amplitude and frequency content in situ for monotonic tensile, fatigue and creep
testing of CMC and MMC.
ta
Time
Figure 3. Schematic showing definition of wave characteristics.
A schematic of a typical dogbone specimen is shown in Fig. 4 [ 16]. For analysis

purposes the time required for the longitudinal wave to propagate from one transducer to
the other is separated into several sections. Each transducer easing has a wear face layer
to protect the piezoelectric element. The time required for the signal to propagate through
the wear face was determined by measuring the propagation times of multiple echoes for
the longitudinal wave in a ultrasonic immersion tank. By comparing multiple signal
echoes it was possible to accurately measure the propagation time through the wear face
layer to.
Figure4. Lengths of different specimen sections and the corresponding wave

propagation time [16].
178 SENSINGSTRESS, STRAIN, AND DAMAGE 1N MATERIALS
After the completion of the test the specimen was cut between sections Ll and Lg.
The length Li and the time of propagation tl were measured for this section. The velocity
of the longitudinal wave vl was computed from the measured values of LI and h.
Assuming that the sections LI and L2 had similar wave propagation velocities, t2 was
calculated from L2 and v2. These time and length measurements provide the data to
calculate the velocity of longitudinal waves in the gage section vg as given by Eq. 1.
Knowing vg, the average density of the material (p) and Poisson's ratio (v), the effective
stiffness of the material was calculated using Eq. (2) for an infinitely wide, isotropic plate
[20].
Lg (1)
Vg- tg
v 0(i-.2) (2)
The thickness of the CMC and MMC specimen was 2.73 mm and 1.35 mm,
respectively. The longitudinal wavelength in the CMC and MMC was 6.2 mm and 6.7
mm, respectively. Consequently, the wavelength to thickness ratio for the CMC and
MMC was 2.3 and 5.0, respectively. An assumption in the derivation of Eq. (2) was that
the wavelength to thickness ratio be greater than five for the error in the measured
velocity to be less than 5% from the theoretical solution. In addition, the test specimen
had a finite width, which further invalidates the use of Eq. 2, but special solutions [21]
have been developed for rectangles of finite thickness and width. Since the wavelength to
thickness ratio was not achieved for the CMC and MMC specimens it is not possible to
accurately calculate the elastic modulus. The measured changes in the effective stiffness
(Eeff), expressed as a ratio with respect to the initial value, can be expected to be valid.
RESULTS
Tension Test on CMC
The stress versus strain data for the tension test on SiC/BMAS is shown in Fig. 5
[16]. The test was performed under stroke control and periodically put on hold to acquire
longitudinal and surface wave data. The elastic modulus (Ec) was 120 GPa. The
proportional limit, identified by the deviation of the stress-strain data from the initial
elastic response, was determined to be approximately 65 MPa.
250 9 I ' I ' I " I
Ec = 120 GPa
200
150
100 data acquisition points
~i~ 9 Proportional
50
Limit = 65 MPa
9 I 9 I 9 I 9 I
0 0.0015 0.003 0.0045 0.006 0.0075

Strain (m/m)
Figure 5. Monotonic tensile stress versus strain response of CMC.
The normalized p-p amplitudes for the longitudinal and surface wave was shown
as a function of stress in Fig. 6. Both signals show very little change with increasing
stress before reaching the proportional limit. The surface wave data show a rapid decline
in p-p amplitude near the proportional limit whereas the longitudinal wave data show a
slightly more gradual decrease [16].
9 " " ' I " i ' ' ' I ' ' ' " I " ' ' '
1.0
O
~~~~//Longitudinal Wave
2 0.8
< 0.6 -
) '', ~ Surface Wave -
' | I
Proportional
0.4 =9: Limit ~ i ," t-~l-ZlE].l.71E]"~
0.2 - I
ZO
0.0 9 9 9 9 I .'% 9 . . I . . 9 9 I 9 9 9 9
0 50 100 150 200

Stress (MPa)
Figure 6. Surface and longitudinal wave P-P amplitude versus stress for the monotonic
tension test of CMC.
The frequencies of the surface and longitudinal wave transducers were centered at
10 MHz and 5 MHz respectively, but due to the properties of the composite the received
ultrasonic signals were centered at 4.2 MHz and 0.83 MHz respectively. The decrease in
the p-p amplitude of the surface wave data can be attributed to the formation of matrix
cracks on the outer plies which attenuated the higher frequency (~-A.2 MHz) signals. The
longitudinal wave data, which consisted of lower frequency (~0.83 MHz) signals were
not as sensitive to the development of small cracks. The surface wave data captured the
initial damage initiation and the longitudinal wave data captured the overall deformation
response of the composite. Cyclic load-unload tensile data showed that the changes in
stiffness from the longitudinal signal correlate with stiffness calculated from cyclic data
[16]. The correlation between different forms of damage development and transmitted
dominant frequency is shown as a plot of dominant frequency of the ultrasonic signals
versus stress level for the tensile test on the CMC in Fig. 7. The dominant frequency of
the surface wave in Fig. 7 displays a sharp decrease near the proportional limit. In
contrast, the longitudinal data showed almost no change in frequency for the entire test.
The data supports the argument that the higher frequency signals in the surface wave data
were attenuated by the development of matrix cracks that blocked the transmission of
shorter wavelength signals.
5 9 9 9 9 | I l l 9 m J 9 9 9 9 | 9 9 9 9
'~" 4 ,, ,., v ~ ,~--,~ / Surface Wave
3
9" Longitudinal 'i
~" 2
9
' ~, c ~.~ ~
.3{3 E3 t~ ~ ~ ~-~-H~-t~+ ++ ~ ~ 3 I ~ E ] [ ] - ~
@
~ 0 . '~ / Proportional Limit
. . . . , , ~ . . . , , , 9 9 , 9 9. ~ .
0 50 100 150 200
Stress (MPa)
Figure 7. Effect of stress on dominant frequency of transmitted waves stress for the
monotonic tension test of CMC.
Fatigue Test on CMC.
The fatigue test was cycled at a frequency of 5 Hz, with a stress ratio of 0.05 and
maximum stress of 155 MPa at room temperature [16]. The test frequency was
periodically reduced to 1 Hz cycling to acquire stress versus strain data and then briefly
put on hold at maximum load to acquire the longitudinal wave data.
1.0
~D 0.9
r~ 0.8
~ 0.7
;~ 0.6
0.5
10~ 101 102 103
Cycle
Figure 8. Normalized modulus versus fatigue cycles on CMC.
The mechanical modulus was calculated from the initial unloading linear portion
of the stress-strain response. The normalized modulus for the two techniques show
similar trends for the majority of the fatigue life consistent with the observations during
the tension test. The ultrasonic data showed a 20% decrease in the stiffness during the
last 100 cycles while the extensometer showed only a small change [16].
Creep Test on CMC
A unique feature of the redesigned grips is the capability of performing elevated

temperature tests with the longitudinal wave transducers. Evaluation tests demonstrated
that the ultrasonic transducers remained less than 50~ with a 1315~ temperature on the
CMC specimen for an extended length of time. The creep tests typically started with a 10
minute ramp to reach the test temperature of 1093~ followed by a 20 minute
stabilization period with zero applied load [15]. The specimen was then mechanically
loaded at 10 MPa/s to the target load of 69 MPa, =10% above the measured proportional
limit for this material. The frequency of the longitudinal transducers was centered at 10
MHz, but due to the properties of the composite the frequency of the received ultrasonic
signal was centered at 1 MHz.
The longitudinal data showed no change in p-p amplitude during the heating of
the specimen to 1093~ (Fig. 9). There was a 4% drop in signal amplitude during the 20
minute stabilization period at zero load. During the seven seconds that the specimen was
ramped to full load there was a 25% drop in the p-p amplitude of the longitudinal
waveform, presumably due to the formation of matrix cracks. The sudden decrease in the
p-p amplitude of the ultrasonic signal coincides with a rapid increase of strain as shown
in the Fig. 9. The p-p amplitude started decreasing gradually with increasing strain rate
between 10,000 and 60,000 seconds at which point the signal amplitude dropped below
the electronic noise of the system [15]. This may be attributed to extensive matrix
cracking in the material. Additional interrupted tests with fractographic examination are
required to confirm these observations.
I I I I iiii I ii i I I IIII i I I I i IIII I I I I I Ilia 0.010

I
1.0 . . . . . . . . -O O 0 . Longitudinal ,~I

~D
2 i JWave J O
0.8 Ramp {[t~0- - "~ _ L 0.005
Temperature ', "9 _ / '\
0,6-\, -- \ g~
0.4[ ~ "1 /LAopap~yl Extensomet~y ~ Extensometry 0.000
@ 0.2
Z
!,/ ', ^ ^o
o.o V . .. -0.005
10 2 10 3 10 4 105 106
Time (s)
Figure 9. Total strain and longitudinal p-p amplitude versus time on CMC.
Fatigue Test on MMC
This section discusses the results of the in situ NDE/mechanical test on Sigma/Ti-
6242 under fatigue loading conditions [17]. Surface wave data were acquired in situ at
room temperature at a frequency of 0.01 Hz. The maximum stress was 805 MPa with a
stress ratio of 0.1. The specimen was also interrupted at three points in the life of the test
to perform surface wave C-scans in an immersion tank. Figure 10 shows the results of the
normalized p-p surface wave amplitude and mechanical modulus obtained from
extensometry versus the applied cycles. The surface wave and extensometer data do not
show any indication of damage below 2700 cycles. After 3000 cycles the surface wave
data show a sharp decline in the signal as the specimen approaches failure. In contrast,
the extensometer indicates only a 10% drop in the stiffness prior to failure. The p-p
amplitude of the surface wave data appears to be more sensitive to impending failure. C-
scans were acquired at cycle counts 1000, 1985, 2922 and at failure. The C-scan at 1000
cycles showed no indication of any damage analogous to the surface wave and
extensometry data. The C-scan at 1985 cycles revealed the onset of surface cracks. The
decrease in the surface wave signal occurs at 2754 cycles which occurs between the last
two C-scans. The loss in surface wave p-p amplitude could be related to the initiation and
development of cracks on the surface that led to failure in the specimen. The details of
this fatigue test and the C-scan figures can be found in Ref. [17].
i I I I I I II I I I I I I I I I
1.0
0.8 / I
!
Extensometry
o ;>
0.6
0.4
Surface Wave ~
N
0.2 Failed,
9 9 9 9 9 9 9 9 9 9 . . . . l I~1 I Illl
o 0"0102 10 2 10 3 10 4
Z
Cycle
Figure I0. Normalized p-p surface wave amplitude and mechanical modulus versus
fatigue cycles on a MMC.
CONCLUSIONS
An unique integrated NDE/mechanical testing system was developed to

characterize the damage progression in CMC and MMC during tension, creep and fatigue
tests. The results of these tests demonstrated that the NDE test system is sensitive to the
initiation and accumulation of damage with increasing stress and/or fatigue cycles.
Surface waves were used to identify the onset of surface and near-surface cracks in CMC
during monotonic tension tests. Longitudinal wave propagation data were successfully
used to monitor changes in the stiffness of the composite during tension and fatigue
loading conditions. The creep test on the CMC demonstrated that the NDE system can be
used successfully at temperatures up to 1093~ The interrupted fatigue test on the MMC
demonstrated a correlation between the development of surface cracks observed with C-
scan results and the decrease of the p-p amplitude of the surface wave with increasing
cycles. The changes in wave characteristic of longitudinal and surface wave data appear
to be an equal or better indicator of damage progression in the composite than the
extensometer data.
ACKNOWLEDGMENT
This research was conducted at Wright Laboratory (WL/MLLN), Materials

Directorate, Wright-Patterson Air Force Base, OH 45433-7817, USA. D.J. Buchanan, R.
John and D.A. Stubbs were supported under on-site contract number F33615-94-C-5200
(Project Monitor: Mr. J.R. Jira). P. Karpur and D.M. Benson were supported under on-
site contract numbers F33615-94-C-5213 and AFOSR F49620-93-1-0461, respectively.
The authors gratefully acknowledge the efforts Mr. A.F. Lackey, under on-site contract
number F33615-94-C-5200 and Ms. J.A. Herzog under contact number
AFOSR/AASERT Program F49620-95-1-0500 for their assistance in the experimental
test setup and data analysis.
REFERENCES
[1] Gabb, T.P., Gayda, J., and MacKay, R.A., "Isothermal and Nonisothermal Fatigue
Behavior of a Metal Matrix Composite," Journal of Composite Materials, Vol. 24,
June 1990, pp. 667-686.
[2] Johnson, W.S., "Fatigue of Continuous Fiber Reinforced Titanium Matrix

Composites," An Invited Lecture for the Engineering Foundation Conference, MCE
Publications, Santa Barbara, CA, 1991, pp. 357-377.
[3] Evans, A.G., and Marshall, D.B., "The Mechanical Behavior of Ceramic Matrix
Composite," Acta Metallurgica, Vol. 37, No. 10, 1989, pp. 2567-2583.
[4] Vary, A., Materials Analysis by Ultrasonics, Noyes Data Corporation, Park Ridge,
NJ, USA, 1987.
[5] Kautz, H.E. and Bhatt, R.T., "Ultrasonic Velocity Technique for Monitoring
Property Changes in Fiber-Reinforced Ceramic Matrix Composites," Ceramic
Engineering and Science Proceedings, The American Ceramic Society, Westerville,
OH, USA, Vol. 12, July-August 1991, pp. 1139-1151.
[6] Achenbach, J.D., Fine, M.E., Komsky, I., and McGuire, S., "Ultrasonic Wave
Technique to Assess Cyclic-Load Fatigue Damage in Silicon-Carbide Whisker
Reinforced 2124 Aluminum Alloy Composites," Cyclic Deformation, Fracture, and
Nondestructive Evaluation of Advanced Materials, ASTM STP 1157, M.R.,
Mitchell and O. Buck, Eds., American Society for Testing and Materials,
Philadelphia, PA, USA, 1992, pp. 241-250.
[7] Tiwari, A., Henneke, E.G., II, and Reifsnider, K.L., "Damage Characterization of a
Cross-Ply SiC/CAS-I~ Ceramic Composite Under Fatigue Loading Using a Real-
Time Acousto-Ultrasonic NDE Technique", Journal of Composites Technology and
Research, JCTRER, Vol. 17, No. 3, July 1995, pp. 221-227.
[8] MacLellan, P.T., Stubbs, D.A., and Karpur, P., "In Situ Ultrasonic Surface Wave
Assessment of Mechanical Fatigue Damage Accumulation in Metal Matrix
Composites," Composites Engineering, Vol. 5, No. 12, 1995, pp. 1413-1422.
[9] Baste, S., "Damage Mechanisms in CMCs As-Assessed by Ultrasonic

Measurements," High-Temperature Ceramic-Matrix Composites I: Design,
Durability, and Performance, C Ceramic Transactions, A.G. Evans and R. Naslain,
Eds., Vol. 57, pp. 131-140.
[10] Kline, R.A., "Quantitative NDE of Advanced Composites Using Ultrasonic

Velocity Measurements," Journal of Engineering Materials and Technology,
Transactions of the ASME, Vol. 112, April 1990, pp. 218-222.
[11] Steiner, K.V., Eduljee, R.F., Huang, X., and Gillespie, J.W., Jr, "Ultrasonic NDE
Techniques for the Evaluation of Matrix Cracking in Composite Laminates,"
Composites Science and Technology, Vol. 53, No. 2, 1995, pp. 193-198.
[12] Anson, L.W., Chivers, R.C., and Puttick, K.E., "On the Feasibility of Detecting Pre-
Cracking Fatigue Damage in Metal-Matrix Composites by Ultrasonic Techniques,"
Composites Science and Technology, Vol. 55, No. 1, 1995, pp. 63-73.
[13] Wooh, S.C., and Daniel, I.M., "Real-Time Ultrasonic Investigation of Damage
Development in Ceramic-Matrix Composite," Review of Progress in Quantitative
Nondestructive Evaluation, D.O. Thompson and D.E. Chimenti, Eds., Plenum
Press, Vol. 11, 1992, pp. 1523-1530.
[14] Hartman, G.A., and Ashbaugh, N.E., "A Fracture Mechanics Test Automation
System for a Basic Research Laboratory," Applications of Automation Technology
to Fatigue and Fracture Testing, ASTM STP 1092, A.A. Braun, N.E. Ashbaugh, and
F.M. Smith, Eds., American Society for Testing and Materials, Philadelphia, PA,
USA, 1990, pp. 95-110.
[15] Stubbs, D.A. and Buchanan, D.J., "In Situ Ultrasonic Monitoring of Damage in
Ceramic Matrix Composites Undergoing Creep and Tension Testing," Final Report,
University of Dayton Research Institute, Dayton, OH, USA, Prepared for United
Technologies Research Center, East Hartford, CT, May 1995.
[16] John, R., Buchanan, D.J., Stubbs, D.A., and Herzog, J.A., "Characterization of
Damage Progression in Ceramic Matrix Composites Using an Integrated
NDE/Mechanical Testing System," Thermal and Mechanical Test Methods and
Behavior of Continuous-Fiber Ceramic Matrix Composites, ASTM STP 1309,
Michael G. Jenkins, Stephen T. Gonezy, Edgar Lara-Curzio, Noel E. Ashbaugh, and
Larry P. Zawada, Eds., American Society for Testing and Materials, Philadelphia,
PA., 1996.
[17] Benson, D.M., "Evaluation of Damage Evolution and Material Behavior in

Sigma/Ti-6242 Composite Using Nondestructive Methods," Master of Science
Thesis, University of Dayton, Dayton, OH, August 1995.
[18] Brennan, J.J., Nutt, S.R., and Sun, E.Y., "Interfacial Microstructure and Stability of
BN Coated Nicalon Fiber/Glass-Ceramic Matrix Composite," High-Temperature
Ceramic Matrix Composites II: Manufacturing and Materials Development, A.G.
Evans and R. Naslain, Eds., Ceramic Transactions, Vol. 58, The American Ceramic
Society, Westerville, OH, USA, 1994, pp. 53-64.
[19] Hartman, G.A., and Buchanan, D.J., "Methodologies for Thermal and Mechanical
Testing of TMC Materials," Characterisation of Fibre Reinforced Titanium Matrix
Composites, 77th Meeting of the AGARD Structures and Materials Panel, AGARD
Report 796, Bordeaux, France, 27-28 September 1993.
[20] Kolsky, H., Stress Waves in Solids, Dover Publications, Inc., New York, NY,
USA., 1963, pp. 79-83.
[21] Morse, R.W., "The Velocity of Compressional Waves in Rods of Rectangular Cross
Section," The Journal of the Acoustical Society of America, Vol. 22, No. 2, March
1950.
J. A. Wang~
USING ULTRASONIC MEASUREMENTS AND A TWO-PHASE COMPOSITE

MODEL TO ASSESS RADIATION DAMAGE IN REACTOR PRESSURE VESSEL
STEELS
REFERENCE: Wang, J. A., "Using Ultrasonic Measurements and a Two-Phase

Composite Model to Assess Radiation Damage in Reactor Pressure Vessel Steels,"
Nontraditional Methods of Sensing Stress, Strain, and Damage in Materials and Struc-
tures, ASTM sTP 1318, George F. Lucas and David A. Stubbs, Eds., American Society
for Testing and Materials, 1997.
ABSTRACT: Ultrasonic methods used in the study of radiation damage and recovery in
single crystals appear to also be useful for similar studies on polycrystalline alloys.
Ultrasonic methods have demonstrated a sensitivity to radiation damage as affected by
neutron fluence, irradiationtemperature, large changes in composition, and possibly, as well,
by neutron energy spectrum. On the microstructure defect evolution, only the residual
defects created through the radiation event will contribute to the final macroscopic material
property change. From a microstructure point, it is generally accepted that radiation
hardening and embfittlement in metals are caused by clusters of vacancies, interstitial, and
solute atoms that impede the motion of slip dislocations. Although vacancy-type defects are
a major contributor to the material hardening, they also indicate the presence of other
interstitial defects. Thus the total volume change of vacancy-type defects before and after
irradiation can serve as a direct index to the final material property changes. The volume
change of the vacancy-type defects can be determined by utilizing the two-phase composite
model (matrix and void-type inclusion)to interpret wave velocities of baseline and irradiated
specimens that are obtained from the ultrasonic wave experiment. This is a relatively
economic and straightforward procedure. The correlation of the volume change of the
vacancy-type defects with the existing destructive mechanical test results may play an
important role in the future for the prediction of the radiation embrittlement and remaining
plant life, especially for the older plants on the verge of exhausting all the available
mechanical test specimens loaded in the surveillance capsules. The above hypothesis was
supported by the limited irradiated data analyzedand presented in this paper. The proposed
ultrasonic methodology also has a potential application to assess creep damage in fossil
power plants.
~Research Staff Member, Computational Physics and Engineering Division, Oak

Ridge National Laboratory, Oak Ridge, Tennessee 37831.
187
Copyright9
(PDVSA by ASTM
KEYWORDS: radiation embrittlement, ultrasonic, nondestructive, vacancy cluster, two-

phase composite, power reactor, pressure vessel steels
INTRODUCTION
It is well known that the fluxes of energetic neutrons and fission fragments in
neutron chain reactors can displace significantnumbers of atoms and thus alter physical and
mechanical properties [!]. From a microstructure point, it is generally accepted that
radiation hardening and embrittlement in metals are caused by clusters of vacancies,
interstitial, and solute atoms that impede the motion of slip dislocations. The degree of
embrittl~nent is conventionallycorrelated with fast neutrons or with total displacement per
atom (dpa). The radiation embrittlementof reactor pressure vessel (RPV) materials depends
on many different factors, primarily flux, fluence, neutron energy spectrum, irradiation
temperature, prdrradiation material history, and chemical compositions [2.3]. These factors
must be considered to reliably predict the pressure vessel embrittlement and to ensure the
safe operation of the reactor. Intensive studies have been and are being made to determine
the parameters which seem to affect sensitivity to radiation embfittlement.
The aging and degradation of light-water-reactor (LWR) pressure vessels is of
particular concern because of their relevance to plant safety and the magnitude of the
expected irradiation embrittlement. The decrease of ductility in PRV materials is a direct
consequence of irradiation hardening and embrittlement. Irradiation hardening can result
in increase of yield and ultimate strengths, but decrease of ductility. Radiation damage of
structural components would undoubtedly contribute to doubts against extending lifetime
&power reactors. Changes in material properties are monitored by periodic testing of RPV
surveillance specimens. Thermal annealing &reactor vessels may require additional testing
to ensure the effectivenessof such treatments. There are only limited surveillance specimens
loaded in the irradiated capsules, especially, for some older plants which may already be on
the verge of exhausting all the available surveillance specimens. Therefore, it is essential to
make the best use of the limited surveillance materials, Hence, there is a considerable
benefit to be able to conduct nondestructive tests on the same surveillance specimen
repeatedly and still be able to characterize the material properties of the reactor pressure
vessel. The combination of the ultrasonic techniques and analytical methods will allow the
determination of material moduli, strength, and toughness, and will provide a nondestructive
approach to characterizing the initial states of materials and their degradation or
modification due to exposure to service environments.
Ultrasonic methods, used in the study of radiation damage and recovery in single
crystals [4], appear to be useful for similar studies on polycrystalline alloys: Ultrasonic
methods have demonstrated a sensitivityto radiation damage as affected by neutron fluence,
irradiation temperature [5], large changes in composition, and possibly, as well, by neutron
energy spectrum. Ultrasonic methods offer further advantage of helping to define the nature
of recovery mechanisms due to a theoretical differing ultrasonic response behavior between
interstitial and vacancy-type defects [6_~.7].
CopyrightbyASTMInt'l(allrightsreserved);SatJan2808:00:22EST2012
Downloaded/printedby
(PDVSALosTeques)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
WANG ON USING ULTRASONICMEASUREMENTS 189
RADIATION DAMAGE INDEX
Irradiation effects, or damage, on the physical, thermal, chemical and mechanical

properties of nuclear reactor materials with time and other factors can greatly influence the
reactor design, operation, performance, and safety in engineering applications. The change
in these properties is attributed to the crystal (lattice) defects produced chiefly by neutron
irradiation during reactor operation. The crystal defects induced by radiation on nuclear
materials not only depend on the neutron flux, energy spectrum, irradiation time, and
irradiation temperature, but also on the crystal structure, impurity atoms, alloying elements,
heat treatment, etc. From the residual defects point, only the point defects that survive from
the bulk recombination and in cascade recombination can contribute to the final mechanical
property changes; the residual defects strongly depend on the material properties and
irradiation environments, such as chemistry composition, heat treatment, irradiation
temperature, and neutron spectrum.
From a mierostructure point, vacancy clusters are thought to be responsible for more
strengthening than are interstitial clusters at a given temperature and damage rate [~.~9.].
Vacancy clusters, such as dislocation loops, voids, and stacking fault tetrahedra, are strong
obstacles to the movement of dislocations and, therefore, may determine the material
hardening [10]. Currently, copper is considered as one of the primary contributors to the
radiation embrittlement of the low alloy steel. Copper and other face center cubic (fcc)
metals compositions are believed to stabilize vacancy-type defects in irradiation material
[1H5].
From a mechanics point, the overall material properties, such as bulk modulus and
shear modulus, depend on the properties of individual material components. If the volume
or the property of the component undergo a large transformation, this component will
definitely play a dominant role in the final overall material property change. The formation
of the vacancy-type defect is such a candidate due to the dramatic change of material
property in the matrix by forming a new phase with zero mass. Whereas for interstitial
clusters, which generally have similar atomic numbers as that of a matrix, the sensing of
change in material property may not be as great as that of vacancy clusters.
On the microstructure defect evolution, all the residual defects created through the
radiation events will contribute to the final macroscopic material property changes. The
residual defects in materials due to neutron-induced displacement damage are a function of
neutron energy, neutron flux, and exposure temperature, as well as the material properties
that determine how neutrons interact with atoms and how defects interact within the
material. Vacancy cluster is a major contributor to the material hardening at higher
temperatures and lower displacement rates, for example, at a commercial power reactor
environment [2]. Thus fi'om microstructure and mechanics points, vacancy-type defects not
only can be considered as a major contributor to the material hardening, but also indicate
the co-existence of other precipitates, such as copper. Furthermore, the total volume
change of the vacancy-type defects may serve as a direct index to the final residual defects
by taking into account the spectral effect, rate effect, and temperature effect, in addition to
the material properties, implicitly. Therefore, the total volume change of the vacancy-type
defects contained in the irradiated material may serve as a direct index to the radiation
embrittlement.
The volume change of the vacancy-type defect will serve as an important index to
the radiation damage, and its correlation to the mechanical test results may play an
important role in the future, especially for older plants on the verge of exhausting all the
available mechanical test specimens loaded in the capsule ofa RPV. The volume change of
the vacancy-type defects can be determined by utilizing the two-phase composite model
(matrix and void-type inclusion) to interpret wave velocities of baseline and irradiated
specimens, obtained from the ultrasonic wave experiment. The detailed description of this
proposed methodology is stated in the following sections.
TWO-PHASE COMPOSITE MODEL
Background of Composites
A composite material consists of a number of distinct homogeneous phases that form

regions large enough to be regarded as continua and that usually bond together at the
interface. Many natural and artificial materials are of this nature, such as some rocks (for
example, sandstone), filled polymers, mortar, concrete, precipitate, porous and cracked
media, etc. The development of mathematical models for composite materials is an
extremely complicated and difficult task because of the large number of factors on which
the behavior of the composite depends, such as phase factors [16-18], composition factors
[19-21], and operational factors [22]. In order to develop a rigorous model for a composite,
these factors must be under control, either by means of a particular model describing their
influence or by suitable assumptions.
From a practical point of view, two kinds of information determine the properties
of a composite material: the internal phase geometry [23] (i.e., the phase interface geometry)
and the physical properties of the phase (i.e., the constitutive relation of these); the former
is far more difficult to classify than the latter. Since the early 1960s, there has been much
work on the elastic properties of heterogeneous materials [24-27]. Much of this work is of
a fundamental nature and provides a sound basis for understanding and discussing the
properties of aggregates and composites. Due to the complexity of composite materials, the
exact analytical theory needs detailed modeling of the various factors, such as interfacial
phenomena, debonding and contacting inclusions, and phase geometry, etc. The modeling
of these factors is not always available, and then the effective bounds are usually too wide
for practical application. For engineering purposes, a more practical and general model is
desired. The modified direct method was developed to meet the above requirements [28].
Modified Direct Method
Two important assumptions are usually made when treating the macroscopic elastic
properties of composite materials: statistical homogeneity and statistical isotropy [29]. For
the problem to be workable, a necessary characteristic of a composite material is statistical
homogeneity. This assumption essentially means that large enough subregions of the
composite are statisticallyidentical with the whole sample. The most frequently considered
subregion of the multiphase medium is the two-phase composite in which one phase may
WANG ON USING ULTRASONIC MEASUREMENTS 191
be regarded as matrix in which arbitrarily shaped inclusions of other phases are embedded.
The effective properties of a composite material, when their space variation is statistically
homogeneous, define the relations between averages of field variables, such as stress and
strain. The evaluation of an average elastic modulus of a composite material or polycrystals
is one of the classical problems in micromechanics. The pioneering works have been done
by Voigt and Reuss. The computation of effective properties in terms of the average is
called the direct approach. In general, it requires determination of the appropriate fields in
the phases as defined by the field equations, interface continuity conditions, and external
homogeneous boundary conditions in order to compute the required average.
The modified direct method [28] is a scheme for the estimation of elastic moduli of
composite materials and is based on micromechanical theory and classical elasticity. Using
the statistical homogeneous assumption and the two-phase composite approach, one takes
the average field of the composite. The spherical shape inclusion is the mean geometry of
inclusions with a range of spectrum of arbitrary shapes, and is used in the analysis. In ref.
28, comparison of modified direct method with experimental data and with other well-
known methods, such as Christensen-Lo, Kuster-Toksrz, and the self-consistent method,
shows that the modified direct method provides a very good estimation of the elastic moduli
in different kinds of problems, such as the sott and hard inclusion cases, porous materials,
at various concentrations and/or various porosities.
Based on the theoretical derivations in the modified direct method for the
composites, the effective moduli, g and/~, for the isotropic materials, can be written as
= rt + c 2 ( g * - g )
g = K + c2(X* - (1)
For small concentrations, ~ and ~ can be written as
= (15K+20g)g ~ _ (4g+3K)
, (2)
(9K + 8g)g + (6K + 12~)~* 4~ + 3K*
where g and la*are the shear moduli for the matrix and inclusion, respectively. K and K* are
the bulk moduli for the matrix and inclusion, respectively, and c2 is the concentration of the
inclusion.
For materials with a high concentration of inclusion or high porosity, the concept of
an infinite matrix in the composite domain cannot be used. Instead, a finite subdomain
containing the inclusion lumped and originating at a central position, maintaining the same
volume ratio as in the original composite material, was used in the modified direct method.
These subdomains are evenly distributed throughout the whole domain. The derived
effective stiffnesses can be written as
g = K +r -
(3)
where K'u can be written as
= 1 +(1-R)(~-I) = ~(1-R)+R O)
and R is defined as the modification factor, and can be written as
_ - _ _ (5)
The Relation Between Acoustic Velocities and Elastic Modulus
The elastic properties can be determined by measuring the acoustic velocities, V?

(compression wave) and V s (shear wave), in a sample specimen. Once the wave speeds
and density are known, then the elastic moduli can be determined directly from the
following relations:
(
K = p V2 - ~ , g : PV2 (6)
where p is mass density, which is a function of c~. The Young's modulus E and the
Poisson's ratio v can be determined as follows:
E- 9Kla , v - 3K-2p. (7)

3K + la 6K + 2g
The experimental apparatus used to measure the wave speeds of the sample specimens
is depicted in Fig. 1, which contains a block diagram of the experimental set-up. The
system is composed of a pulsed oscillator, digital oscilloscope and Carver press, wave
switch box, transmitter and receiver transducer, data acquisition system, mini computer,
and plotter. The input signals are provided by a pulse generator for velocity
determinations. A delay line incorporated in the oscilloscope is used for travel time
measurements.
Interpretation of Radiation Damage by Ultrasonic Waves and Two-Phase Composite

Model
The total volume change of vacancy-type defects before and after the irradiation can
be used as a primary index to the radiation damage. Therefore, the overall radiation damage
to the material can be estimated through the proper calibration of total volume change of
vacancy-type defects.
!,.,oo !
i|1
"~Computer ~ S
sF(eceiver
.,o... j i I 1
FIG. 1--Ultrasonic experimental set-up.
The originalunirradiated sample can be defined as the matrix, and the post-irradiated
sample can be defined as a two-phase composite, which includes the original matrix plus the
void-type inclusions(the vacancy-type defects generated through neutron radiation event).
Based on this two-phase composite model and applying Eqs 1-6 to analyze wave vdocities
for both unirradiated and irradiated samples obtained from the ultrasonic waves experiment,
the quantity of the total volume change of vacancy-type defects before and after irradiation
can be determined.
A preliminaryanalysisof TrantovCsultrasonic data on 304 stainless steel [30], which
incorporated a two-phases composite modeling, shows that the estimated total volume
change of vacancy-type defects increased with increasing neutron fluence, as illustrated in
Fig. 2. These analytical ultrasonic test results demonstrated a very similar trend as that
obtained from the destructive mechanical test results; namely, increased embfittlement
damage with increased fluence or dpa, see Fig. 3 [31]. Thistrend demonstrates the potential
of using the proposed new parameter (total volume change of vacancy-type defects) for
evaluating neutron damage in typical low-alloy RPV steels.
Furthermore, Fig. 2 clearlyindicatesthat the defects generation rate of vacancy-type
defects are strongly dose dependent, that is at relative low dose the damage production rate
of vacancy-type defects is small. This stage can be analogous to the incubation or
nucleation stage, where aiter a certain threshold dose, the defects production start to take
off, in analogy to the growth stage.
Currently, dpa is used as a measure of the potential to create point defects. It is not
necessarily equal to, or even proportional to, the actual number of residual point defects.
Clearly, a complete and accurate description of microstructural development during
irradiation has not been attempted and seems to be beyond the present state-of-the-art.
However, a better correlation parameter for radiation embrittlement can be expected
through the upgraded dpa that incorporates the concepts of residual defects.
194 S E N S I N G S T R E S S , S T R A I N , A N D D A M A G E IN M A T E R I A L S
0.020
i
#
0.018 t
304 Stainless Steel I
I
l 0.016
0.014
371 < Temp_irr <378~
(700 < Temp_irr < 713~
!
I
Ip
I
I
I
0.012 I
i
I
0.010 I
1
I
15 I
o 0.008 I
r
I
I
0.006 I
a I
I
0.004 I
1
I
0.002
0.000 J
1.0E§ 1.0E§
F l u e n c e , E > .1 M e V [ n l c m =]
FIG. 2--Plot of total volume change of vacancy-type defects vs fluence.
a HFIR A212B key 6,7 with #

525 DPA(Neutron+ Gamma) jr
o HFIR A212B in O R R o /
# ,
O.
=E Fit per
U) 475
l,kl
1--
O9
425
9
>-
375
0
325
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01
DISPLACEMENTS PER ATOM [dpa]
Fig. 3--Plot of tensile yield stress vs. dpa. [31]
CONCLUSION
This paper has demonstrated the feasibility of using ultrasonic nondestructive

measurements and analytical techniques to characterize the primary parameter of radiation
damage. The correlation of the analytical ultrasonic test results (volume change of the
vacancy-type defects) with the existing destructive mechanical test results may play an
important role in the future for the prediction of the radiation embrittlement and remaining
plant life. Moreover, the linear and nonlinear characteristics of the ultrasonic testing
techniques also have potential applications to assess creep damage in fossil power plants,
fatigue damage in structural components, and to estimate damage of structural members
in components of the nation's infrastructures during service (such as bridges).
Furthermore, the new integrated analytical/experimental ultrasonic testing technique will
enable in-situ monitoring of material property changes during service. Then it will be
possible to carry out inexpensive, faster and more comprehensive assessments of both
material quality and structural integrity.
REFERENCES
[1] Mansur, L. K., "Mechanisms and Kinetics of Radiation Effects in Metals and Alloys,"
Kinetics ofNonhomogeneous Processes~ ed. by Gordon R. Freeman, 1987.
[ ~ Odette, G. R., Lombrozo, P. M., and WuUaert, R. A., "Relationship Between

Irradiation Hardening and Embrittlement of Pressure Vessel Steels," Effects of
Radiation on Materials, Twelfth International Svmoosium. ASTM STP 870, eds.
Garner, F. A. and Perrin, J.S., American Society for Testing and Materials,
Philadelphia, 1985, pp.840-860.
[3] Lucas, G. E., Odette, G. R., Lombrozo, P. M., and Sheckherd, J. W., "Effects of
Composition, Microstructure, and Temperature on Irradiation Hardening of Pressure
Vessel Steels," Effects of Radiation on Materials; Twelfth International Symposium
ASTM STP 870, eds. Garner, A. and Perrin, J., American Society for Testing and
Materials, Philadelphia, 1985, pp. 900-930.
[4] Thompson, D. O., and Pare, V. K., "Use of An elasticity in Investigating Radiation
Damage and Diffusion of Point Defects," Physical Acoustic Princioles and Methods.
Volume III-Part A, Mason, W. P., ed., Academic Press, New York, 1966, P. 293.
[5] Hunter, D. O., "Ultrasonic Velocities and Critical-Angle-Method Changes in Irradiated

A302-B and A542-B Steels," BNWL-988, May 9, 1969.
[6] Dienes, G. J., "A Theoretical Estimate of the Effect of Radiation on the Elastic
Constants of Simple Metals," Physical Review. Vol. 86-2, p, 228; April, 1952.
[7] Dienes, G. J., "Effect of Radiation on Elastic Constants," Physical Review. Vol, 87-4,
P. 665, Augur, 1952.
[8_] Spitznagel, J. A., and Venskytis, F. J., "The Annealing of Copper Vacancy Aggregates
in Neutron Irradiated Ferritic Pressure Vessel Steels, TRANS ANS 1975, p. 161.
[_9] Stoller, R. E., "Modeling The Influence of Irradiation Temperature and Displacement
Rate on hardening Due to Point Defect Clusters in Fen-itic Steels,"Effects of Radiation
on Materials: 16th International Symposium. ASTM STP 1175, Arvind S. Kumar,
David S. Gelles, Randy K. Nanstad, and Edward A. Little, Editors, ASTM, 1993,
pp.394-423.
[10] Jung, P., "Relevance of the Displacement Damage Concept for the Evaluation of
Radiation Effects in Metals," Radiation Effects and Defects in Solids. Vol. 113, 1990,
pp. 109-118.
[11] Brenner, S. S., Wagner, R. and Spitznagel, J.A. "Field Ion Microscope Detection of
Ultra-Fine Defects in Neutron-Irradiated Fe-34% Cu Alloy",
Transactions. Vol. 9A~ 1978, pp. 1761.
[12] Smitd, F.A. and Sprague, J.A. "Property Changes Resulting from Impurity-Defect
Interactions in Iron and Pressure Vessel Steel Alloys," Effects of Radiation on
Substructure and Mechanical Properties of Metals and Alloys, ASTM STP 529, 78,
(1973).
[13] English, C. A., "Recoil Effects in Radiation Damage," Radiation Effects and Defects
in Solids, Vol. 113, 1990, pp. 15-28.
[14] Simons, R. L., "Damage Rate and Spectrum Effects in Ferfitic Steel NDTT Data,"
Influence of Radiation on Material Pronerties: 13th International Svmnosium (Part II).
ASTM STP 956. F.A. Garner, C. H. Henager, Jr., and N. Igata, Eds., American
Society for testing and Materials, Philadelphia, 1987, pp. 535-551.
[.1~ Odette, G. R., and Sheeks, C. K. ,"A model for Displacement Cascade-Induced
Microvoid and Precipitator Formation in Dilute Iron-Copper Alloys," Phase Stability
During Irradiation, J. R. Holland, L. K. Mansur, and D. I. Potter, Eds., Metallurgical
Society of the American Institute of Mining, Metallurgical and Petroleum Engineers,
(1982), p. 415.
[16] Bland, D.R. (1960). The Theoryof Linear Viscoelasticity. Pergamon Press, Oxford,
London.
[17] Tobolsky, A.V. (1956). "Stress Relaxation of the Viscoelastic Properties of Polymers."
Journal of Applied Physics. Vol. 27. pp. 673-685.
[18] Timoshenko, S., and Goodier, J.N. (1969). Theoryof Elasticity, Mcgraw-Hill.
WANG ON USING ULTRASONIC MEASUREMENTS 197
[19] Hashin, Z. (1970). "Mechanics of Composite Materials." proceedings: 5th SvmDosium.

Naval Structure Mechanics.
[20] Wu, I.N., and Sendeckyj, G.P. (1974). "Multiple Circular Inclusion Problems in Plane
Elastostatics." Journal of Annlied Mechanics. Vol. 41. ser. E, pp. 215-221.
[21] Goree, J.G., and Wilson, H.B. (1967). "Axisymmetric Torsional Stresses in a Solid
Containing Two Partially Bonded Rigid Spherical Inclusions." Journal of Applied
Mechanics; Vol. 34. ser. E, pp. 313-320.
[22] Theocaris, P.S., and Paipetis, S.A. (1978). "Indentation Studies in Aluminum-Filled
Epoxies." Journal of Applied Polvmer Science. Vol. 22~ pp. 2245-2252.
[23] Papanicolaou, GC., and Paipetis, SA. (1978). "The Concept of Boundary Interphase
in Composite." Colloid and Polymer Science. Vol. 256, pp. 625-630.
[24] Wilson, H.B., and Hashin, Z. (1962). "The Elastic Moduli of Heterogeneous
Materials." Journal of Applied Mechanics. Vol. 29. pp. 143-150.
[25] Budiansky, B. (1965). "On the Elastic Moduli of Some Heterogeneous Materials."
JOurnal of Mechanics Phvsics Solids. Vol. 13, pp. 223-227.
[26] Hill, R.A. (1965). "Self Consistent Mechanics of Composite Materials." Journal of
Mechanics Phvsics Solids, Vol. 13, pp. 213-222.
[27] Walpole, L.J. (1969). "On the Overall Elastic Moduli of Composite Materials." Journal
of Mechanics Phvsics Solids, Vol. 17, pp. 235-251.
[28] Wang, J. A., Lubliner, J., and Monteiro, P. J. M, "The Modified Direct Method for
the Calculation of the Elastic Modulus of Composite M a t e r i a l s , " ~
First International Conference on Comoosites in Infrastructure. Eds. H. Saadatmanesh,
and M. R. Ehsani, Tucson, Arizona, pp. 80-94, January 1996.
[29] Hashin, Z. (1964). "Theory of Mechanical Behavior of Heterogeneous Media." Applied

Mechanics Reviews. Vol. 17. pp. 1-9.
[30] Trantow, R. L., "Ultrasonic Measurement of Elastic Properties in Irradiated 304

Stainless Steel," HEDL-TME-73-92, December 1973.
[31] Remec, I., Wang, J. A., Kam, F. B. K., and Farrell, K., "Effects of Gamma-Induced
Displacements on HFIR Pressure Vessel Materials," J. Nuclear Mater. 217 (1994) 258-
268.
Christopher S. Welch 1, K. Elliott Cramer~, Jon R. Lesniak 3, and Bradley R.
Boyce3
AN ARRAY MEASUREMENT SYSTEM FOR THERMOELASTIC

STRESS ANALYSIS
REFERENCE: Welch, C. S., Cramer, K. E., Lesniak, J. R., and Boyce, B. R.,
"An Array Measurement System for Thermoelastic Stress Analysis,"
Nontraditional Methods of Sensing Stress, Strain, and Damage in Materials
and Structures, ASTM STP 1318, George F. Lucas and David A. Stubbs,
Ecls., American Society for Testing and Materials, 1997.
r
ABSTRACT: Thermoelastic Stress Analysis (TSA) is defined, and an overview

is given to show its potentially wide applicability and brief history. A simplified
calculation of instrumental response characteristics shows the substantial
improvement in data acquisition speed that can be gained in a system using a
staring array in comparison to that using a scanned detector. Descriptions are
given of several methods which have been used to obtain data for thermoelastic
stress analysis. Finally, image representations are presented of TSA data taken
with a staring array imager showing the resolution attainable in practice using
some laboratory fatigue samples and a prototype part.
KEYWORDS: thermoelastic stress analysis, full-field stress analysis, stress

imaging, staring infrared detector arrays
INTRODUCTION
Thermoelastic stress analysis (TSA) is an experimental technique for

stress analysis. When conditions arise or may be arranged in which
time-variable stress can be applied to or observed in a test object, TSA can
lSr. Research Scientist, Applied Science Department, College of William

and Mary, Williamsburg, VA 23187
2Aerospace Technologist, Nondestructive Evaluation Sciences Branch,
Materials Division, NASA Langley Research Center, Hampton, VA 23681
3Vice-President and President, respectively, Stress Photonics, Inc., 3002
Progress Road, Madison, WI 53716
198
Copyright9
(PDVSA Los Teques)by ASTMtoInternational
pursuant License Agreement.www.astm.org
WELCH ET AU ON AN ARRAY MEASUREMENT SYSTEM 199
generally be employed to provide nondestructive, full-field data from which

stress may be inferred. The recent availablility of staring array infrared
detectors has permitted development of a TSA system which provides results in
a timely manner--a matter of minutes or even seconds. The opportunities to use
stress measurements enabled by timeliness of the results bring to mind a
variety of uses from the fields of product design, manufacturing process control,
materials evaluation and testing, and nondestructive evaluation.
The potential available in the thermoelastic effect and its curious

coupling of stress to reversible temperature fluctuations have attracted interest
over the years from a variety of investigators. The earliest feasibility study of a
stress measurement system using radiometric temperature measurement was
reported by Belgen [1]. A description of a commercially available system was
provided by Oliver, et aL [2], and the commercial availability of a thermoelastic
stress measurement system attracted worldwide industrial and academic
attention towards the development and application of TSA. A review of the
emerging field is given in Harwood and Cummings [3], and a comprehensive
bibliography has been compiled by Zhang and Sander [4].
One of the attractions of the thermoelastic effect is that the relation

between stress change and temperature change can be expressed in terms of
physical characteristics of the material in question. In the industrially important
case of homogeneous, isotropic materials, the thermoelastic equation is
expressed [3] as:
AT : - (odC) To Aoii, (1)
where AT is the change in temperature, Aoii is the sum of the principal stresses,
To is the mean absolute temperature, cr is the coefficient of linear thermal
expansion, and C is the volumetric heat capacity (the product of density and
heat capacity). The equation takes this form as an approximation valid at most
laboratory temperatures for which To is much larger than the temperature
fluctuations (on the order of 0.1K) attained by most load-carrying materials at
half the yield stress.
The physical origin of the temperature change is the conversion of

external work to internal energy through the application of load under elastic
(well below yield stress) and adiabatic conditions (that is in such a way that no
significant heat conduction can take place). Thermodynamically, it is very
similar to the adiabatic heating which occurs during the compression of gases,
and so it is a reversible process. If the load is immediately released, the
temperature immediately reverts to its previous pattern. If the temperature
pattern under load is not uniform, corresponding to non-uniform stress
distribution, it is subject to the irreversible process of thermal conduction, and
tends to dissipate in surrounding material. This is the reason that Eq. 1 relates
changes in temperature to those of stress. Also important is that the stress itself
is not measured, but the sum of principal stresses, the first stress invariant. This
also is similar to gas compression, for it involves a change in volume. The sum
of principal stresses vanishes in pure shear stress, so the thermoelastic

measurement requires some interpretation before engineering decisions are
made on its basis.
The small values of observable temperature and the tendency for

patterns of temperature to dissipate lead to conflicting time requirements for the
measurement, long times being needed to attain precise measurements and
short times being available before dissipative degradation of the data. An
effective method of resolving the conflict is to produce a rapidly oscillating stress
at a constant frequency and measure the amplitude and phase of the
associated temperature oscillations. The frequency is chosen to be high
enough that negligible dissipation occurs during an oscillation period, and the
temperature oscillation can be measured for long enough that high precision
can be attained. This measurement method is sometimes called complex
demodulation. If this method is used, the stress and temperature changes in
Eq. 1 can be interpreted as amplitudes of the oscillation, and the minus sign
becomes a 180 degree phase shift.
THE MEASUREMENT REQUIREMENT
To gain some intuition for the measurement precision which can be

attained with careful practice in many situations, 0.001K has been used by
several practitioners as a good benchmark for temperature resolution attainable
with a practical radiometric measurement. Table 1 shows the stress resolution
TABLE 1--Stress levels in various materials which produce a thermoelastic

temperature change of 0.001K. Sample temperature is considered to be
300K unless otherwise noted.
Material Stress, MPa Stress, psi Yield Stress, (ksi)
Pure Aluminum (1100, annealed) 0.313 45.4 11.
Ice (@ 260 K) 0.094 13.7
Steel (AISI M1010, hot rolled) 1.12 162.5 30.
Invar 4.27 619.4
Wood (Oak) Parallel to fiber 0.908 131. 6.9
Wood (Oak) Perpendicular to fiber 0.082 11.9
Plexiglass 0.073 10.6 14.
WELCH ET AL. ON AN ARRAYMEASUREMENTSYSTEM 201
attainable in some different materials at 0.001 K temperature resolution based

on Eq. 1 and typical handbook values for representative material properties,
and the numbers presented are meant to be suggestive only. Aluminum, steel
and wood represent common classes of structural materials. Invar is a trade
name for an iron-based alloy which is formulated to have a small coefficient of
thermal expansion. Plexiglass is a common commercial plastic which may be
used for making models or demonstration objects. In all cases, the dynamic
range (maximum allowable signal divided by noise level) available with
measurements of .001K within the elastic range of the materials is sufficient to
describe many stress features.
Next, we examine some factors which determine the instrumental

response from a photon-limited radiation measurement system. Because the
measurement is limited by photon counts, the limiting physical process is the
statistical uncertainty associated with counting statistics. Equation 2 is Planck's
formula for black-body radiation expressed in terms of the photon count rate [5].
n(X,'r) = 2 ~ c / { ;L4 [exp(h c / ;L k T )- 1]} (2)
This rate is of photons exiting a surface into a surrounding hemisphere per unit
of surface area per unit of time per unit of wavelength. In Eq. 2, ~. denotes the
wavelength of the radiation, T denotes the temperature of the black body, and c,
h and k denote the speed of light, Planck's constant and Boltzmann's constant
respectively. The number of counts expected over a period of time, t, is just the
rate times the period. With a staring radiometer, t is the frame time, while with a
scanning radiometer, it is the frame time divided by the number of pixels per
frame. To account for the wavelength sensitivity of the detector, Equation 2
must be multiplied at each wavelength by the transmissivity of the detector
optics and the product integrated over wavelength between the limiting
wavelengths of the appropriate window over which the detector is sensitive. For
presently available staring arrays, the optics are designed to pass in the
atmospherically clear region from 3 to 5 microns, while for scanning detectors,
that window or one from 8 to 12 microns is often used. For this calculation, the
optics are considered to be perfectly transmitting between the limiting
wavelengths, and the detector is considered to have a quantum efficiency of
0.7, which is characteristic of some presently available systems. From each
location on the surface of the object, the fraction of emitted radiation actually
passing through the imager aperture and striking the detector must be
determined. This factor can be shown to be equivalent to the solid angle of the
lens as seen by the detector times the area of the detector. A final factor is
associated with Lambertian radiation distribution, typical of a majority of
surfaces encountered in mechanical testing, and a conversion between
fractional hemispheres and steradians for solid angle representation.
These are the factors required to evaluate the sensitivity which would be
obtained with a perfect detector in an infrared imaging system. The limiting
noise in such a detector is given by counting statistics as the square root of the
number of counts. Only some of the factors, the aperture of the imager, the
number of pixels in an image, the field of view of the imager, the frame time,
and the wavelength window for the detector, may be chosen by a designer. Of
course, a designer is also balancing other factors not considered here such as
weight, size, stability, reliability, availability of parts, target market, and cost. A
comparison of optimal performance for a scanning and staring array detector
will serve to show the advantages of a staring detector system. This
comparison is shown in Table 2. The values chosen are intentionally not
descriptive of any particular instrument, but rather generic and, in our view,
appropriate to the two approaches. From the table, one can see that the
scanning imager has a basic 50:1 advantage in counts at room temperature,
because single detectors for scanning systems are readily available for the 8-
12 micron atmospheric spectral window, which is close to the peak of the
blackbody spectrum at 298 K. Photovoltaic staring arrays are not yet readily
available in the 8-12 micron window, but are available in the 3-5 micron
window, which produces much less blackbody radiation at room temperatures.
Comparing the counts at 298.001 K with those at 298 K, one can calculate that
the percentage sensitivity is about twice as large in the 3-5 micron range as the
8-12 micron range, somewhat offsetting the greater intensity for the 8-12 micron
radiation. The overwhelming difference in potential performance comes from
the Pixel Integration Time per Frame, which accounts for the ability of the staring
detector to gather data continuously from all of its pixels, while the scanning
detector must gather its data sequentially from each pixel in turn. Following the
photon-limited calculation through to its end, the potential performance which
might be attained with staring detectors is much better than the present stage of
development.
In current practice, available staring detectors beginning to approach the

performance associated with the photon counting limit, for they are limited by
readout electronics and a demand for increased resolution. The smaller number
of counts in the 3-5 micron range becomes an advantage here, so the extra
counts in the 8-12 micron range are of limited advantage. Presently available
imaging thermoelastic equipment achieves temperature resolution of 0.001 K in
about 30 seconds, several hundred times slower than the photon limit, yet still
30 times the speed of a scanning detector. Because the limiting factors seem
more like matters of technical development than fundamental physics, we may
expect continued improvement in the near future.
TECHNICAL APPROACHES
Since the original feasibility study [1], a number of technical approaches

have been followed to produce stress images using the thermoelastic effect.
Belgen himself used an optically impressive spot radiometer with
Cassegrainian reflecting optics with an aperture diameter of 0.20 m (8 inches)
and a spot size of 2.5 mm (0.1 inch) at a distance of 2.54 m (100 inches) with a
resonantly vibrating beam source and an immersed bolometer detector driving
a narrow-band filtered amplifier. Mountain and Webber[6] developed a
WELCH ET AL. ON AN ARRAY MEASUREMENTSYSTEM 203
TABLE 2 -- Optimal performance calculation for a scanning imager and a

staring array imager suitable for thermoelastic stress analysis.
Factor Scanning Imager Staring imager
Wavelength range 8-12 microns 3-5 microns
Hemispheric photon rate @ 298 K 5,888 E 21 1,223 E 20
Additional photon rate for 1 mK

temperature increase 9,630 E 16 4.445 E 15
Common Factors for Both Imagers:
Optics f/2.3
Percent of hemisphere subtended

by aperture 2.28 %
Quantum efficiency 0.7
Imager frame rate 30 frames/sec
Detector area (40 micron detector) 1.6 E -9 m2
Pixels per image (256 x 256) 65536
No of detectors 1 65536
Pixel integration time per frame 5.086 E -7 s 3.33 E -2 s
Pixel photon counts per frame 151569 2.063 E 8
Additional pixel photons for

1 mK temperature increase 2.479 7499.2
Pixel noise counts per frame 389.318 14362.2
Noise equivalent temperature difference 157.05 mK 1.92 mK
Time to achieve temperature resolution of 0.001K 822.13 s 0.122 s
thermoelastic detection system which synchronously detected the thermoelastic

signal and used the synchronous detection to drive a strobe signal to capture
the infrared signals at the peak of their oscillation, subtracting the values for the
positive and negative peaks to produce the therrnoelastic part of the signal.
Images were produced by building up many measurements of points in a raster
pattern. Oliver et al. [2] describe a further development of this system to utilize
an analog correlator which utilizes all of the signal rather than simply peak
values for increased efficiency. In an alternative approach based on
manipulation of image data, Heyman [7] reported a thermoelastic measurement
of stress distribution around a hole in a thin plate obtained by averaging
multiple subtractions of image pairs taken before and after application of a rapid
stress change in a sample. In a further development, Welch et al. [8] reported
thermoelastic data obtained by post-processing a number of images for which
the driving force was synchronized to the imager time scale. As noted by Bouc
and Nayroles [9], who obtained data on a line-by-line basis from an uncoupled
cyclic driver and imager, the stability of frame times is not sufficient simply to
perform Fourier analysis on the time series obtained by the imager, but a
reference signal is also required to maintain phase stability in the analysis long
enough to reduce noise levels to the 0.001 ~ range. They addressed the
problem by digitizing the reference signal along with the video signal for use in
data analysis. In all of these techniques, a number of spatial samples of the
infrared radiation are obtained, but only a single reference signal is needed to
produce the desired correlations. Because the data are photon limited, signal-
to-noise ratio, and hence stress resolution, is increased if array radiometer data
are available continuously for all pixels (10).
SOME EXAMPLE DATA
The full-field data available from digitally-processed imaging are stored

in the form of arrays of numbers, each possessing the full precision of the
measurement. These arrays are available for further processing, but the array
representation is difficult to comprehend immediately in terms of stress analysis.
Modem imaging instruments are able to take advantage of image presentation
techniques to provide the data in an intuitively interpretable manner within
seconds of data acquisition. These display products are more easily compre-
hended, but they often involve some loss in data precision. The precision of the
underlying data remains, however, and other display representations may be
calculated as desired by an analyst. We present four gray scale images from a
full-field stress imager to provide examples of the stress patterns available to an
industrial analyst within one minute from the start of the data cycle using
commercially available equipment 4. All images were obtained at stress levels
between 10 and 20 percent of the yield stress of the material, in this case
aluminum, under test. Three are of modified hole-in-plate samples of 47.5 mm
(1.88 in) in width with a centered hole having a diameter of 5.0 mm (0.19 in).
These are shown in Fig. 1. The fourth is an aircraft fitting, which represents
4DeltaTherm 1000, manufactured by Stress Photonics, Inc., Madison, Wi.
WELCH ET AL. ON AN ARRAY MEASUREMENTSYSTEM 205
FIG. 1 -- Three stress images of holes in plates with fatigue cracks. Shown is
the central region surrounding a hole 5 mm in diameter in a 48 mm width plate.
The plate is pulled with 51 mm grips at located more than 100 mm from the hole
center. In la, the crack has been initiated with a saw cut visible to the left of the
hole.. In lb, a crack has propagated almost a hole diameter from an initiating
notch. In lc, the plate is thick, and the crack has not yet penetrated the visible
face, but the redistribution of stress from it is apparent.
a fabricated part which might be inspected, as in Fig. 2, to verify the results of a

numerical model. A similar inspection might be used prior to service to verify
appropriate stress distribution in comparison with a standard part to indicate
FIG. 2 -- Thermoelastic stress image of an aircraft fitting in a test jig (see

schematic) configured to produce realistic working loads, on the order of 1000
lb. High stress areas correspond to the findings of a computer model.
freedom from flaws. In all cases, the samples are subjected to small oscillations
at 20 Hz around a tensile load, which is applied in the vertical direction seen in
the data images. The three hole-in-plate samples have cracks and notches on
the left-hand side of the holes, with the tensile load from a servohydraulic load
frame applied vertically in the images. This loading produces additional tensile
stress on the sides of the holes and small compressive~ regions above and
below the hole. The holes themselves show zero stress, with some artifacts
producing highlights or shadowing on the edges. The artifacts are caused by
slight motion of the hole with the load and, in some cases, slightly oblique
viewing angles between the imager and the hole, with exposure of the inside
wall of the hole over half of its perimeter. From a nondestructive evaluation
diagnosis perspective, it is notable that all of the compressive parts of the stress
patterns, on the top and bottom of the hole, are not laterally symmetric with the
hole for all of the flaws, which are located on one side of the hole. For unflawed
holes, the compressive areas are laterally symmetric with the hole. In Fig. 2, the
aircraft fitting, a single piece with two hoops mounted on a plate with an oblong
hole between the hoops, is fastened with four bolts to a custom jig in the load
frame, simulating its attachment to the airframe. For the thin plate samples, the
tensile load is used to avoid sample buckling under compression. For the
aircraft fitting, the tensile load is maintained to keep a linkage from becoming
slack, it is loaded with a pin, which extends just beyond the right edge of the
image. The pin and bolt heads are missing some or all of their
emissivity-enhancing coating, so they appear with speckles caused by moving
reflections as the assemblage strains slightly under the load. Two small, high
stress regions, one compressive and the other tensile, are seen near the left-
hand corners of the oblong hole. In this instance, these features provide
verification for similar features which were found in a finite element analysis
model of the same part. In all cases, the image data shown were produced with
measurement times of less than one minute.
SUMMARY AND CONCLUSION

This paper has described an advancement in the instrumentation
available for full-field thermoelastic stress analysis. The advancement is the
use of newly available staring arrays for infrared radiation detection. The
potential advantage of such arrays, only partly realized in the equipment, is
indicated in a calculation of comparative radiometric performance with a
hypothetical scanning and staring array system. Images of thermoelastic stress
data obtained with staring array equipment are presented to provide an
indication of presently available instrumental capability. The rapidity of
acquisition of these images, less than a minute, is anticipated to make a number
of applications of thermoelastic stress analysis economically attractive in
industrial design and inspection situations where the time required for data
acquisition had previously kept thermoelastic stress analysis out of
consideration.
WELCH ET AL. ON AN ARRAYMEASUREMENTSYSTEM 207
REFERENCES
(1) Belgen, Milo H., "Infrared Radiometric Stress Instrumentation Application
Range Study," NASA Contractor Report CR-1067, May, 1968, 142 pp.
(2) Oliver, D. E., Razdan, D., and White, M. T., "Structural Design
Assessment Using Thermoelastic Stress Analysis (TSA),"
Joint Conference:" 'State of th~ Art' in Measurement Techniaues," The
University of Surrey, 6-9 September, 1982, 23 pp.
(3) Harwood, N. and Cummings, W. M., eds., Thermoelastic Stress Analysis,
Adam Hilger, Philadelphia, 1991.
(4) Zhang, D., and Sandor, B. I., "Thermographic Stress Analysis," in

Nondestructive Testina Handbook, edited by R.K. Stanley, Second
Edition, Volume 9, Section 6, Part III, Published by the American Society
for Nondestructive Testing, 1995.
(5) Wolfe, W. L and G. J. Zissis, eds., The Infrared Handbook, The

Environmental Research Institute of Michigan, 1978.
(6) Mountain, D. S., and Webber, J. M. B., Proceedinas of the Society of

Photo-Ootical Instrument Enaineerina. 164, (1978), pp. 189-196.
(7) Heyman, J. S., "Advanced NDE Technology for Materials

Characterization," DoD Manufacturina Technoloav Advisory Grouo Test
and Insoection Mini-Symposium, November 25-2~Ji 1984, 17 pp + figs.
(8) Welch, C. S., Cramer, K. E., and Dawicke, D. S., "Thermographic

Inspection of Fatigue Cracks in Riveted Plates," Societv of Exoerimental
Mechanics Sorina M~etin_a, Milwaukee, Wl, June 10-13, 199i, pp. 239-
243.
(9) Bouc, R. and Nayroles, B., "M~thodes et r~sultats en Thermographie

Infrarouge des Solides," Journal de M~canieue Th~orieue et A opliau~e
4(1 ), 1985, pp. 27-58.
(lO) Lesniak, J. R. and Boyce, B. R., "A High-Speed Differential

Thermographic Camera," Proceedinas of the Society of ExDerimental
Mechanics Sorina Conference, June 6-8, 1994, Baltimore, MD, pp. 491-
497.
Jon R. Lesniak ~, Daniel J. Bazile ~, Bradley R. Boyce ~, Michael J. Zickel ~, K. Elliott
Cramer:, Christopher S. Welch 3
STRESS INTENSITY MEASUREMENT VIA INFRARED FOCAL PLANE

ARRAY
REFERENCE: Lesniak, J, R., Bazile, D. J., Boyce, B. R., Zickel, M. J., Cramer, K.
E., and Welch, C. S., "Stress Intensity Measurement via Infrared Focal Plane
Array", Nontraditional Method~ of Sensing Stress, Strain. and Damage in Materials and
Structures. ASTM STP 1318, George F. Lucas and David A. Stubbs, Eds., American
Society for Testing and Materials, Philadelphia 1997.
ABSTRACT: A full-field thermoelastic technique for measuring stress intensity

factors in dynamically loaded fatigue crack specimens is outlined. Thermoelastic Stress
Analysis (TSA) uses an infrared focal plane array detector and high-speed processing
electronics to produce stress images of cyclically loaded specimens. The TSA data is then
used in combination with a fitting algorithm to determine stress intensity factors for mode I
and mixed mode I and II crack geometries. The technique is non-contacting, fast and
accurate, and has the potential to greatly enhance the efficiency of fatigue and fracture
testing. An analytical solution for the complete stress state in terms of the first stress
invariant is derived in support of the full-field TSA techn,que. Results using the technique
on mode I and mixed mode specimens are presented, and show good agreement with
accepted values.
KEY WORDS: Stress, Thermoelastic, Crack, Stress Intensity Factor, Fracture,

Infrared, Thermal
i Stress Photonics Inc., 3002 Progress Rd., Madison WI 53716

2NASA LangleyResearchCenter, Mail Stop 231, HamptonVA 23681-0001
3The College of William and Mary, Applied Science Dept., WilliamsburgVA 23186
2O8
Copyright9
(PDVSA Los Teques) by ASTM
pursuant to International
License Agreement.www.astm.org
LESNIAK ET AL. ON STRESS INTENSITY MEASUREMENT 209
INTRODUCTION
There are many tools available that can monitor crack length in the laboratory including
microscopes, potential drop, and crack mouth opening techniques. These measurements
combined with analytical solutions, can be used to infer stress intensity factors and crack
lengths. However, these approaches may not be as applicable to real components as they
are to test coupons. This paper illustrates that Thermoelastic Stress Analysis (TSA) can be
as powerful a tool in materials laboratories as it has been consistently demonstrated to be in
component test facilities. TSA is a full-field measurement technique that shows great
promise for measuring stress intensity factors on arbitrary structures. This paper describes
recent progress in the measurement of mixed mode stress intensity factors of monolithic
materials. The results are compared with analytical solutions of predetermined mode I and
mixed mode I and II geometries. The solutions used in this application are intentionally
general and are intended to demonstrate the methodology of the damage assessment
system. Of course, no analysis is a perfect match for every situation. Each user of this
technology will have a unique set of concerns, but the layout of the thermoelastic damage
assessment system accommodates improvements as well as facilitates further evolution of
the technology into field applications, composite damage, test control, hostile environment
testing and more.
THERMOELASTIC PRIMER
Thermoelastic Stress Analysis (TSA) is a full-field, non-contacting stress-analysis

technique [1]. An infrared array camera is used to measure small, load-induced
temperature changes (AT) described by
off + AOy) (1)

AT = - pCp (AOx
where p is the density, ~ is the thermal expansion coefficient, Cp is the specific heat, T is
the absolute temperature and AOx and Aoy are the normal stress components. To improve
measurements, the temperature changes induced by the thermoelastic effect are repeated
and time-averaged with continuous dynamic loading [2]. Although sinusoidal loading is
most common, more complex load histories like those induced by traffic on a bridge can
be used. Figure 1 is a block diagram of a typical laboratory setup. A closed-loop hydraulic
load frame provides the loading while a TSA system correlates the load-induced infrared
signals with the reference signal from the load cell or strain gage. Because the cameras
employed in TSA measure the amplitude of the temperature oscillation, not just the
absolute temperature; they are referred to as differential thermographic cameras. Most
often the differential thermographic camera operates in a lock-in mode, providing in-phase
and out-of-phase images. When properly adjusted, the stress map resides in the in-phase
image and the out-of-phase image is null. All experimental results presented in this paper
were acquired with at TSA system similar to the one depicted in fig. 1.
Fig. 1 TSA system block diagram.
L I N E A R E L A S T I C STRESS F U N C T I O N S
In classical fracture mechanics, equations describing the stress singularity at the tip of a
crack are often derived. TSA has been used before to determine stress intensity factors
[3][4][5], but without considering higher order terms. A photoelastic technique that looked
at isochromatic fringes used a very similar over deterministic method to the one used in
this paper [6]. Because it is only the singular terms that drive failure or crack propagation
the arguments rarely yield a complete solution for the stress distribution away from the
crack tip. In order to measure stress intensity factors on arbitrary geometries using full
field data, a general solution to all potential stress distributions must be sought.
To accomplish this solution a stress function approach is taken. In review, a plane elastic
stress function in the presence of no body forces is a function that satisfies the biharmonic
equation. This relation is derived considering the two equations of equilibrium as well as
the stress compatibility equation. In polar coordinates the biharmonic equation is written
V4*(F,O) = ~-+7~'~r + r2 O02Jtcir 2 -I-7~-r -t7~J=O

(2)
where ~ is the stress function. Of course, the stress function does not describe the state of
stress directly, rather the exact state of stress can be derived via the following differential
operations
(~r =_.] ~..~+ 1 ~)2~

r 3r r 2 302 (3a)
a2~ (3b)
l a~ t b2,
1:0 r230 r 3r30 (3c)
It is important to note that the above defined stress distributions are valid for any plane-
elastic case including plane strain and plane stress. For the zero body force problem it is
not until strains must be calculated that conversion from one to the other is necessary.
Williams sununarized the derivation of the stress functions of cracks [7], but made several
typographical errors in his calculations. The process is revisited here for completeness
even though he corrected his mistakes in a later paper [8]. The generic stress function for a
notch with stress free edges is of the form
O(r,O,~,)= rX+'[c, sin(;L + 1)O+ c 2 cos(Z, + l)O + c 3 sin(~, -l)O + c 4 cos(;L -l)O]
(4)
Stress Pattern
Crack Tip
\" )
t /
\ /
J
Fig. 2 Fracture mechanics coordinate system
where 0 is defined in Fig. 2. The values of ~. are described by the eigenequation
sin(Xoc) = • (5)
where cx is the angle subtended from edge to edge, which for a closed crack, c~=2rc,yields
L=n/2 where n=1,2,3,4 ..... (6)
The generic stress function for a closed crack becomes
~)(r,O,n) = r n/2+l [c l sin(n/2 + 1)0 + c 2 cos(n/2 + 1)0 + C3 sin(n/2 - 1)0 + C4 cos(n/2 - l)O] (7)
The enforcement of boundary conditions at the free edges of the crack requires that,
~o=0 For 0 = 0,27t (8a)
%0=0 For 0=0,2~ (8b)
The stress function can then be written in terms of c 3 and c 4 as
~(r,O,n) = rn/2+'{c3Isin((n/2-1)0)- [n~22 ] sin((n/2 + l)O)] + c4[cos((n/2-1)0) - cos((n/2 + 1)0)]} (9)
To convert this to a more recognizable form a coordinate transformation can be performed

(Fig. 2) such that
~=0-~
For n=1,3,5 .... Let n ~ 2 n - 1 n=1,2,3 (10)

For n=2,4,6 .... Let n ~ 2 n n=1,2,3
Making these and the appropriate trigonometric substitutions and then separating the
resulting equation into even and odd functions, so that the stress function is written as
%(r,0,n) = X b2n-' (-l)n+' rn+'/2[sin( (n - 3/2)~)- sin((n + I/2)~F)]

n=l (lla)
+ ~ b2n(-1) n+l sin((n +,/~,/1
n-I L " L2n+zj " "j
%(r,O,n) = a2n-,(-1)n+nrn+'/2[cos((n " ),1

(lib)
+ E a2n(-1)n+l rn+l[c~ (n- 1)V)- c~ (n + I)~F)]
n=l
The even stress function relates to the symmetric mode I stress intensity distribution where
the odd function relates to the antisymmetric mode II stress intensity distribution.
Sparing the algebraic detail the first stress invariant (~r+~0) simplifies to
I(~=~__,(-1)n+l(4n-2)b2n_lrn-312sil+n4~(-l
n=l
(n-311J,
)n+lb2nnrn-lsin(n-1)~
n=l
(12a)
Notice that the sin(n+l/2)~t and the sin(n+l)~ terms completely cancel. A similar
derivation can be performed to determine the first stress invariant for the even stress
functions.
3 t~ + 4 Z (_l)n+la2nnrn_l cos(n _ 1)t~
I~ = Z(-1)n+'(4n - 2)a2n_lr n-3/2 cos n - ~- (12b)
n=l n=l
Redefining a and b to absorb all constants and defining the leading terms to be compatible
with the common definition of stress intensity factors, the complete stress state in terms of
the first stress invariant is
FITTING ALGORITHM
The analytical solution describes all the possible stress states that could occur within a zone
immediately surrounding a crack tip irregardless of the geometric details outside of the area
(Fig. 3).
Fig. 3 Even modes of analytical fits, K I, A o, n=l, 7
A fit window is established to designate the data to be used in the fit (Fig. 4). Although the
fit window is not necessarily circular or concentric about the crack tip a sufficient number
of stress terms will describe an arbitrary planer stress state about the crack tip. A boundary
collocation method is used by Gross and Srawley to determine the stress intensity factors
for specimens in bending or combined bending and tension [9]. The method is similar to
the technique described in this paper, except that this technique uses a full-field fitting
algorithm. An exclusion zone is established in the immediate area of the crack tip to
eliminate the use of high stress-gradient data which can be tainted by heat conduction or
nonlinear material behavior. A fully elastic body loaded at an adiabatic frequency can be
phased in so that all the oscillating thermal signal is perfectly synchronized to the loading.
The in-phase image resulting from the lock-in electronics represents the signed amplitude
of the thermal oscillations and the out-of-phase image should be null. Any material
experiencing nonlinear behavior or thermal conduction will show a deviation in phase;
therefore, the radius of the exclusion zone can be determined by examining the out-of-
phase image or the actual phase image. To understand the compromise that must be made
between algorithm application expediency, stress resolution and the generality of the model
stress distribution a closer examination of the fit algorithm is offered.
Fig 4. Data set with square fit window and circular exclusion zone
L E A S T SQUARES F I T T I N G
Once the appropriate fit window is established, a matrix form of multiple regression least
squares fitting is performed. The individual model stress distributions can be viewed as
elements of a basis that must span all potential stress distributions. A specific stress image
can be constructed as described in the following relationship:
KI
K,
ao
{S}=[TK~ TK. Ao AI B2 , . . A~ Bn. a]
- b 2
an
.bn (14)
Where the elements of the basis are defined in Eq. 13 so that TK~ and T~I~describe the
stress singularities associated with the stress intensity factors, A o is a uniform value at, and
A,, B, describe all other terms. In the context of these arguments, the two dimensional
data and stress terms are considered to be reorganized in one dimensional vectors.
Equation 14 can be written more simply as
{s}=[Q]{c} (15)
The least squares algorithm applied to Eq. 15 yields
[Q]T {S} = [Q]r [Q]{C} (16)
To simplify the least square algorithm the basis is orthogonalized and normalized over the
fit window so that
[Q']T[Q'] = I (17)
this simplifies the solution of Eq. 16 to
[Q']T{s} = {C'} (18)
The orthonormilization allows for rapid application and reapplication of the fit. The
orthogonalization process is ordered in such a manner so that the solution matrix C' yields
K I directly; however, a history of the orthogonalization must be maintained in order to
extract K n from the solutions vector.
STRESS AND TIP LOCATION RESOLUTION
The signal to noise ratio for stress intensity measurement is related to the root of the total
points used in the fit window. If a system resolution of 5mK in a 10 second scan is
possible and a fit window utilizing 400 data points is used then a final intensity resolution
of 0.25inK is possible. The crack tip location is optimized automatically by a search
routine that reapplies the fit in the vicinity of the user estimated crack tip. The tip search
algorithm is essential to crack tip following. The actual location of the tip can be
interpolated to a partial pixel size. For example, a scan area of 50mm x 50mm the crack tip
location is accurate to approximately 0.20mm. This resolution can be improved on as the
scan area is reduced. Scans as small as 12.8ram x 12.8mm are routinely captured.
RESULTS
Mode 1
A series of Mode I tests were performed to demonstrate the accuracy of the technology.
Pin-gripped specimens with crack length ratios of 0.225, 0.33 and 0.5 were tested (Fig.5).
Fig. 5 Mode I data sets
The specimens were fabricated from 1.6mm thick machine steel. The image acquisition
time was 60 seconds for these example images. This lengthy acquisition time is not
necessary to make accurate measurements for crack growth work. Stress images can be
acquired in just a few load cycles. Analytical solutions in terms of the normalized load for
this specimen are readily available and are plotted with measured values in Fig. 6. As a test
three different loads were used for the same geometry. The three measurements of the
dimensionless intensity factor Y do not deviate more than 3%.
Stress Intensity Factors

Normalized for Load
Kt : Y G ~ a
7.00
6.00
5.00
4.00
3.00
2.00
1.00
0.00 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::
6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6
alW
Fig. 6 Mode 1 stress intensity factors, analytical and measured
Mixed Mode
A mixed mode specimen was manufactured to demonstrate the abilities of TSA to separate
the ratio of mode I and mode II (Fig. 7). The analytical solutions for the stress intensity
factors versus measured results are described in Table 1.
Fig. 7 Mixed mode data
Table 1. Mixed mode stress intensity measurements
CONCLUSION
It has been demonstrated that TSA is capable of making accurate non-contacting mode I
stress intensity measurements. Limited data has been collected on mixed mode cracks;
improvements in this area are expected. Efforts are underway to apply this damage
evolution work to high-temperature situations. Plasticity and other material behaviors will
be considered further. Even though great effort can be taken to match analytical solutions
with experimental data, accounting for all nonlinearities and material properties, the
technique as it stands can be used to predict the likelihood of failure if the measurements
are consistent for a given material and thickness over a range of load conditions. Future
efforts of this work will test the damage evaluation system's ability to use fracture
mechanics arguments to predict actual failure. This will be the true test of the usefulness of
this technology.
REFERENCES
[ 1] Oliver, David E., Stress pattern Analysis by Thermal Emission, in Handbook on

Experimental Mechanics, edited by Albert S. Kobayashi, Pretentice Hall, Inc., Englewood
Cliffs, N.J., 1987, p. 610-618.
[2] Lesniak, J. R., and Boyce, B. R., A High-Speed Differential Thermographic

Camera, Proceedings of the Society of Experimental Mechanics Spring Conference,
Baltimore, 1994, p. 491-497.
[3] Stanley, P. and Chan, W. K., Mode H Crack Studies Using the SPATE Technique,
Proceedings of the Society of Experimental Mechanics Spring Conference, New Orleans,
1986, p. 916-923.
[4] Stanley, P, and Chan, W. K., Thermoelastic Stress Evaluation of a Proposed K2

Specimen, Proceedings of the 2nd International Conference on Stress Analysis by
Thermoelastic Techniques, London, SPIE Vol. 731, 1987, p. 102-109.
[5] Stanley, P. and Dulieu-Smith, J. M., Progress in the Thermoelastic Evaluation of

Mixed-Mode Stress Intensity Factors, Proceedings of the Society of Experimental
Mechanics Spring Conference, Dearborn, MI, 1993, p. 617-623.
[6] Sanford, Robert J. and Dally, James W., A General Method for Determining
Mixed-Mode Stress Intensity Factors from Isochromatic Fringe Patterns, Engineering
Fracture Mechanics, Vol. 11, 1979, p 621-633.
[7] Williams, M. L., On the Stress Distribution at the Base of a Stationary Crack,
Journal of Applied Mechanics, Vol. 24, No. 1, March 1957, p. 109-114.
[8] Williams, M. L., Bending Stress Distribution at the Base of a Stationary Crack,
Journal of Applied Mechanics, 1961, p. 78.
[9] Gross, B., Srawley, J, E., Stress Intensity Factors for Single-Edge-Notch
Specimens in Bending or Combined Bending and Tension by Boundary Collocation of a
Stress Function, NASA TN D-2603, January, 1965.
Koji Hyodo, t Masahiro Yamada,2 and Tetsuya TateishP
THERMOELASTIC STRESS ANALYSIS OF THE HUMAN TIBIA
REFERENCE: Hyodo, K., Yamada, M., and Tateishi, T., "Thermoelastic Stress
Analysis of The Human Tibia," Nontraditional Methods of Sensing Stress, Strain and
Damage in Materials and $Iroqtorr ASTM $TP 1318, George F. Lucas and David A.
ABSTRACT: The thermoelastic stress analysis method is a non-contact technique

which utilizes the thermoelastic qualities of hard tissues such as bones. It enables easy
measurement and imaging of surface stress distributions. Using this approach, we
performed stress analyses of human tibias in simulated loading conditions at various
modes ofvarus valgus loading on the knee joint. In normal loading, the greatest stress
image was obtained on the posterior aspect, indicating that the majority of the stress was
concentrated in the posterior part of the tibia. The epiphysis of the proximal tibia, which
consists predominantly of cancellous bone, showed tensile stress and the diaphysis which
consists of compact bone showed compressive stress in every plane. These results
indicate that the tensile stress in the epiphyseal surface indirectly reflects the impact-
absorbing properties of cancellous bone. Also, results demonstrated that small changes in
the alignment of the knee joint have a great influence on the stress pattem of the proximal
tibia.
KEYWORDS: thermoelastic stress analysis, biomechanics, tibia, knee joint
One of the major factors concerned with knee osteo-arthritis is the imbalance of
the biomechanical condition of the joint. It is important to analyze stress distribution of
proximal part of the tibia to understand the biomechanical background &the disease
~SeniorResearcher, Mechanical Engineering Laboratory, Agency of Industrial

Science and Technology, Ministry of International Trade and Industry, Namiki 1-2,
Tsukuba, 305 Japan.
2Orthopedist, Kobe University School of Medicine, Kusunoki 7-5-1, Chuo,
Kobe, 650 Japan.
3Director &Bionic Design Group, National Institute for Interdisciplinary
Research, Agency of Industrial Science and Technology, Ministry of International Trade
and Industry, Higashi 1-1-4, Tsukuba, 305 Japan.
221
Copyright9
(PDVSA by ASTM
and for evaluation of efficacy of medical treatment. Even though, there are only a few
reports which describe the 3 dimensional stress analysis of the proximal tibia. Finlay and
Bourne [_4]reported in vitro analysis of tibia by using strain gauge methods and Little [10]
used Finite Element Method for 3 dimensional analysis.
The thermoelastic stress analysis method utilizes the thermoelastic properties
of materials. Compared with former methods, such as strain gauge methods, photoelastic
modeling or laser holography methods, it enables easy measurement and imaging of the
distribution of the sum of principal surface stresses. It is a non-contact, whole stress field,
analysis method, and is free from some of the difficulties inherent in other methods. For
example, it requires no strain gauges, no two-dimensional plastic modeling, and no precise
optical systems for making holograms. We applied this method to perform experimental 3
dimensional stress analysis of the proximal tibia.
EXPERIMENTAL METHOD
Thermoelasticity
The thermoelastic effect is the change in temperature that accompanies the

adiabatic elastic deformation of a body. It was firstly described analytically by Thomson
(Lord Kelvin) [16] and further developed by Biot [_2].Recently, basic research for imaging
was performed [1] [4] [15], and thanks to the development of measurement technology,
high-precision imaging is now possible [7"-9] [14] [17--~18].
The thermoelasticity is govemed by the following relations.
A T=-kT A a (1)
where T is the absolute temperature, k, the thermoelastic constant, and e, the volumetric
strain. The thermoelastic constant, k, is given by
k=KT/3 V / C v (2)
where KT is the isothermal bulk modulus, /3, the coefficient of cubic thermal expansion,
V, the molar volume, and Cv, the molar heat at constant volume. The KT,/3 and A e
are given b y
A a=((1-2 V) / E T ) " A(O 1+O 2) (3)

Kv=ET/(3(1-2 v )) (4)
t~ =3 o~ (5)
HYODO ET AL. ON TSA OF THE HUMAN TIBIA 223
where a , is the principal stress, v, is Poisson's ratio, ET, is Young's modulus at

constant temperature and ~, is the coefficient of linear thermal expansion. Typical
values for v, ET and ct for cortical bone are 0.242 (the radial direction), 26.0 GPa (the
axial direction) and 0.27• 10"4/K [3][18].
From Equation (1), using Equations (2) to (5), the following equation can be
derived.
A T=-k ' T A ( o l + a 2) (6)
where k" , is given by

k" = a V / C v ~ a/(o Cv" ) (7)
O, is density, Cv" , is the specific heat at constant volume. Typical values for 0 and
Cv" of cortical bone are 1920 kg/m3and 1300 J/kg-K [3][18].
Vanderby and Kohles [18] thermographically measured heat flux from loaded
specimens of cortical bone and correlated the results with strain gage data. The correlation
coefficients demonstrated a significant linear relationship between thermal data and
measured and computed stress, strain data. The local temperature changes measured from
the emitted infrared radiation provide detailed information on the associated stresses.
Test Procedorr
Material,
Five fresh human cadaveric legs were used. These specimens were obtained
following above knee amputations performed for peripheral vascular diseases (male 4,
female 1, average age 62 years old (51 "~65years)). These specimens were frozen at -20~
until used for the experiments. No significant damage &the specimens was found by X-
ray inspection. The femur was prepared for approximately 70 mm length above the knee
joint and the foot was removed at the level of the ankle joint. All extraneous soft tissues,
other than the extensor mechanism, capsular, ligamentous structures, and fibula were
removed to expose bony structure surface. The surfaces of the specimens were polished
with sandpaper (grid #100), and then painted frosted black with aerosol lacquer paint
(Kanpehapio Co., Japan) in order to suppress heat reflection. The proximal part of the
femur and the distal part of tibia were fixed in the custom-made device outlined below.
These specimens were mounted in the load testing device which was similar to a
device reported by Finlay and Bourne et al.[5]. Four stainless steel plates and three metal
balls were used to simulate varus-valgus loading of the leg. Plate 1 and plate 2 were fixed
to the oscillator and the femur and the tibia were fixed to plate 2, 3. Two 9 mm diameter
metal balls connected plates 1 and 2 and a 20 mm diameter metal ball connected plates 3
and 4. (FIG. 1). In this test setup, the 9 mm diameter metal balls connected plates 1 and 2
provided for rotation of the femur only in the coronal plane, while the 20 mm diameter
metal ball connected plates 3 and 4 provided for axial rotation and rotation in the sugittal-
coronal plane of the tibia. Therefore, the knee joint was fixed in a maximum extension
position with free varus-valgus movement. By changing the position of the two metal
balls which connects plates 1 and 2, 0 ~ ,2.5 ~ and 5 ~ varus-valgus loading conditions
could be simulated.
FIG. 1-- The load testing device for the femur and tibia
Method
The measurement apparatus consisted of two units, the first being the oscillator
unit (Tokyo Shikenki) and the second the thermoelastic stress analysis unit (Fujitsu,
Tokyo) (FIG. 2). The oscillator unit applied 1-Hz sinusoidal loads to the materials. As
the maximum value of the knee joint force during level walking was 2.1 to 4.0 times body
weight [13], and to get high thermoelastic Signal/Noise ratio, 2.8 kN (max 2.9 kN, rain
0.1kN) of oscillatory amplitude was applied. Thermal radiation from the surface of the
materials was detected with an infrared camera (Fujitu, Tokyo; HgCdTe infrared
semiconductor sensor which has 10 tt m wavelength detection peak) with 60-Hz scanning.
A 0.01 ~ was detectable when integrated 16 times. Data collection was synchronized
with the load cycle to obtain the temperature difference.
The data was converted to relative stress changes (A ( o ]+ cr 2)) and imaged by
224 dots • 224 lines by the tbermoelastic stress analysis unit. A number of 224 load
cycles were required to get one stress image by line scanning, The stress imaging was
displayed on a TV monitor, with stress levels represented by 50 gray scales (+/- 0.25K,
approximately 3.16 MPa for 1 scale or 0.01K [3][18]). The stress data was stored on
computer (Fujitsu, Tokyo).
The thermoelastic stress image in the proximal tibia was measured in the
anterior-posterior and medial-lateral plane by using the measurement apparatus.
FIG. 2-Thermoelastic stress measurement apparatus
RESULTS
Thermoelastic stress images were demonstrated in the medial, lateral, and

posterior aspects of the tibia. These results were recorded after three pre-imagings (672
load cycles) to stabilize the experimental conditions. To complete the experiment (anterior,
posterior, medial and lateral stress images with normal, varus or valgus loading ), it
required at least 4480 load cycles to the specimens. There was no significant difference
between stress images at different cycles.
In normal loading, little stress image was observed on the anterior side (FIG. 3-
1). The greatest stress image was obtained in the posterior aspect, indicating that under
these testing conditions, the major stress was concentrated on the posterior surface of the
tibia. The bony surface of the epiphysis and the metaphysis in the tibia, consisting
predominantly of cancellous bone, showed tensile stress, and that of the diaphysis, which
consists of compact bone, showed compressive stresses in every plane.
In varus loading, with a simulated 2.5 ~ angle between the tibia and femur, the
compressive stresses in the medial aspect of the diaphysis increased, while the
compressive stresses in the lateral aspect decreased, indicating that little stress existed on
this side (FIG. 3-2, gray lines in Fig. 4). In this testing condition, the high stress image had
shifted medially and the medial and posterior aspects of the tibia showed similar stress
concentration. In 5 ~ varus loading (FIG. 3-3, dashed lines in Fig. 4), the compressive
stress was further increased in the medial aspect of the diaphysis, while tensile stress was
demonstrated in the lateral aspect of the tibia. ( In 5 ~ varus loading, anterior and posterial
stress distributions were undetectable because of M-L movement of the specimens during
sinusoidal loading). These results indicate that the tibia is loaded equally between the
medial and the lateral condyle and that the load is concentrated posteriorly in normal
alignment. In moderate varus loading, with 2.5 ~ valgus deformity, the major
compressive stress had shifted to the posteromedial aspect of the tibia. In marked varus
loading, with 5 ~ varus deformity, the load transmission had shifted to the medial aspect.
Valgus loading showed that, with 2.5 ~ valgus deformity, the load was
concentrated on the posterolateral aspect of the diaphysis of the tibia, while with 5 ~
valgus deformity, the load was concentrated almost entirely on the lateral aspect. These
results were similar to those observed in varus loading, except that the medial and lateral
aspects were reversed.
These results demonstrate that small changes in the alignment of the knee joint
have great influence on the stress pattern of the proximal tibia.
DISCUSSION
Stress distribution in the proximal tibia
In the bony surface of the epiphysis and the metaphysis part of the tibia,
consisting predominantly of cancellous bone, a tensile stress distribution was found,
whereas in the diapysis, consisting of compact bone, compressive stress was found in all
planes. The greatest stress image was found in the posterior aspect. Although FEM
results by Little [11] show larger compressive stresses in the proximal tibia than in this
experiment, the overall stress distribution patterns are very similar.
Anterior Posterior Medial Lateral
3 - 2 2.5 ~ varus
FIG. 3 -- Thermoelastic stress image of the tibia

Stress levels represented by 50 gray scales (+/- 0.25K, approximately
3.16 MPa for 0.01K)
Anl Post Med Lat

)
'7 ~ "t
"4
I ! I I I " I
(AT measurement line) -0.16 0 +0.16 -0.16 0 +0.16 -0.16 0 40.16 -0.16 0 +0.16
AT (K)
FIG. 4 -- A T of the tibia at various aspects
black lines: normal alignment, gray lines: 2.5 ~ varus, dashed lines :
5 ~ varus (approximately +0.16K = 50.6 GPa (compression))
Cancellous bone has porous structure and is filled with marrow fluid. The
elastic modulus of cancellous bone is 10"--102 times smaller than that of compact bone
[3]~]. The metaphysis of the tibia (cancellous bone surrounded by thin compact bone)
was supported with the diapysis (compact bone), and covered with cartilaginous plate.
These conditions cause collapse of the cancellous bone area when loaded. That is why
tensile stress was observed around the metaphysis in the tibia. This structure plays very
important role in shock absorbing (FIG. 5).
: t~'r~reulve
........ : t'~litle
FIG. 5 -- Shock absorbing deformation of proximal tibia
Stress distribution and varus-valuus andes
Maquet [12] reported that 15 ~ varus deformation of the knee joint raises the
condyle stress by about 4 times as much as in normal position. It occurs also in valgus
deformity. Kettelkamp et al. I[LQ]calculated tibial condyle loads in the bipedal standing
state. The load became almost same as body weight in lateral condyle when small valgus
deformity, and in medial condyle when small varus deformity. Finlay ~]concluded that
3.7 ~ ofvarus deformity raises the strain of the medial surface of the tibia by about 2-',-3
times compared to the normal position and reduces the compressive strain of lateral
surface of the tibia to zero. Our thermoelastic stress images demonstrated that medial
compressive stress increased about 2 times with a 5~ varus deformation, and tensile
stress was found in the lateral surface of the tibia.
These results show that small changes in the alignment of the knee joint have a
great influence on the stress pattern of the proximal tibia. One of the major factors
concerned with knee osteo-arthritis is an imbalance of the biomechanical condition of the
joint. The alignment, varus valgus angle, is one of the most important factors in this
disease.
CONCLUSION
The thermoelastic stress analysis method was very effective for performing
stress analysis of the tibia. It was shown that the tensile stress in the epiphyseal surface
indirectly reflects the impact-absorbing properties of cancellous bone, and small changes
of alignment in the knee joint have a great influence on the stress pattern &the proximal
tibia.
The thermoelastic stress analysis method detects the sum &the main surface
stress ( a 1+ cr 2). Combining this with photoelastic stress analysis, which detects( cr 1-
0 2), would produce interesting results.
REFERENCES
[!] Baker, L.R. and Webber, J. M. B., "Thermoelastic stress analysis," Optica acta,
Vol. 20, NoA, 1982, pp 555-563.
[2] Biot, M.A., "Thermoelasticity and irreversible thermodynamics," Journal of

Applied Physics, Vol.27, No.3, t956, pp 240-253.
[3] Duck, F. A., Physical Properties of Tissue, Academic Press., 1993
[_4] Duncan, J. L. and Cummings, WM., "Thermoelastic stress analysis of fresh

bone," Abstracts of the first international conference of stress analysis by
thermoelastic techniques, Sira, London, Section 10, 1984, pp 1-5.
[5-] Finley, J. B., Bourne, R. B. and McLean, J., "A technique for the in vitro
measurement of principal strains in the human tibia," Journal of Biomechanics,
No. 15, 1982, pp 723-739.
[6] Gibson, L. J., "The mechanical behavior of cancellous bone," Journal of

Biomechanics, No. 18, 1985, pp 317-328.
[_7] Hyodo, K., Shirono, T., Lee, I.K. and Tateishi, T., "Consideration in
Thermoelastic stress analysis method for dental biomechanics," Journal of
Mechanical En~neeringLaboratory, Vol. 44, No.5, 1990, pp 173-180.
[8_] Hyodo, K. and Tateishi, T., "Application ofthermoelastie stress analysis

method to joint biomechanics,"Hip Biomeehanics, Spfinger-Vedag, 1993, p p
277-285.
[-9.] Hyodo, K. and Tateishi, T., "Application of thermoelastic stress analysis

method to orthopedic biomechanics," Medical!ma~in~Technolo~, Vol. 13,
No. 5, 1995, pp 716-721.
Kettelkamp, D. B. and Chao, E Y., "A method for quantitative analysis of

medial and lateral compression forces at the knee during standing," Clinical
Orthop., No. 83, 1972, pp 202-213.
[11] Little, R.B., Wevers, H.W., Siu, D., et at., "A three-dimensionalfinite element
analysis of the upper tibia," ASME JoBrnBl9f BiQmcchanicalEn~neering. No.
108, 1986, ppl 11-119.
Maquet, P. G., Biomechanicsof the knee, Springer-Vedag, 1976
113~] Morrison, J. B., "The mechanicsof the knee joint in relation to normal
walking," JournalofBiomechanics, Vol. 3, 1970, pp 51-61
[.t~.] Okano, M, Tateishi, T. and Hyodo, K., "Studies on Stress Distribution under
SimulatedMuscles on Dry SkullUtilizing Thermoelastic Stress Imaging
Method," Journal of Japane~;eProsthodontic Society.,No. 37, 1993, pp 887-
895
rL~] Reifsnider, K.L., Henneke, E.G. and Stinchcomb,W.W., "The mechanicsof

vibrothermography," Mechanic~of Nondestructive Testing, Plenum Press,
1980, pp249-276.
]Lt~] Thomson, W., "On the dynamicaltheory of heat, with numericalresults

deduced from Mr. Joule's equivalent of a thermal unit, and Mr. Regnault's
observations on steam," Transactions &the Royal Society of Edinburgh, Vol.
20, 1853, pp 261- 288.
U-/] Tateishi, T., Hyodo, K, Homma,K. and Yamada,M., "Visualization

methods in biomechanics," HeimkeG, SolteszU, Lee AJC (eds) Clinical
implant materials. Advan~:r in biomaterials,Elsevier, vol 9, 1990, pp651-656.
[.L~] Vanderby, Jr.R. and Kohles, S.S., "Thermoelastic stress analysis in cortical
bone," ASME Journal of BiomechanicalEn~neering, No. 113, 1991,
pp 418-422.

STP1318 Eb.1415051 1 PDF

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

STP1318 Eb.1415051 1 PDF

Uploaded by

Copyright:

Available Formats

Copyright by ASTM Int'l (all rights reserved); Sat Jan 28 08:00:22 EST 2012

Nontraditional Methods of Sensing

George F. Lucas and David A. Stubbs, Editors

ASTM Publication Code Number (PCN):

IITI'" 100 Barr Harbor Drive

Copyright 9 1997 AMERICAN SOCIETY FOR TESTING AND MATERIALS, West

Peer Review Policy

Use of an Optic Lever as a Short Gage Length Extensometer--G. M. HOUSER

New Applications of the Interferometric Strain/Displacement Gage--

An Extensometer for Soft Material Strain Measurement--N. SHRIVE,

NONTRADITIONAL CRACK MEASUREMENT TECHNIQUES

Short Fatigue Crack Characterization and Detection Using Confocal

An Advanced Multiple Frequency ACPD System for Crack Detection and

Observation on the Growth Behavior of Small Surface Cracks Using Remote

Use of Portable/In-Situ Stress-Strain Microprobe System to Measure Stress-

OPTICAL, NoN-CONTACTING STRAIN MEASUREMENT DEVICES

BISMAP: A Non-Contact Biaxial Displacement Measurement System--

Optical Measurement of Biaxial Strain in Thin Film Polymers--J. L. RAND AND

A Review of Strain Measurement Techniques Using the Grid Method--

New Developments in Full Field Strain Measurements Using Speckles--

Ultrasonic Longitudinal and Surface Wave Methods for In Situ Monitoring of

Using Ultrasonic Measurements and a Two-Phase Composite Model to Assess

An Array Measurement System for Thermoelastic Stress Analysis--c. WELCH,

Stress Intensity Measurement via Infrared Focal Plane ArraynJ. LESNIAK,B.

Thermoelastic Stress Analysis of the Human TibianK. HYODO,M. YAMADA,AND

The understanding of how materials and structures behave under conditions

This Special Technical Publication is based on the symposium on Non-

USE OF AN OPTIC LEVER AS A SHORT GAGE LENGTH EXTENSOMETER

REFERENCE: Houser, G. M. and Walrath, D. E., "Use of an Optic Lever as a Short

KEYWORDS: fiber optic, strain measurement, optic lever

IDirector of Research and Development, Model and Instrument Development,

2Associate Professor, Mechanical Engineering Department, PO Box 3295,

One such strain transducer design is an extensometer, an adaptation of the foil

FIBER OPTIC LEVER

Figure 1. Operating Principle of the Fiber Optic Lever.

Figure 2. Conceptual Model of a Fiber Optic Lever Displacement Measurement System.

SEVEN FIBER OPTIC LEVER MODEL

Figure 4. Reflected IUuminance Radius at the Fiber Plane

Figure 5. Illuminated Receiving Fiber Area (~/2 < 7d2)

where xo is related to the fiber center spacing as

defining the reflected iiluminance ratio such that

as long as W/2 is an acute angle. If ~P/2 is obtuse,

A, = 6p'I)c: ~ + , - sin-'0: sin~l - ~:xo sin~l,

Figure 6. Illuminated Area Ratio as a Function of Target Displacement for a Seven-Fiber

Three important regions are defined:

where E0 is the reflected illuminance at the fiber plane,

Regression analysis resulted in the cubic representation of

d = 0.11295 + 4.5191x10 "2V + 5.5324x10"3V2 + 4.5065x10"4V3

'-"0"--- F.O. Lever jO

0.00 ,I. ~ , . I . . . . I . . . . I 9 ' ' ' I 9 9 o ,

0.0 0.5 1.0 1.5 2.0 2.5

Figure 10. Comparison of Measured Strains as Functions of Cross-head Displacement for

Calculated Strain (10 4)

Overall, the objectives of the development effort were attained, including

Development of a seven-fiber optic lever design model,

Determination of important optic fiber parameters including their relevance to optic

Construction of a prototype short gage length extensometer for strain measurement,

Preliminary evaluation of the prototype transducer including comparison to strain

[8] Hecht, Eugene, Optics, Addison-Wesley,Reading, MA, 1988, pp. 79.

NEW APPLICATIONS OF THE INTERFEROMETRIC

ABSTRACT: The Interferometric Strain/Displacement Gage (ISDG) is an established

KEYWORDS: interferometry, lasers, microspecimens, polysilicon, thin films,

The Interferometric Strain/Displacement Gage, or ISDG, is a non-contacting optical